The Pandemic Election

An Analysis of Five Key Dimensions of the Information Landscape in the 2020 U.S. General Election

Lead Author

Anna Sakellariadis, D3P Fellow

 

Section Authors

Jaim Coddington, D3P Harvard Kennedy School
Yousef Ibreak, D3P Harvard Kennedy School
Amanda Powers, D3P Harvard University
Yi-Ting Lien, D3P Harvard Kennedy School
Chris Stewart, D3P Harvard Kennedy School

 

Contributors

Eric Rosenbach, Co-Director, Belfer Center; Director, D3P
Maria Barsallo Lynch, Executive Director, D3P
Robby Mook, Co-Founder, Senior Advisor, Senior Fellow, D3P

Gabe Cederberg, D3P Harvard University
Danya Chudacoff, D3P Harvard Kennedy School
Amina Edwards, D3P Fellow
Stefani Jones, D3P Harvard Kennedy School
Kelly Mallon, D3P Harvard Kennedy School
Aleksandra Srdanovic, D3P Harvard University
David Stansbury, D3P Harvard Kennedy School

 

A Note to Readers

The Belfer Center has a dual mission: (1) to provide leadership in advancing policy-relevant knowledge about the most important challenges of international security; and (2) to prepare future generations of leaders for these arenas. 

Since July 2017, the Belfer Center’s bipartisan Defending Digital Democracy Project (D3P) has worked alongside election administrators, national security experts, political and civic thought leaders to develop and deliver practical strategies, tools, and recommendations to protect democratic processes and systems from cyber and information attacks.

These efforts have taken the form of national training events for state and local election officials in person and online, tabletop exercises, playbooks for countering and responding to cybersecurity attacks and information operations, incident communication response plan among other resources. The project has also provided recommendations through congressional testimony and reports. The D3P team includes talented students, alumni, fellows and Senior Advisors. Students are vital to the work of the project and are central to the purpose of the Belfer Center’s work. 

Ahead of the 2020 Election, a team of D3P fellows and students developed a National Election Data Set. In a quickly evolving information environment affecting election and communication processes, these changes created vulnerabilities for potential mis and disinformation. The team’s goal was to compile relevant state-state specific data, some of which was changing in real time, and provide analysis as a resource for better understanding election processes during an unprecedented time as the nation navigated the COVID-19 pandemic. The data set was shared as a potential resource to media, officials, public leaders and voters. 

The team’s curiosity and creative thinking sparked their interest in collecting additional data detailed in this report. This report shares analysis from a unique election cycle and provides relevant insights for future elections in an evolving information and threat landscape

 

Eric Rosenbach, Co-Director, Belfer Center; Director, D3P

Maria Barsallo Lynch, Executive Director, D3P

 

---

Executive Summary

The 2020 general election was unlike any other experienced by the American public. The COVID-19 pandemic impacted every aspect of the election system, and created an urgent need to adapt and explain new election administration processes to the public. Unavoidable changes to the tempo of Election Night, ballot tabulation timelines, and reporting of election results created major departures from public expectations. This divergence between expectations and reality posed a serious threat to public trust in the integrity and consistency of the electoral system. Complicating this was the fact that each U.S. state sets its own election process and pandemic response measures, so any public awareness efforts would have to convey the differences among 50 different election administration systems. Further, since 2016, the increasingly decentralized information ecosystem across social and traditional media platforms, as well as the documented influence of foreign and domestic actors in spreading disinformation, have become ever more important factors mediating the level of public trust in U.S. elections. Taken together, these factors created an information threat landscape with fertile conditions for election-related mis and disinformation to take root.

Mis and disinformation thrive by capturing imaginations through obfuscation, oversimplification, and speculation. In contrast, this report documents and analyzes five key areas related to the shifting information threat landscape in the 2020 U.S. general election with clarity, in their complexity, and on the basis of rigorously collected data. Our goal is to provide relevant insights for future elections in an evolving information and threat landscape.

We find that:

• The near doubling of mail-in voting across the U.S. from the 2016 to 2020 general election can only comprehensively be understood on a state-by-state basis. While some states saw over 80% of their voters cast mail ballots, others saw less than 10% use mail ballots; while some states saw the use of mail-in voting more than double, others saw small decreases in the use of mail-in voting compared to 2016. Demystifying what drove these variations and changes, and quantifying by how much voting patterns actually shifted in each state can help reduce confusion around the increase in mail-in voting throughout the U.S. in 2020. Both the significant demand for mail voting evidenced in 2020 and the positive impacts of greater voting flexibility on turnout should be taken into account in planning for future elections.
• To counter and respond to the information threat in the 2020 election, a variety of stakeholders across civil society, government, and the private sector launched broad and sometimes overlapping campaigns to blunt the impact of misinformation and disinformation. We term these endeavors to address persistent and pervasive information threats as “Election Defense.” As long as mis and disinformation continue threaten public trust in elections, these multilateral and proactive Election Defense measures will likely be a necessary feature of efforts to protect the bedrock of American democracy. 
• Disparities between prevailing narratives on traditional and social media emerged in the first 48 hours after Election Day. Topline news coverage across 26 national and regional traditional news outlets in this period proved relatively consistent in conveying confidence that election processes were following the usual course for accurately tallying votes. In contrast, groundless narratives of election fraud went viral across Twitter, Facebook, and Google in this same period. These differences highlight how important the 48 hours after Election Day were in the seeding of unsubstantiated claims of widespread election fraud.
• The 2020 election saw an unprecedented amount of election-related litigation, particularly in the post-election period. The pattern of remedies sought and states targeted in post election cases indicated that a politically expedient strategy to secure outcomes favorable to then-President Donald J. Trump was the motivating force behind this litigation, not a desire to root out election fraud. While these cases overwhelmingly failed to win in the courts, their mere existence may have served to legitimize baseless election fraud claims in the public square.
• During the 2020 U.S. general election, the People’s Republic of China (PRC) mounted overt influence operations through public statements from state-backed organizations, largely targeting a domestic Chinese audience. Understanding the messaging employed remains relevant to Americans, as the PRC has increasingly sought to broadly promote the superiority of the “China option” to the U.S.’s liberal democratic model. An analysis of these narratives on Chinese social and traditional media reveals an emphasis on portraying the U.S. model as a failing one by characterizing the 2020 election as chaotic, disorderly, and polarized.

 

Background: How the COVID-19 pandemic presented new, fundamental problems for the 2020 Election

When the World Health Organization determined that the COVID-19 outbreak could be characterized as a pandemic on March 11th, 2020, the United States was partway into a presidential election year. Twenty-five states still needed to conduct their primary elections, and all states and territories were preparing for the general election in November. Public health concerns called for limiting the transmission of the virus by increasing social distancing amongst the population. But most states largely relied on in-person voting procedures that required the congregation groups of large groups in polling locations on the same day. Consequently, COVID-19 would present an additional challenge for U.S. election administration: how to maintain the integrity of the election from known, persistent threats while safely and securely conducting the election in the face of new risks posed by the pandemic.

Election officials became the frontline workers responding to meet this extraordinary operational challenge. Anticipating a significant uptake in mail-in voting where existing state laws allowed, officials worked to publicize those voting options and prepared to administer a greater number of mail ballots. Further, while state legislatures and courts shaped the relative expansions in voting flexibility across states—primarily in terms of access to early or mail-in voting—election officials faced the task of implementing new policies in record time. As a result of these efforts, the number of mail ballots cast nearly doubled from 2016 to 2020¹.

For states projected to see significant increases in mail-in voting for the first time, it was possible that they would not be able to post a complete tabulation of results by the evening of the end of voting on November 3rd, Election Day². This possibility would become increasingly likely if both of the following conditions were met: 1.) that the election outcome was extremely close, such that the remaining number of mail ballots to be tabulated exceed the margin of victory and therefore could determine the winner, and, 2.) that a state’s laws limited the pre-processing of mail-in ballots before November 3rd, such that election officials would only be able to begin tasks such as removing ballots from envelopes, or verifying signatures, so late in the voting period that tabulation would go on beyond the evening of November 3rd.

Such delays in the tabulation of votes would cause a subsequent delay in results reporting and, in turn, the projection of an unofficial winner—though it would not, importantly, affect the weeks-long processes through which states have always certified official election results³. But, with Americans accustomed to knowing the projected winner of the U.S. election by Election Night, the abnormality of a delay in projected results could be exploited to further cast doubt on the integrity and security of the election. This was particularly a concern given that disinformation meant to intentionally confuse the public about elements of the election process was already occurring months before November 3rd⁴. Indeed, months before the general election, then President Donald J. Trump himself repeatedly questioned mail-in voting⁵ without providing any substantive evidence⁶, seeding concerns about election integrity and security that could be exploited by any apparent delay in the projected results.

In fact, mail-in voting has a long and successful history in the United States. It was first widely used amongst the military during the Civil War as a way to ensure that soldiers could vote “absentee,” and was subsequently expanded to allow civilians to vote absentee as well⁷.” By 2020, before states implemented any changes in response to the COVID-19 pandemic, five states conducted their elections entirely by mail, 29 offered no-excuse absentee voting, and 16 offered absentee voting provided voters had a certain excuses⁸. While many states increased access to mail voting in light of the pandemic, mail-in voting procedures to protect the integrity and security of the results had already been well established and proven to be effective at preventing widespread fraud⁹. The challenge for election officials was not in building these procedures from scratch, but in shifting election operations to meet relatively greater demand for mail voting from the electorate.

Though perhaps the most memorable moment of this election season occurred on January 6th, 2021, when the U.S. Capitol Building was overrun by a mob coming from a rally in support of then President Trump, this report centers on earlier periods in the election cycle. We focus on the pre-election period before Election Day, the interim period after Election Day before news outlets projected a winner on Saturday, November 7th, 2020, and the vote tabulation period that culminated with the certification of official results and selection of electors on December 14th, 2020.

 

Section 1: State-by-state analysis of changing vote-by-mail patterns in the 2020 election

Section Authors: Anna Sakellariadis and Chris Stewart

Key Data Points

➠ States with the greatest 2020 vote-by-mail (VBM) rates (generally above 55%) were those that offered the greatest access to VBM by using entirely by-mail voting systems.

➠ Conversely, states that strictly limited access to VBM saw the lowest VBM rates (less than 17%).

➠ Among states that neither used a entirely mail-in voting system, nor strictly limited access to VBM, 2020 VBM rates ranged widely; this suggests that other policies, election administration practices, or factors across these states influenced preferences for voting method.

➠ Of the ten states with the largest gross increases in VBM rates, eight of them automatically either mailed a ballot or a ballot application to every voter.

➠ States that made major policy changes for the 2020 general election saw a much greater mean percentage increase in vote-by-mail rates, versus those states that did not make major policy changes.

➠ Of the states that did not implement policy changes, even those with the most restrictive VBM policies saw a notable percentage increases in their mean VBM rate in 2020. On the other hand, the states that continued expansive VBM policies from prior elections saw nominal percentage increase in their mean VBM rate.

➠ Where states maintained in-person voting as the de-facto voting method but allowed voters to choose how to vote—either through policy changes that relaxed absentee voting excuse requirements or by maintaining policies that did not require an excuse—many more voters chose to vote by mail than did in 2016.

 

Overview

Only 25 states had conducted their contest¹⁰ for the U.S. presidential primary elections by March 11th, 2020, when the World Health Organization determined¹¹ that the COVID-19 outbreak could be characterized as a pandemic. For the remainder, the Center for Disease Control (CDC) recommendations¹² to limit the transmission of the virus by practicing social distancing presented new operational challenges: prior to 2020, most states largely relied on conducting in-person voting operations on a single “election day,” a method that required large numbers of voters to gather closely in polling locations to cast their votes.

To minimize crowding at polling locations on a single day due to pandemic-related health concerns and to give voters greater choice in how to cast their ballot given new safety considerations, states began to expand access to mail-in and early voting¹³. Whether through new state laws, court decisions, or decisions made at the discretion of election officials, states made changes to rules surrounding voter registration, in-person early voting, absentee ballot eligibility, and convenience reforms such as ballot drop-boxes or extended polling place hours. While these changes occurred at the state and local levels, we will examine only major, state-level changes, and we focus on changes to access to mail-in voting.

Any large, sudden shift in voting methods and election operations increases the relative risk that mis or disinformation might undermine public trust in election outcomes. Despite the secure and successful history¹⁴ of mail-in voting across the U.S., the newness of such changes creates the potential for confusion that complicates the information threat landscape. In response to the COVID-19 pandemic, the changes to expand state capacity to administer mail ballots either under existing policies or under new policies drawn from best practices in other states were unique in the rapidity of their application and the scale of their impact. The fact that each state had its own set of existing election laws and responded differently to the pandemic further complicated this picture, as it is only possible to understand the effects of these changes on a state-by-state basis. Quantifying how voting methods and election administration shifted in 2020 in each state is, thus, a key initial step in understanding the scope of the information threat landscape in this election cycle.

 

Data and Analysis

The analysis in this section draws 2016 and 2020 vote-by-mail (VBM) rates, as well as information on state policies governing VBM access, from D3P’s National Election Dataset. The Dataset collated information from a variety of sources. Specifically, data regarding 2016 VBM rates came from the 2016 Election Administration and Voting Survey (EAVS) reporting the official election results, while data regarding 2020 VBM paratestterns came from the U.S. Elections Project’s summary of unofficial reports from state election officials of VBM ballot returns¹⁵. Finally, Information regarding state policies around VBM came from the National Conference of State Legislatures and the Brookings Institution.

 

Vote-by-Mail Utilization in 2020 and Change from 2016: State by State Breakdown

State policy governing ease of access to vote-by-mail (VBM) was a major driver of voting patterns in 2020. Table 1 groups states according to four distinct types of policies governing access to mail ballots, namely:

1. States that mailed a ballot to every registered voter
2. States that mailed an absentee ballot application to every registered voter
3. States that either did not require an excuse to vote absentee or only required a flexile excuse to vote absentee
4. States that limited eligibility to vote by mail

Across these categories, Table 1 documents the absolute number of mail ballots returned in 2016 and 2020, the VBM utilization rate as a percentage of voter turnout in 2016 to 2020, and the gross change in VBM rates from 2016 to 2020 in each state.

 

Table 1: Vote by mail rates 2016-2020 by by 2020 method of accessing mail ballots requesting
by State Policy Governing How to Access Mail Ballots in 2020

 

➠ Key Point: States with the greatest 2020 VBM rates (generally above 55%) were those that offered the greatest access to VBM by automatically mailing registered voters mail ballots, a practice known as “universal vote-by-mail.”

All of these states saw the highest 2020 VBM rates (above 55%), with the exception of Nevada, which had a 41% VBM rate in 2020. Eight out of ten states in this category had VBM rates above 68%.

While Nevada’s VBM rate appears relatively low for a state with a universal VBM policy, it is at least partially explained by the robust, in-person early voting option that Nevada also offered. With 23% of Nevada voters choosing to vote early in person, the total Nevada early voting rate—whether by mail or in person—was a combined 64%, on par with the other states in the universal VBM category¹⁶.

➠ Key Point: Conversely, states that limited access to VBM saw the lowest VBM rates (less than 17%).

Amongst all these states, all saw small increases in their 2020 VBM rates (less than 9%). Indiana actually saw a 17% decrease in its VBM rate.

➠ Key Point: Among states that automatically sent mail ballot applications to registered voters, or that accepted requests to vote by mail from most voters, 2020 VBM rates ranged widely; this suggests that other policies, election administration practices, or factors influenced voter preferences for voting method.

Of the states that automatically sent mail ballot applications to registered voters, Iowa had the largest VBM rate (59%), while North Carolina had the smallest VBM rate (18%). Of states that did not require an excuse or allowed COVID-19 concerns as an excuse to vote by mail, Montana had the largest VBM rate (86%), while Arkansas had the lowest VBM rate (10%). Montana’s VBM utilization rate, as well as Arizona’s (72%) were notable because they were on par with the VBM rates of states using universal vote by mail systems. In Montana’s case, the state implemented a new policy which allowed local county officials to decide whether they would automatically mail ballots to all eligible voters¹⁷ (because this was not a statewide mandate, we do not categorize this state as a universal VBM state). In Arizona’s case, its high VBM utilization rate is explained by Arizona’s history long history of absentee and early voting, starting with a 1991 law that allowed residents to vote by mail for any reason, and the fact that most early voting is done by mail¹⁸.

 

States with the Largest Gross Increases in Vote by Mail Rates

The ten states with the largest gross increases in VBM rates saw increases ranging from 33% on the low end, in Alaska, Pennsylvania, and Wisconsin, to approximately 70% on the high end, in New Jersey and Vermont (Table 2).

Table 2: Top 10 increases in gross vote-by-mail rates, 2016 to 2020

 

➠ Key Point: Of the ten places with the largest gross increases in VBM rates, eight of them automatically either mailed a ballot or a ballot application to every voter.

Across the four areas with the largest increases in gross VBM rates, all three states and the District of Columbia employed a universal VBM system for the first time in 2020, suggesting this policy change expanding access to vote-by-mail was a significant reason for the increase in VBM participation. While Pennsylvania and Alaska did not proactively send a ballot or ballot application to voters, they also did not restrict access to mail in voting, allowing voters to vote by mail without an excuse. Pennsylvania’s idiosyncratic situation also helps explain its presence on this list: it had a particularly low vote-by-mail rate in 2016 (4%), given that prior to a 2019 act passed by the state legislature, the state did not have significant absentee voting or in-person early voting options¹⁹. Due to this combination of a low baseline VBM rate and the first-time availability of expanded modes of VBM voting, Pennsylvnia saw a large percentage increase in its 2020 VBM rate, though its overall vote-by-mail rate in 2020 remained only 38%. These patterns amongst the ten states with the largest gross increases in VBM rates suggests that the easier it was for voters to vote by mail, the more likely voters were to take advantage of this option. Of these ten regions, only three-Nevada, Wisconsin, and Pennsylvania, were states whose results were repeatedly questioned in post-election litigation.

 

How Policy Changes Affected Vote-by-Mail Rates

Both policy changes to expand access to vote-by-mail and the relative flexibility of existing voting laws and policies were important determining factors in 2020 VBM utilization, and in changes in utilization from 2016.

Table 3 compares the percentage increase in VBM rates amongst states implementing newly expanded voting policies in 2020, using the Cybersecurity and Infrastructure Security Agency’s (CISA) designation of states that implemented major mail in voting policy changes.

 

Table 3: Impact of Policy Changes on Vote-by-Mail Utilization Rates

 

➠ Key Point:  States that made major policy changes saw a much greater mean percentage increase in vote-by-mail rates, versus those states that did not make major policy changes.

States that automatically mailed a ballot to all registered voters for the first time and states that relaxed the excuse requirements for voters to access mail ballots saw the largest percentage increases in mean VBM rates (296% increase among the former, 299% among the latter). These two categories represented the states that made the most significant policy changes in terms of increasing voter access to vote-by-mail. States that automatically mailed a ballot to registered voters had the highest mean vote-by-mail rate (63%), attributable to the fact that such universal vote-by-mail systems, while still providing the option for in-person voting, make mail-in voting the de facto voting method. On the other hand, states that changed their voting policies by relaxing the excuse requirements to be able to vote by mail had a mean utilization rate of 26%. Additionally, Nevada was the only state whose results would later be contested in public discourse and in the U.S. courts that implemented a major policy change in response to the COVID-19 pandemic.

➠ Key Point:  Of the states that did not implement policy changes, even those with the most restrictive VBM policies saw a notable percentage increases in their mean VBM rate in 2020. On the other hand, the states that continued expansive VBM policies from prior elections saw nominal percentage increase in their mean VBM rate.

The group of states with the most restrictive VBM policies—those that required an excuse to vote absentee—nonetheless saw  36% increase in mean VBM rates from 2016, with every state experiencing increased utilization of VBM from 2016. However, the mean VBM utilization rate remained low (14%), reflecting the more restrictive nature of their policies. The five states that had administered universal VBM elections prior to 2020 saw a nominal percentage increase in their mean 2020 VBM rates (4%). This reflected differing trends amongst states in the group: Hawaii’s VBM rate increased from 44% in 2016 to 83% in 2020, but Utah, Colorado, Oregon, and Washington saw they VBM rates decrease relative to 2016 by 5%, 7%, 11%, and 13%, respectively²⁰. Nonetheless, this group of universal VBM states had the highest mean VBM rate for 2020 (84%), notably greater than the mean VBM utilization rate of the states that implemented a universal vote-by-mail system for the first time (68%), which was the group with the next highest mean VBM utilization rate in 2020.

➠ Key Point:  Where states maintained in-person voting as the de-facto voting method but allowed voters to choose how to vote—either through policy changes that relaxed absentee voting excuse requirements or by maintaining policies that did not require an excuse—many more voters chose to vote by mail than did in 2016.

Among states that relaxed absentee voting requirements in 2020, the mean VBM utilization rate more than quadrupled, from 6% of voters in 2016 to 26% of voters in 2020. Among states that maintained the policy of not requiring an excuse to vote absentee, the mean VBM utilization rate nearly doubled, from 22% in 2016 to 42% in 2020. Of the seven states whose results would be much contested in the U.S. courts and the court of public opinion, six (Arizona, Georgia, Michigan, North Carolina, Pennsylvania and Wisconsin) were all states whose laws already allowed no-excuse absentee voting before the advent of the COVID-19 pandemic.

 

Section Conclusion

The 2020 election cycle saw the use of mail-in voting double across the U.S., with over 32 million more mail ballots returned in 2020 than in 2016²¹. In many states, this reflected an incredible effort on the part of election officials to update election administration operations within a very short period of time in response to the COVID-19 pandemic. Though some bumps in primaries conducted after the onset of COVID had created concerns about the November election²², the general election was a noted success. Election officials adeptly and securely administered an election that had record levels of voter turnout²³ and lower mail-ballot rejection rates than in the primaries²⁴. Ben Hovland, a member of the Election Assistance Commission appointed by former President Donald J. Trump, called it “one of the smoothest and most well-run elections that we have ever seen²⁵.” The Cybersecurity and Infrastructure Security Agency described the election as “the most secure in America history.²⁶ And, after a U.S. Justice Department investigation, then Attorney General William Barr concluded that “to date, we have not seen fraud on a scale that could have effected a different outcome in the election²⁷.” To have implemented this unprecedented, election administration mobilization in the midst of a pandemic, while still supporting record voter turnout, is a testament to the dedicated effort of countless officials and civil servants.

However, despite the fact that there was no evidence of election irregularities on a scale that would have changed the election result, many Americans believe that the 2020 presidential election was invalid²⁸, with the increased use of mail-in voting a frequently-cited reason for this mistrust²⁹. But changes in the use of mail voting were not uniform across the country, as each state determines its own election procedures. Consequently, the expansion of mail-in voting can only comprehensively be understood at the state level. Demystifying what drove these changes and how much voting patterns changed in each state since 2016 can help reduce confusion and distrust around the doubling of the use of mail ballots from 2016 to 2020 throughout the United States.

A clear driver of vote-by-mail (VBM) utilization is state law. States with the most restrictive VBM laws, such as Louisiana, Tennessee, and Texas, had the lowest VBM utilization rates. States with universal VBM systems, such as Colorado, Oregon, and Washington, had the highest VBM utilization rates. In terms of changing VBM rates since 2016, states implementing new, more expansive VBM policies for the 2020 election, either by moving to a universal VBM system, as in New Jersey, or by reducing restrictions for absentee voting, as in South Carolina, saw the largest increases in VBM utilization rates. But even states that did not change statewide voting policies in response to the pandemic generally saw increases in the use of VBM. Interestingly, the only group of states that saw small, marginal decreases in VBM utilization were states that used universal VBM systems before 2020, but, still, the absolute VBM rates in these states remained among the highest in the country.

The importance of a state-specific view of changing VBM patterns, and of a given state’s VBM laws, is particularly relevant with respect to the states whose results were the most contested. Of the states implementing major policy changes in response to COVID, only one, Nevada, was subject to multiple post-election legal challenges. The other six states that faced large numbers of post-election legal challenges, Arizona, Georgia, Michigan, North Carolina, Pennsylvania and Wisconsin, all already had laws allowing for the use of no-excuse absentee voting before the COVID outbreak. In other words, while these states did all see increases in the use of VBM, this reflected the reality that more voters were choosing to use this voting option in the middle of a pandemic than had in previous elections. In other words, the core structure of an expansive, no-excuse absentee voting system was already in place among these six contested states well before the onset of COVID, and election officials in these states had already had experience in and significant time to prepare for ensuring the security and integrity among mail ballots³⁰. Understanding these specific facets of the increase in mail ballots in 2020 clarifies the oversimplified concerns about the rapid increase of mail voting across the country, and consequently might help to limit the unsubstantiated beliefs about election fraud that followed from those concerns.

Ultimately, the increases in VBM utilization across states is a measure of significant voter demand for this voting method. Further, the fact that voter turnout reached record levels in 2020 suggests that the greater flexibility afforded to voters by allowing for a mail-in voting option led more people to vote than in prior elections, even in the midst of a pandemic. Both the demonstrated voter demand for mail voting, and the positive impacts of greater flexibility in voting methods on voter turnout should be taken into account in planning for future elections.

 

Section 2: Election Defense Inventory: mitigating the information threat to U.S. Elections in 2020

Section Author: Jaim Coddington

Key Insights

➠ The sum of all efforts made to protect U.S. elections in 2020 lacks a coherent label. We propose ‘election defense’ to describe these efforts going forward.

➠ The 2020 election cycle spurred the rise of an assertive election defense mentality among stakeholders. Actors across civil society, government, and the private sector established dedicated election integrity, security, and monitoring groups, often on continuous, 24/7 operational cycles and with a high degree of information sharing.

➠ The emergence of CISA as a primary enabler and organizer for election defense suggests a prominent federal role for future defensive efforts, even as state and local officials continue to lead election administration.

➠ While federal government support and resources may be welcomed by other actors, future politicization of CISA could undermine the election defense zeitgeist and hamper collaboration.

➠ Social media companies struggled to balance freedom of speech principles with the need to combat misinformation and disinformation, settling on an algorithmically -driven deamplification policy to reduce the virality and prevalence of damaging content without resorting to widespread censorship.

➠ Algorithmic annotation and flagging of digital content may accelerate and become more sophisticated if deamplification is ultimately judged as a successful election defense policy.

➠ Crisis planning efforts were vindicated by the events of January 6, 2021. The future prospect of physical threats to the U.S. democratic system should motivate election defense planners to redouble crisis response and physical security planning, particularly during presidential elections.

 

Overview

The information ecosystem today is a complex, decentralized space. The rise of social media has exponentially increased the number of information sources and the reach of personal communication networks amongst users of platforms like Twitter, Facebook, Reddit, YouTube, and Snapchat. Where once the news was filtered through a funnel tightly controlled by a few major news outlets, now information flows to the public through a disparate and largely disordered sieve. In this environment, not only is it easier for well-meaning users to unintentionally spread misinformation, it is also easier for ill-intentioned actors to purposefully spread disinformation designed to influence and erode Americans’ trust in election processes and outcomes. Malicious actors use highly visible and interactive public forums, including social media, to spread disinformation in support of Influence Operations (also known as Information Operations). These operations use disinformation to manipulate public opinion or influence behavior³¹.

Disinformation is not a novel threat. In the 2016 U.S. election, a bipartisan Congressional commission found that the Russian government targeted the election by a coordinated Influence Operation (IO) seeking to increase societal divisions and undermine public faith in the democratic process³². Other actors, like China, are also conducting long-term influence operations designed to sway American audiences to support pro-China policies³³. Perhaps most worryingly, national security experts recognized significant activity by domestic actors in amplifying and promoting foreign-source disinformation to serve their own election agenda³⁴. Foreign actors were also found to be amplifying domestically-produced disinformation, creating a perverse reciprocal relationship³⁵. Facing such tactics in 2018 by both foreign and domestic actors, public and private leaders expected these tactics to continue in the 2020 election cycle and worked to develop strategies to make election systems more resilient.

Offensive responses to IO, such as targeting foreign IO sponsors with cyber weapons, are considered costly in case of escalation and provide no guarantee of effective deterrence³⁶. On the other hand, repressive defensive responses to counter-IO methods have thus far been the domain of autocratic foreign regimes. Mimicking China’s Great Firewall, or Russia’s clampdown on civil society, would impose unacceptable costs on American civil liberties and constitutional rights³⁷. The established and intuitive options in the IO domain all presented their own ethical, legal, and practical dilemmas.

 

Data and Analysis

Stakeholder strategies to counter and respond to the IO threat in the 2020 election ultimately followed a different paradigm: not offensive measures, but multilateral and proactive election defense. Essential election actors including federal agencies, traditional and social media organizations, election officials, and civil society groups, launched broad and sometimes overlapping campaigns to blunt the impact of misinformation and disinformation on U.S. elections³⁸. This amounted to a whole-of-society mobilization to protect the bedrock of American democracy; because the disinformation threat was both persistent and pervasive, it required a holistic effort to counter it.

We characterize this effort as ‘election defense.’ The election defense concept acknowledges the complexity of both the disinformation problem and its solutions. It also implies a greater degree of proactivity and cohesion than the more generic term ‘election security.’ In 2020, election defense was a unifying cause across all sectors of American society: public servants, private industry, and civil society. Moving forward, it is an open invitation for all election stakeholders to join the effort of protecting democracy - to build a coalition that can defeat the disinformation threat. This coalition mentality will shape the outcome of future election defense efforts.

 

Election Defense Themes

Many observers have explored and critiqued the actions of individual election stakeholders in 2020, but the public discourse has largely bypassed a holistic view of the election defense ecosystem. While deep-dive analysis on the success or failure of individual actors’ election defense is vitally important, any attempt to saddle a single actor or stakeholder group with responsibility for election defense outcomes would create a faulty premise for future policy. Planners need an accurate, objective, and complete picture of the efforts made to defend the 2020 election cycle.

This picture cannot be painted without a clear understanding and codification of major actor groups and the roles of high-profile election defense organizations. To use national defense as an analogy: the public is generally aware of the U.S. armed forces and what they do. The average citizen probably knows that we have multiple service branches like the Army, the Navy, and the Air Force, and has a general sense of what roles those organizations play in national security. Most Americans are likely aware of our intelligence services, such as the Central Intelligence Agency, and how the intelligence community serves U.S. national security interests.

Conversely, a comparable public awareness of election defense does not exist. Although the security of our elections surely matters as much as our national security, the average American citizen may have no idea what the Cyber and Infrastructure Security Agency (CISA) is, or what it does. By painting a picture of major election defense actor groups, we hope to give both policymakers and the public an analytical baseline for further planning and deeper understanding. Our selection of representative actors from high-level election defense groups focused on a subset of organizations with a clear and enduring influence (and interest) in election defense.

For example, in regard to social media stakeholders, we examined the public statements of Facebook and Twitter, the dominant platforms in their sector in terms of active users in the US population. For traditional news media, we focused on the Associated Press, whose reporting infrastructure of 4,000 stringers across the U.S. supports election results reporting for The New York Times, The Wall Street Journal, the Washington Post, Fox News, NPR, and PBS, among others³⁹.

This survey is intended as a representative sample, not a conclusive profile. A wider look at other election stakeholders across all sectors would be a welcome addition to our initial findings.⁴⁰ There are many organizations and individuals we have not explicitly named here who deserve credit for their efforts to protect the 2020 election cycle.

All stakeholder actions we list and define here are derived from a combination of public statements by each actor and by publicly available reporting. We do not attempt to judge the impact of actions taken, or determine the extent to which a stakeholder carried out stated intentions to protect the election.

The resulting product, a 2020 election defense inventory, is intended to give the public a general framework for understanding the election defense ecosystem. Just as we think about organizations like the Army, Navy, Air Force, and Marine Corps as key actors in national security, we seek to identify the most influential and critical actors in election security. This inventory is not exhaustive, and is intended as a starting point for further analysis.

Our inventory of stakeholder activities reveals four areas of election defense actions, listed in Table 4, along with broad goals and methods across these areas. These themes are:

1. Trust Building: To prevent loss of public trust in electoral administration before, during, and after the election due to misinformation and disinformation.
2. Deamplification: To reduce the velocity and virality of disinformation and misinformation in the information ecosystem.
3. Voter Access: To offset the impact of election-related misinformation and disinformation which seeks to limit a specific group’s access to voting in any way.
4. Crisis Response: To enhance the resiliency of our election system in the face of crises, such as natural disasters on Election Day, and limit worst-case disinformation scenarios.

 

Table 4: The four Election Defense Action Areas, their goals, and methods

 

These action areas are sometimes mutually reinforcing and interdependent, but are conceptually distinct. For example, a trust building campaign to promote reliable and authoritative election information sources may also contribute to deamplification by dissuading citizens from seeking out misinformation and disinformation. Likewise, loss of public trust may precipitate the need for crisis response to post-election violence resulting from that loss of trust. But crisis response factors vary significantly from trust-building factors and require a separate planning methodology; post-election crisis is episodic and not inevitable, whereas trust-building is a constant and enduring need.

As a framework, these action areas point to the strengths among the various categories of actors in our inventory. Some actors are inherently positioned to cover specific themes — for instance, social media companies may have the most significant influence over deamplification of disinformation and misinformation within the public discourse because they have full control over user interface and user experience modifications on their digital platforms. In contrast, traditional media has little control over non-journalistic online media content, and in the deamplification context is largely resigned to a reactive fact-checking role. Comparing actor strengths across all four themes, we can infer that a comprehensive election defense approach requires a broad and diverse coalition of stakeholders.

 

Election defense measures by the National Association of Secretaries of State (NASS) & National Association of State Election Directors (NASED) (Election Officials)

Together, NASS and NASED are membership organizations that represent each state’s top election officials. Their members operate at the front lines of the electoral system, and serve as the primary and original source of authoritative, jurisdiction-specific voting information for the general public. Because NASS and NASED membership bodies are comprised of chief election officials, these organizations create a critical forum for coordination among the states on nationally relevant election issues⁴¹. Given their interlinked and complementary relationship, as well as their high level of collaboration through the Election Infrastructure Subsector Government Coordinating Council (EIS-GCC), we categorized these two key organizations together in our inventory⁴². Their notable collaboration points include supporting the creation of the Election Infrastructure Information Sharing and Analysis Center (EI-ISAC)⁴³, as well as targeted mobilization of licensed attorneys to act as poll workers during the election⁴⁴. Both organizations amplified voter access messaging from their individual members and from external partners like the Cyber & Infrastructure Security Agency (CISA)⁴⁵. Individually, NASS engaged with a large number of civil society groups on the #TrustedInfo2020 campaign, which sought to drive voters to credible online information sources on voting procedures, election processes, the impact of COVID-19 on election norms, and to validate election officials as the key trusted sources of information for the public. NASED worked in lockstep with NASS to promote #TrustedInfo2020⁴⁶, and also spearheaded concrete voter access measures in the months leading up to the election, including the provision of hand sanitizer for poll workers and voters in the months leading up to the election⁴⁷. NASED and NASS’s innovative, continuous, and highly coordinated efforts throughout the country’s electoral district helped ensure smooth and timely administration of the 2020 election. Many of the activities and collaborations conducted by NASED and NASS members, particularly those related to cybersecurity, were directly enabled by funds from the Help America Vote Act of 2018⁴⁸.

 

Table 5: Election Defense Inventory for the National Association of Secretaries of State & National Association of State Election Directors

Sources: EIS-GCC: An Open Letter to American Voters, 2018⁵⁰, #TrustedInfo2020, n.d.⁵¹, Elections Infrastructure Information Sharing and Analysis Center, n.d.⁵², I Joined the EI-ISAC - Now What?, n.d.⁵³, ‘This is How It Was All Supposed to Work’: The EI-ISAC Readies for Election Day, 2020^54, 55, ‘No Bar’ to What Election Officials Shared on Election Day, DHS Says, 2020⁵⁶, Voting In-Person After Requesting an Absentee Ballot, 202057, How We Prepared for the 2020 Election During a Pandemic, 2020⁵⁸, NASED and NASS 2020 Election Preparations and Reminders, 202059, ABA joins NASS and NASED to mobilize lawyers as poll workers for Election Day.⁶⁰

 

Election Defense Inventory for the Associated Press (Traditional News Media)

A closer look at one specific news organization, the Associated Press (AP) serves as the basis for our election defense inventory of traditional news media. This actor group was particularly difficult to choose from given the broad and diverse disposition of its constituents. We selected the AP as a representative actor because of its unique reach with live election reporting and its deployment of the AP VoteCast tool. The Associated Press (AP), provided sustained election coverage in every US state, with a focus on live reporting in battleground states⁶¹. AP deployed its polling analytics platform, AP VoteCast, to provide broadcast and print media organizations with polling data before, during, and after the election. VoteCast directly reported the vote counting effort in national and state races, and underpinned AP’s ability to declare winning candidates in each race⁶². Likewise, AP’s peer and competitor news networks rely on VoteCast data to make their own election outcome predictions, making AP a crucial link in the larger election information ecosystem⁶³. AP emphasized accuracy and timeliness in election-related coverage, allocating personnel to a dedicated Fact Check team and using push alerts to disseminate breaking news to the information ecosystem quickly⁶⁴. A renewed emphasis on explanatory journalism, with a concerted focus on the unique impact of early voting (both in person and by mail) on this election cycle, sought to build resilience in the electorate against misinformation, disinformation, and general confusion around the election, amplifying accurate information on election procedures, voter registration, and ballot sites⁶⁵. Multilingual election coverage in English, Spanish, and French sought to reach the broadest possible audience in analyzing and explaining the election outcome⁶⁶. AP’s reporting on other actors’ election defense planning sought to amplify a narrative of security and reliability regarding the election system.

 

Table 6: Election Defense Inventory for the Associated Press

Sources: AP Announces 2020 General Election Plans, 2020⁶⁷, AP VoteCast, n.d.⁶⁸, NORC at the University of Chicago n.d.⁶⁹, How We Call Races, n.d.70, and Explaining Election 2020, n.d.⁷¹, In a Hot Election, the Cool-Headed Associated Press Takes Center Stage⁷².

 

Election defense measures taken by Facebook (Social Media Companies)

With an estimated 190 million users in the United States⁷³, Facebook is a dominant information platform for the American electorate. Facebook’s election defense planning focused on voter registration and combating misinformation and disinformation on the platform⁷⁴. In 2018, Facebook began a hiring surge for human content moderators to complement its algorithmic content review process^75,76. In 2020, Facebook stood up a Voting Information Center within its user interface and regularly prompted users to register to vote, encourage others in their network to register/vote, visit voter registration portals such as TurboVote, and heed reminders on voter registration deadlines and voting procedures⁷⁷. The company detailed a range of content and user activity moderation actions, including: removing content under grounds of voter interference, marking inaccurate content with warnings and ‘informational labels’ for users, investigating and removing pages and groups involved in “coordinated inauthentic behavior”, rejecting suspicious or unauthorized ads, placing a moratorium on new political ads close to the election, and removing fake accounts from the platform⁷⁸. Governance of this deamplification regimen was conducted through a dedicated cross functional team working in an Election Operations Center and relied in large part on algorithmic identification of content⁷⁹. Facebook pledged to share information with federal and state authorities on election-related threats using sophisticated user activity analytics tools such as its Crisis Assessment Dashboard, and provided government officials with a security toolkit to help protect their Facebook account security during the election cycle⁸⁰. Finally, the company explicitly stated that it would defer to established, credible news sources in any situation regarding the projection of a winning candidate in a given electoral contest, with the intent of avoiding undue influence over public perceptions⁸¹.

 

Table 7: Election Defense Inventory for Facebook

Sources: “Preparing for Election Day,” 2020⁸², Leading countries based on Facebook audience size as of January 2021⁸³, “Preparing for Elections,” n.d.⁸⁴, Roose et al., 2020⁸⁵, Seetharaman, 2020⁸⁶, Why Am I Seeing a Post from Facebook about Registering to Vote? | Facebook Help Center, n.d.⁸⁷, 15 Months of Fresh Hell Inside Facebook⁸⁸, Facebook Clamps Down On Posts, Ads That Could Undermine U.S. Presidential Election⁸⁹

 

Election defense measures taken by Twitter (Social Media)

Like Facebook, Twitter commands a substantial US audience with nearly 70 million users⁹⁰. Twitter occupies a unique space in the election information ecosystem because it serves as a platform for key political leaders, allowing figures like former President Donald Trump to directly transmit election-related messages to a wide audience. Twitter made significant changes to its platform to attack misinformation and disinformation for the 2020 election, including wide-reaching deamplification measures and a permanent ban on all political advertising (initiated in 2019) on the platform⁹¹. While both Facebook and Twitter’s deamplification methods increased flagging and removal of malicious or false information content, Twitter also made extensive structural changes to its user interface and user experience. Innovations included limiting content recommendation criteria to two degrees of separation and mutual likes/follows between user accounts, redirecting users to ‘authoritative’ information sources via multiple pathways, and prompting users to quote Tweets and add value or context (rather than simply re-Tweeting)⁹². The company also launched a bilingual Election Hub to give voters a one-stop election information resource composed of credible sources⁹³. Twitter deliberately promulgated a sense of civic responsibility in its corporate culture, releasing a clear and transparent civic integrity policy forbidding any use of the Twitter platform which could undermine elections and civic processes, as well as making election day a paid holiday for all of its employees worldwide⁹⁴. Like Facebook, Twitter tooks steps to help secure the accounts of major political figures which could be exploited by malicious actors⁹⁵. In 2018, Twitter began sharing user activity data with civil society researchers with the specific goal of detecting and thwarting state-sponsored information operations, and maintains a public Application Programming Interface (API) to facilitate data transparency⁹⁶. Finally, starting in 2018 Twitter began its ongoing voter registration campaign using hashtags, push alerts, and partnerships with nonprofit groups⁹⁷.

 

Table 8: Election Defense Inventory for Twitter

Sources: Leading countries based on number of Twitter users as of January 2021⁹⁸, Operations on Twitter, n.d.⁹⁹, Elections Integrity, n.d.¹⁰⁰, Additional Steps We’re Taking Ahead of the 2020 US Election, n.d.¹⁰¹, The 2020 US Elections and Twitter, n.d.¹⁰², Kelly, 2020¹⁰³, Twitter Says Steps To Curb Election Misinformation Worked, 2020¹⁰⁴, Twitter Keeps Some Measures It Says Slowed Election Misinformation : NPR, 2020¹⁰⁵, Culliford, 2020¹⁰⁶, Guynn, 2020¹⁰⁷.

 

Election defense measures taken by the Cybersecurity and Infrastructure Security Agency (Federal Agencies)

The Cybersecurity and Infrastructure Security Agency (CISA) provided the security backbone for 2020 election systems with important implications for public trust, and employed a broad and innovative array of election defense techniques across all election defense themes. CISA was put forward as the primary national authority for election security in 2017¹⁰⁸ when Congress designated election systems as critical infrastructure, giving CISA the responsibility of securing all IT and physical infrastructure in the country related to the election process (voter databases, ballot counting systems, polling sites including early voting locations, election certification and validation systems, ballot storage facilities, etc). Since 2017, CISA has collaborated with other actors in the election defense community and taken a wide range of election defense measures anticipating the 2020 election cycle. Major CISA initiatives include the Countering Foreign Influence (CFI) task force, which embarked on a nationwide public education campaign, #Protect2020, to conduct rumor control and alert the public about the threats of foreign and domestic influence on elections¹⁰⁹. CFI used a variety of methods attempting to reach diverse demographics, including a graphic novel about election meddling and cybersecurity titled Real Fake¹¹⁰. On the cybersecurity front, CISA released a software package called Crossfeed.

Crossfeed gave state and local election teams a ready-made tool to scan individual election systems and assess vulnerabilities well ahead of the election¹¹¹. CISA funded the EI-ISAC cyber threat monitoring center, endorsed by NASS and NASED and run by the Center for Internet Security, creating a key multilateral node for election defense monitoring and situational awareness¹¹². CISA also published extensive information products, including mail-in voting policy change maps and infographics, to help voters understand where and when they could access mail-in voting¹¹³. Throughout the 2020 election cycle, CISA adopted an operational election defense mentality with constant, proactive activity aimed at securing election systems, keeping the public informed, fighting false narratives and rumors, and maintaining crisis response capacity.

 

Table 9: Election Defense Inventory for the Federal Government: Cybersecurity and Infrastructure Security Agency

Sources: Marks, 2020¹¹⁴, Joint Statement from DOJ, DOD, DHS, DNI, FBI, NSA, and CISA on Ensuring Security of 2020 Elections, 2019¹¹⁵, Fessler, 2020¹¹⁶, Krebs, 2020¹¹⁷, and Election Infrastructure Security, n.d.¹¹⁸, Fired Director of U.S. Cyber Agency Chris Krebs Explains Why He Says Vote Was “Most Secure in American History”¹¹⁹. Election Infrastructure Security Funding Considerations¹²⁰.

 

Election defense measures taken by the National Task Force on Election Crises (Civil Society)

The Nation Task Force on Election Crises (NTFEC) formed primarily as an apolitical safeguard against potential voting chaos during the pandemic, and to ensure a peaceful transfer of power after the election. NTFEC’s central premise was to preempt the possibility of or mitigate the fallout from election crises, with the goal of enhancing a smooth election season, maintaining public confidence in the outcome under any scenario, and facilitating a peaceful transition or continuation of power¹²¹.

As a bipartisan, non-governmental organization comprised of legal experts, retired public servants, and other credible authorities, NTFEC offered an independent voice in the public discourse, seeking to clarify and reassure the news media and the general public about election norms, procedures, legal precedents, and the handling of potential disputes¹²². The organization periodically released detailed memorandums on topics of concern (election postponement, incapacitated candidates, election calling procedures, election-related law enforcement activities, deployment of troops around elections, voter intimidation, electoral college procedures, etc)¹²³.

It can be inferred the NTFEC’s primary targets in its information campaign were not members of the general public, but rather the political class: members of the news media, lawmakers and their staffs, executives, law enforcement and military officials, and others with elevated awareness of and interest in legal procedure surrounding the election. NTFEC’s interactions with the news media are especially noteworthy. NTFEC offered media organizations customized resources and briefings explaining election law and procedures in order to support their election reporting efforts. It also made recommendations to major outlets about their methods for reporting election results, given the unique circumstances brought about by the pandemic¹²⁴. This activity positioned NTFEC as both a source for the news media by serving as subject matter experts on a variety of esoteric election topics, and as an adviser regarding the best election reporting practices in a complicated election year. In so doing, NTFEC filled a unique and important election defense niche left largely untouched by other election defense actor categories in our inventory.¹²⁵

 

Table 10: Election Defense Inventory for the National Task Force on Election Crises

Sources: There’s a Reason the Election Went So Smoothly, 2020¹²⁶, This Year’s Primaries Were a Mess. Here’s What Needs to Happen for November to Be Better, 2020¹²⁷, Can A Group of Policy Experts Prevent an Election Catastrophe in 2020?, 2020¹²⁸, National Task Force on Election Crises Releases Post-Primary Report, Detailing Lessons For Safeguarding November’s General Election, 2020¹²⁹, Watchdogs Demand Election Night Clarity from the Media, 2020¹³⁰, Resources, n.d.¹³¹.

 

Section Conclusion: Early Lessons from 2020 Election Defense Efforts

This inventory provides a baseline for understanding the holistic 2020 election defense approach, and reveals several common patterns among stakeholder groups. First, the 2020 election cycle spurred the rise of an assertive election defense mentality. Actors across civil society, government, and the private sector established dedicated election security and monitoring groups, often on continuous, 24/7 operational cycles with peak intensity on election day. In many cases, these groups were activated weeks or months ahead of Election Day. Permanent or cyclical election defense teams could become a fixture for stakeholders with the resources and incentives to maintain them¹³².

As revealed in public statements and interviews by CISA officials, 2020 election defense efforts included a high degree of information sharing and collaboration between most actor groups¹³³. CISA served as the primary organizing authority for much of this collaboration as a major funder and backer of EI-ISAC, and helped to identify needs and channel federal resources for other actors. The emergence of CISA as a primary enabler and organizer for election defense suggests a prominent federal role for future defensive efforts, even as state and local officials remain in control of election administration. While federal government support and resources may be welcomed by other actors, future politicization of CISA could undermine the election defense zeitgeist and hamper collaboration.

All actors in our inventory championed various forms of trust building techniques, such as explanatory journalism, election hubs and voting information centers, and civic integrity policies (measures for which D3P has also advocated since 2018). Individually, election defense actors recognized the need for deliberate public messaging to explain their efforts to the electorate. Election defense transparency and documentation alone may not convince diehard conspiracy believers that US elections are secure, and their impact on public perception is difficult to measure, but they may become a key part of public relations strategy going forward.

Social media companies struggled to balance freedom of speech principles with the need to combat misinformation and disinformation, settling on a deamplification policy to reduce the virality and prevalence of damaging content without resorting to widespread censorship. Deamplification embraced an tiered response methodology, with a spectrum of options ranging from content marking to content censoring. Annotation and flagging of digital content may accelerate and become more sophisticated if deamplification is ultimately judged as a successful election defense policy. Regardless of affiliation or concerns about public relations fallout, key actors all acknowledged the criticality of voter access in the 2020 election cycle, with special concern for the effects of COVID-19 and the accessibility of early voting. Facebook, in particular, targeted voter access as a key election defense metric and claimed to have registered four million new voters in this cycle¹³⁴.

Finally, this election triggered an exceptional degree of planning among all stakeholders for crisis scenarios regarding transitions of power and the escalation of violence after the election. These efforts were vindicated by the events of January 6, 2021, when thousands of protestors violently attacked and entered the United States capitol under the pretense of defeating supposed election fraud. The prospect of physical threats to the U.S. democratic system should motivate election defense planners to redouble crisis response and physical security planning, particularly during presidential elections. The likelihood of such future challenges should drive the election stakeholder community to further develop a planning and evaluation framework for ongoing election defense efforts.

 

Section 3: Traditional media coverage and social media narratives in the 48 hours after Election Day

Section Author: Yousef Ibreak

Key Data Points

➠ On November 3rd, scans of frontpage headlines from a representative pool of 26 traditional, print media outlets—including Breitbart News—show that none of these outlets led with reports of significant election administration mishaps or disasters.

➠ On November 4th, frontpage, traditional media coverage remained measured and consistent across the 26 outlets scanned: no outlets led with frontpage stories declaring a presidential winner, or reporting major instances of fraud or other election irregularities.

➠ November 5th was the first day where “election fraud” emerged as as a keyword in frontpage headlines across the 26 outlets monitored, appearing in frontpage stories in three of these outlets; all three outlets referenced fraud claims in reporting that they were baseless.

➠ In the post-election period, to his audience of over 80 million followers, then President Donald J. Trump questioned the legitimacy of the election approximately 22 times per day. His claims of widespread fraud and irregularities went far beyond, and starkly contrasted with, traditional media coverage in the initial days after November 3rd.

➠ Google Trends shows that searches for the words “election fraud” were at an all time high in the days after the election, while “Stop the Count” became a trending topic on Twitter, and a new “Stop the Steal” Facebook group became one of the fastest-growing groups in Facebook’s history.

 

Overview

News outlets play a unique role in U.S. elections: they project, unofficially, the winners of elections in the immediate aftermath of voting and often weeks before the states complete their audits, canvasses, and certifications of official results. Thus, effectively, the U.S. public learns the results of elections from news outlets based on these projections, not from election officials¹³⁵.

In the unique circumstances of the 2020 election cycle, news outlets not only strove to keep the public informed on the rapidly transforming election administration landscape under the pandemic; they also needed to prepare for the likelihood that they might not be able, as in past years, to confidently project a winner by the evening of Election Day because some swing states might not have counted a sufficient number of votes at that time to allow them to do so. This was a key concern for maintaining confidence around election results: a significant delay in news organizations’ projection of a winner could appear, without the relevant context, to be the result of serious election administration problems. In fact, while news outlets were not able to project a winner until the Saturday after the Tuesday election, this delay only signified that news outlets had to wait longer for election officials to complete their initial tabulations of votes. All of the post election audit, canvass, and certification processes that determine the official results proceeded along the same rules, regulations, and pace as in previous years.

Traditional news media spent months preparing for the idiosyncrasies of the 2020 election. Decision desks updated their predictive models and publicly reinforced their commitment to patience and transparency in projecting results¹³⁶. Newsrooms published fact sheets about how they track and project results¹³⁷. Outlets explained how voting by mail works¹³⁸, how to track an absentee ballot¹³⁹, the difference between projected and official results¹⁴⁰, and why states would take different amounts of time to complete a full count of ballots¹⁴¹. Record voter turnout and mail-in voting rates¹⁴² and a poll showing that only 17% of voters expected to know the projected winner by November 3rd¹⁴³ indicate that these efforts to educate voters on how to vote and on why a delay in projected results was not out of the ordinary were largely successful.

News producers and editors also implemented new practices regarding the reporting of results once voting ended on November 3rd to better reflect the realities of an election with large percentages of mail-in ballots and to continually reinforce, in real time, how this impacted their ability to project an unofficial winner¹⁴⁴. For instance, outlets shifted from tracking the percent of precincts reporting, which obscures the number of mail ballots remaining to be counted¹⁴⁵, to tracking measures such as the percent of outstanding vote. On-air anchors regularly reminded viewers where large outstanding votes remained to be counted, that the leading candidate in a given state could change if there remained large numbers outstanding votes¹⁴⁶, and that a seeming delay in projecting a winner did not mean that anything was wrong with the process¹⁴⁷. Further, outlets such as The Washington Post embedded a warning banner on its election results page that warned readers that increases in mail-in voting could lead to a delay in projecting election results, with a link to further details.

But, while Americans largely expected a delay in projections of an unofficial winner and understood its causes, other issues remained. A Cornell University survey conducted in the week before the election found “overwhelming evidence that Americans will view the election as legitimate only if their own candidate wins¹⁴⁸.” Further, since his candidacy, then President Donald J. Trump sowed doubt in the U.S. election system generally¹⁴⁹, while from at least May 2020, Trump regularly focused on questioning the integrity of mail-in voting, specifically¹⁵⁰. In hindsight, a poll from November 13-17—after every major news outlet had projected then candidate, now President Joseph R. Biden as the winner of the president election—showed that “52% of Republicans said that Trump had ‘rightfully won’” the election and that “68% of Republicans said they were concerned that the election was ‘rigged¹⁵¹.’”

These conditions raise the following questions: How did traditional news outlets manage reporting on the election in the days of uncertainty after the close of voting, but before they had projected a presidential election winner? How did the major outlets differ in their coverage? When and from where did narratives of election fraud occur? The following analysis focuses on answering these questions.

 

Data and Analysis: 2020 Election Day in Headlines

For the first time in 20 years, since the 2000 election between George W. Bush and Albert A. Gore¹⁵², the winner of the presidential election could not be confidently projected by the end of Election Day. We monitored and documented the frontpage lead story of 26 national and local media outlets multiple times per day from November 3-5th, collecting 142 headlines in this period. These outlets included the major TV news networks (ABC, CBS, CNN, Fox, NBC) national print outlets including the Associated Press, the New York Post,New York Times, Wall Street Journal, Washington Post, and Breitbart, and leading local papers like the Houston Chronicle, the Minneapolis Star Tribune, and the Miami Herald (see Appendix 1 for the full list of news outlets monitored). For comparison and context during this period, we also monitored Trump’s Twitter feed and sourced information on broader social media trends across Twitter, Facebook, and Google. We focused this analysis on the immediate days after November 3rd because this was the period in which the greatest uncertainty in election results existed, therefore, was the period in which the information environment was ripe for the spread of mis and disinformation.

 

Examining patterns in commonly-used keywords in headlines from frontpage

For a high level overview of topline election coverage from November 3rd-5th, we used word cloud visualizations of the 142 lead headlines documented—comprised of 2,864 words—to highlight common patterns in coverage. Word clouds summarize patterns in text inputs by sizing words according to the frequency of their use, with the most frequently used words assigned a larger size than less frequently used words. Changes in which words are used most frequently over time provides one measure of the focus of news coverage in different periods. We monitored coverage at three intervals after polls began to close on November 3rd, at two intervals on November 4th, and at one point on November 5th.

At 5pm on November 3rd, the most commonly used words in frontpage headlines were “voter,” “vote,” “early,” “ballot,” “election,” “states,” “day” and “polls.” Neither “Trump” nor “Biden” were used as frequently as these other words. This indicates that coverage in this period focused on Election Day voting operations, and not yet on the outcomes of voting. By 9pm ET on that same day, “Biden,” “close,” and “Trump” became the dominant terms used, suggesting that, at this point, outlets began focusing coverage around the closeness of the race between these candidates, and on the outcome of the contest. Near the conclusion of Election Day at 11pm ET, the most frequently used words in frontpage headlines were “Biden,” “Trump,” and “State,” indicating a continued focus on the outcome of the presidential contest between Biden and Trump. (Image 1)

 

Image 1. Word cloud visualizations from frontpage headlines of 26 major news outlets on November 3rd, 2020.

 

On November 4th, as ballot counting continued in key swing states and the unofficial winner remained unknown, the focus of frontpage stories shifted. Headlines increasingly used the word, “count” (highlighted in yellow, our emphasis), to the point where “count” was used as frequently as the names of the candidates themselves (Image 2). This indicated that stories increasingly covered the ballot counting procedures as the country waited for more information about the results to be shared.

 

Image 2. Word cloud visualizations from frontpage headlines of 26 major news outlets on November 4th, 2020.

 

By the evening of November 5th, “count” remained one of the top three most used words, alongside “Trump” and “Biden,” while the words “thin,” “razor,” and “margin” were increasingly used. This suggests that ballot counting stories remained prevalent, and that that the closeness of the presidential race was also a top theme. Of further importance, this is the first period in which “fraud” also emerged as a common word. (Image 3)

 

Image 3. Word cloud visualization from frontpage headlines of 18¹⁵³ major news outlets on November 5th, 2020.

 

➠ Key Point: November 5th was the first day where “fraud” emerge as as a keyword in lead headlines across different outlets.

“Fraud” (highlighted in yellow, our emphasis) only became a common word in frontpage headlines on the evening of November 5th, a change from topline election coverage up until this point. Three publications, the New York Post, USA Today, and Reuters, led with headlines that used the term, indicating that this was a turning point at which traditional media began to discuss fraud in their lead coverage. However, all three of these outlets, reporting on election fraud claims Trump had made in a White House briefing that day, described these claims as “baseless,” “false,” or “with zero evidence” (Image 4). As we will describe below, election fraud narratives were going viral on social media by this time.

Image 4. November 5th headlines that mentioned fraud on the frontpage

 

 

Examining prevailing election narratives from frontpage headlines on November 3, 2020

Beyond just examining patterns from commonly used words in headlines, we categorized headlines into common categories according to the type of story they described, tallying how many outlets led with these types of stories. We conducted this analysis for media scans made at the three intervals after polls began to close on November 3rd, and at two intervals on November 4th (full detail on the exact headlines each outlet published in each period can be found in Appendix 2).

Across the traditional media outlets monitored, around 5pm ET the majority of outlets led with frontpage stories focused on relatively uneventful Election Day operations, local voting stories, and record turnout. For instance, Reuters reported that “short lines and calm prevail at polls early on Election Day.”The Miami Herald highlighted a local story at the Hialeah polling location, where Trump supporters were vigorously and vocally making their support for him known in the parking lot, but, where the polling line actually formed, voters were “shy” and differential to election officials.Later in the evening of November 3rd, both at 9pm and 11pm, outlets focused on stories about swing states, record turnout, and the need to wait for more returns to be reported before projections could be made. For instance, The New York Post led with a headline that stated that the presidential race was “‘too close to call’ for battleground states.” No outlets led with stories about reports of election fraud or significant problems on November 3rd. (Table 11)

 

Table 11. Summary of Front Page Lead Stories from 26 Major Media Outlets on November 3rd, 2020

Note: for more detail on specific headlines and outlets included in each category, see Appendix 2

 

➠ Key Point: Scans of frontpage, traditional media from the evening of November 3rd show that none of the 26 news outlets monitored across the U.S. led with reports of significant election administration mishaps or disasters in this period. Neither were they prepared to project a winner of the presidential race. In contrast, on Twitter and in a live statement, Trump claimed victory late on Election Night, and that the election was being stolen from him.

 

The lack of stories about significant Election Day operational disasters was a mark of how well-administered the general election was, contrasting starkly with coverage of the primaries that had occurred after the COVID-19 pandemic set in that had focused on stories of operational shortfalls, such as hours-long waits at polling locations¹⁵⁴ or high rejection rates of mail-in ballots¹⁵⁵. No lead stories on November 3rd reported any evidence of election fraud or election manipulation, indicating that on the ground reporters across the country did not discover evidence of such incidents by this stage. Given the large number of outstanding, legally cast ballots remaining to be counted in key swing states at this point, no outlets were prepared to project a winner of the presidential contest.

In contrast, late on Election Night, just before 1am on November 4th, Trump first claimed he had won the election, and then alleged that some people were trying to steal the election from him on Twitter (Image 5). Around 2:30am that same morning, he made live statements to the same effect¹⁵⁶.

 

Image 5. Election Night tweets from President Trump

 

These statements were incongruous with reporting from all outlets monitored, including right leaning sources such as the Wall Street Journal, the New York Post, Fox, and Breitbart, none of which reported that the presidential winner was known at all, let alone that it was Trump who had won.

 

Examining prevailing election narratives from frontpage headlines on November 4th, 2020

By the morning of November 4th, coverage had largely focused on the status of the results. A plurality of the 26 news outlets monitored led with stories focused on the status of key swing states. Six of them led with stories about Trump’s declaration of victory around 12:45am on November 4th, but did not affirm Trump’s claims. Two outlets, Breitbart and Fox News, led with stories that painted the election as chaotic, but did not affirm either that fraud or malfeasance had occurred or that Trump had won . Breitbart ran with the headline: “Election Insanity: They Stop Counting Votes?! Trump Furious as Election Drags On,” while Fox ran with “Election Decision Delayed; Trump and Biden campaigns dig in for legal fight with key battleground state winners undeclared.” Breitbart continued with this same headline through the afternoon of the 4th. Across other coverage that afternoon, ten outlets led with stories focusing on updates to results, ten with stories focusing on the Trump campaign’s post-election legal efforts, and four emphasizing that votes were still being counted. (Table 12)

 

Table 12. Summary of Front Page Lead Stories from 26 Major Media Outlets on November 4th, 2020

Note: for more detail on specific headlines and outlets included in each category, see Appendix 3.

 

➠ Key Point: Frontpage, traditional media coverage on November 4th remained measured and consistent across the 26 outlets scanned; no outlets projected a presidential winner, and neither did they report major instances of fraud or other election irregularities.

 

By the afternoon of November 4th, election officials had counted enough ballots to give news outlets the confidence to project that Biden had won Wisconsin and Michigan¹⁵⁷, though no outlets projected a presidential winner. Instead, most outlets throughout November 4th continued to report on the status of races in remaining swing states and updates in results as more legally cast ballots were counted. Trump’s declaration of victory and his campaign’s post-election litigation efforts to challenge election results were highly covered topics as well, though no outlets, including Breitbart and Fox News, affirmed his claims of victory (for more on this litigation effort, see Section 4). However, Breitbart and Fox News did lead with headlines that described the election in negative terms, highlighting delays in vote counting and general “election insanity¹⁵⁸.”

In contrast, Trump continually and assertively disputed the legitimacy of updated results on November 4th by insinuating that unspecified actors were striving to illegally sway the election towards Biden. In various tweets, Trump claimed that these unspecified actors were “finding Biden votes,” “secretly [dumping] ballots,” and “working hard to make up” the interim, apparent leads that Trump had held in certain states. He also reiterated claims of victory specifically in Pennsylvania, Georgia, and North Carolina, where his interim leads were eroding as more ballots were counted. (Image 6)

 

Image 6: Sample Tweets from President Trump on November 4th, 2020 focused on swing states.

 

 

These tweets contrasted significantly with traditional news coverage from that period, which focused on the expected course of ongoing ballot counting procedures, did not report on interim vote tallies as final, and highlighted where the number of outstanding, uncounted, legally-cast ballots could impact the final result.

 

Comparison point: examining social media trends in the days after November 3rd

In typical mis and disinformation cycles, incorrect information spreads rapidly online, only making it to mainstream, traditional media sources after amplifying the conversation online¹⁵⁹. Stories about election fraud or irregularities in the 2020 followed this same pattern: evidence from Google, Twitter, and Facebook from November 3rd to November 6th show that pejorative election allegations about potential election anomalies and fraud spiked on social media platforms earlier than on traditional news media.

➠ Key Point: In the days after November 3rd, “Stop the Count” became a trending topic on Twitter, while a new “Stop the Steal” Facebook group became one of the fastest-growing groups in Facebook’s history. Google Trends also shows that searches for the word “election fraud” were at an all time high in the days after the election.

Over different platforms, discussion of election fraud and questioning of results went viral after voting ended on November 3rd and in the 48 hours afterwards where ballot counting continued. “Stop the Count” was a phrase used directly by President Trump in his Twitter feed to urge that vote counting stop at a point where he maintained artificial, premature leads in key swing states. By November 5th, “Stop the count” had become the 8th top trending topic on Twitter¹⁶⁰. Further, a Facebook group called Stop the Steal was created on Wednesday, November 4th with posts and videos claiming that Biden was stealing the election, and by the following morning, it had garnered over 320,000 members, becoming “one of the fastest-growing groups in Facebook’s history¹⁶¹.” Meanwhile, Google searches for “election fraud” first spiked on November 4th, peaked on November 5th, and declined thereafter, but still remained a more frequently employed search than before Election Day for a couple months (Image 7).

 

Image 7: Daily frequency of “Election Fraud” Google searches

 

 

The virality of these three examples from Twitter, Facebook, and Google show how rapidly narratives of election fraud spread across social media in the immediate aftermath of Election Day. In addition to his tweets, starting on Election Night alleging election fraud, an address that Trump made from the White House Briefing Room on November 5th may also explain the viral spike across social media platforms, because the TV networks carried this speech and most outlets reported on it. In the address, Trump expanded upon the allegations he had been making on Twitter: “If you count the legal votes, I easily win. If you count the illegal votes, they can try to steal the election from us,” he stated, alongside explicit claims that Democrats committed fraud in order to win¹⁶².

By late on November 5th and certainly by November 6th, likely in large part due to this White House Briefing Room speech, Trump’s claims had truly jumped from Twitter and social media and landed in the mainstream. The Washington Post reported on the speech at 8:21pm ET on November 5th by describing it as “a litany of falsehoods and grievances, with some baseless conspiracy theories thrown in for good measure¹⁶³.” A Wall Street Journal Editorial Board article from November 6th also addressed the speech by focusing on pointing out the lack of evidence to back up Trump’s claims, but stating that the President would have to—and had every right to—seek to prove his allegations in court¹⁶⁴. Fox News reported on the briefing on November 5th after 8pm by leading with a headline highlighting criticism that the speech elicited from some Republicans¹⁶⁵. In fact, the skepticism across the Fox network about Trump’s allegations was so notable that the New York Times reported on it¹⁶⁶. Breitbart covered the speech with an article on November 6th by reiterating Trump’s without either affirming or denying them¹⁶⁷.

Even though traditional media outlets did not affirm Trump’s allegations, and most denied them, the mere fact of reporting them, reiterating Trump’s claims in the process, may have amplified the viral spikes in election fraud themes already seen across Twitter, Facebook, and Google before the speech. Moreover, the fact that tens of millions of voters continue to believe in Trump’s allegations of election fraud¹⁶⁸ suggests that the 48 hours after Election Day in which these stories were already starting to go viral on social media, but were not yet the focus of traditional news coverage, were critical in the spread and seeding of these narratives.

 

The Activity and Reach of then President Trump’s Twitter Feed

Beyond these general trends, the activity and reach of Trump’s Twitter feed, specifically, is relevant to understanding prevailing narratives in the days immediately after the end of voting.

➠ Key Point: In the post election period, to his audience of over 80 millions followers, Trump questioned the legitimacy of the election approximately 22 times per day. His claims of widespread fraud and election irregularities went far beyond, and starkly contrasted with, traditional media coverage in the initial days after November 3rd.

In terms of activity, from November 4th to 23rd, Trump posted to Twitter approximately 29 times per day, attacking the legitimacy of the election approximately 22 times per day¹⁶⁹. Between November 3nd and early on November 5th, Twitter labeled 38% of Trump’s tweets and retweets with warnings that his posts may contain misleading claims¹⁷⁰. As we have shown above, the messages and statements that Trump made on Twitter consistently contradicted topline reporting in traditional news outlets in this period, including in right-leaning outlets generally favorable to Trump.

In terms of reach, these tweets reached his audience of 81.1 million followers¹⁷¹. Only eight people have more followers than that¹⁷². Further, Trump’s tweets have high engagement rates. For instance, of Trump’s top five Tweets from November 3rd to December 30th with the greatest total engagement, Trump received around 1 million likes per Tweet; of his top ten tweets with the most engagement over this period, eight involved claims that he had won the election or either directly stated or implied that illegal election fraud was occurring to overturn his win¹⁷³. Thus, for a number of reasons, from his position as President, to his wide-reach on Twitter, Trump’s voice in this period was one of the loudest, consistently questioning the legitimacy of the election to one of the broadest audiences of followers on Twitter.

 

Section Conclusion

Traditional news outlets generally managed the uncertainty around the outcome of the presidential election by focusing on the closeness of the race in key swing states and any update to election results as ballot counting continued. No outlets projected a winner until the Saturday after Election Day, reflecting the uniqueness of a mid-pandemic election that relied significantly more on vote-by-mail and saw many states continuing to count large numbers of ballots after Election Day. While the tone of election coverage differed amongst outlets, particularly in the case of Breitbart, the substance of topline reporting generally did not, including with Breitbart.

Yet, from November 3rd-5th, a Google trends analysis of word searches for “election fraud” spiked in concert with “Stop the Count” becoming a top 10 trend on Twitter and a “Stop the Steal” Facebook group becoming one of the fastest growing Facebook groups in the platform’s history. The virality of these narratives in this period underlines the importance of those initial 48 hours after the end of voting. The unprecedented election administration challenges of vast numbers of uncounted ballots after election day, coupled with the uncertainty about the outcome of the presidential race created a significant vulnerability in the information threat landscape. Right-leaning, verified “blue check” influencers, whether intentionally or not, used this vulnerability to amplify mis and disinformation around election fraud narratives¹⁷⁴.

Trump, with his audience of 81.1 million followers, was one of these important, right-leaning blue check accounts. From November 3rd to 5th, as well as beyond, he tweeted with great frequency during this period questioning the legitimacy of the election, and received high engagement on these Tweets, echoing these statements in live comments. Trump’s tweets differed significantly from traditional media coverage in two dimensions. First, while traditional media outlets exercised restraint in projecting an election winner, Trump did not. He declared victory late on Election Night, before a significant portion of votes had been counted in key states. Across 26 traditional news media outlets in the U.S., none, including such pro-Trump outlets such as Breitbart and Fox, affirmed his claims of victory¹⁷⁵. Second, while no major news outlets ran with lead stories about allegations of election fraud¹⁷⁶—which would have been frontpage news if credible claims had arisen—such claims were frequent features in Trump’s Twitter feed. These facts indicate that Trump himself was a key driver and amplifier of the election fraud narratives in the critical 48 hour period after Election Day and beyond.

 

Section 4: Patterns and impacts of election litigation on election outcomes and narratives

Section Author: Amanda Powers

Key Data Points

➠ Over 400 election-related legal cases, an unprecedented number, were argued across the country before Election Day, including 26 cases tried in the final month before November 3rd, creating uncertainty during the voting period about what the final election administration laws and procedures would be.

➠ Over 60 election-related cases were filed from November 3rd, Election Day, to December 14th, when the Electoral College met and voted; of these, 58 sought to secure outcomes that would be favorable to then President Donald J. Trump.

➠ All but one of these 58 post-election cases failed to secure favorable rulings in the courts, and none succeeded either in proving that instances of election fraud or other irregularities had occurred, or in changing the official results.

➠ These 58 cases targeted states whose results were still uncertain on Election night, increasingly focusing on those states where Trump had held an interim lead, but which then candidate, now President Joseph R. Biden eventually won.

➠ Post-election cases first strove to stop vote counting in states where Trump’s interim lead was narrowing, then to target and disqualify certain categories of ballots from voting pools known to have greater proportions of Biden voters, and then to stop the certification of official results in states won by Biden.

➠ Throughout this period, Trump frequently referenced on Twitter the existence of these cases to support his claims of fraud and election irregularities, but largely ignored the failure of these cases to prove their arguments in the courts.

 

Overview

The 2020 election cycle set a new high watermark for litigation before and after Election Day. Before the election, at the onset of the COVID-19 pandemic, state legislatures and election officials scrambled to respond to the new public health challenge by updating mail-in and in-person voting procedures¹⁷⁷,¹⁷⁸. Many of these changes were rejected or formalized through over 400 legal cases, an unprecedented number that vastly eclipsed the 196 lawsuits filed both before and after the contested 2000 presidential election¹⁷⁹. These cases strove to clarify the interpretation and application of state election law as states worked to adapt election administration to the COVID environment. The results of these lawsuits materially impacted election administration and voter choices about how to cast their votes (as documented in the changing voting methods described in the first section of this report). With dozens of lawsuits ongoing in the month before the election, litigation also played an important role after voting had ended, when over 60 lawsuits were filed contesting various elements of the presidential election process¹⁸⁰. Both pre and post-election litigation increased the complexity of election administration in the month before and after the election, and played a role in prevailing election narratives in these periods. Further, the pattern of post-election litigation suggests that the motivation behind the cases was not to root out instances of election irregularities everywhere, and that this litigation served a political narrative.

 

Data and Analysis

 

Pre-Election Litigation

By late October, 2020, more than 400 election-related cases had been filed in the 2020 election cycle. Many of these cases focused on adjusting voting policies during the pandemic, and, thus, on policies governing mail-in voting. In some of these cases, political parties, campaigns, or civil society groups hoped to clarify voting laws that were ambiguously written; in others, these plaintiffs asked the courts to implement new policies or to overturn policies passed by state legislatures or by a governor’s executive order. The primary conflict in these legal moves was advocacy either in favor of increased flexibility in voting, or in favor of increased voter fraud prevention measures.

D3P tracked the 26 cases that would impact voting and ballot counting procedures for the November 3rd election in all 50 states and the District of Columbia between October 1st and November 3rd (approximately one month in advance of the final day of voting) in the D3P Election Dataset—Voting Access Litigation Tracker tab. These pre-election lawsuits dealt with specific parts of the mail-in voting or election process, such as the deadline to receive mail-in ballots or whether or not mail-in ballots needed to be signed by a witness in order to be counted. They were filed primarily in swing states—specifically, in Alabama, Arizona, Colorado, Georgia, Indiana, Michigan, Minnesota, Missouri, Montana, New Jersey, Nevada, North Carolina, Pennsylvania, South Carolina, Texas, Wisconsin.

➠ Key Point: Democrat-affiliated litigation sought different outcomes from Republican-affiliated litigation, but, with respect to litigation filed in the month before Election day, judges largely determined that they did not have the jurisdiction to intervene in how elections were conducted.

In the month before November 3rd, generally, Democrat-affiliated litigators focused on expanding access to voting across states, while Republican-affiliated litigators argued that such expansions would threaten election integrity and were not necessary to ensure that citizens could safely cast their votes during the pandemic (D3P Election Dataset—Voting Access Litigation Tracker tab). For instance, Democratic lawyer Marc Elias led then candidate, now President Joseph R. Biden’s campaign effort on state voter access issues¹⁸¹ by focusing on four areas: 1.) asking states to provide pre-paid postage on ballots, 2.) shifting VBM deadlines to allow for ballots postmarked by Election Day, rather than requiring them to be received by Election Day or earlier, 3.) allowing ballot curing for signature matching issues and third party ballot collection, and 4.) eliminating witness requirements¹⁸². On the other hand, lawsuits brought by Republicans generally challenged the implementation of new voting policies resulting from litigation or decisions made by state and local election officials, arguing that such decisions wrested power from state legislatures, and opened election security and integrity vulnerabilities. For example, multiple lawsuits in Harris County, Texas challenged a decision made by county officials to allow all voters access to curbside voting, which had historically only been used by voters with disabilities¹⁸³,¹⁸⁴. Ultimately, whether led by Democrat or Republican-affiliated litigators, very few of the legal decisions made by the courts in the last month before Election Day moved to change the standing election policy; in most cases, judges found that it was out of the court’s jurisdiction to intervene in how elections were conducted¹⁸⁵, ¹⁸⁶.

➠ Key Point: These 26 election-related cases tried in the month before November 3rd added further complications to an already complicated election year.

By October 1st, 30 states had already started sending mail ballots to voters (North Carolina was the first on September 4th), and nine states had begun early voting operations¹⁸⁷. This introduced uncertainty about how voters could effectively and legally cast their ballots during the voting period. Even as late as the week before the close of voting on November 3rd, there were nine voting cases pending (Table 13). Among these, for instance, were multiple lawsuits in Pennsylvania regarding the receipt deadline for mail-in ballots (the date by which mail ballots needed to be received in order to be counted).

 

Table 13: Voting access litigation governing 2020 election law still pending in the week before the end of voting on November 3rd

Source: D3P Election Dataset. Note: one case in North Carolina, Moore v. Circosta, addressed multiple types of voting policy, so is counted multiple times in this chart.

 

These cases, argued during the voting period, meant that voters and election officials had to operate with information that was subject to change. In an unprecedented year, election officials had to add to their many election preparation tasks the responsibility of continually communicating these changes to voters and continually updating their election administration operations accordingly. Chris Whitmire, Director of Public Information at South Carolina’s Election Commission, highlighted this new challenge: “aside from the administrative issues of polling places, poll managers, COVID-19 supplies, and processing absentee ballots, our ability to educate voters on changes (or lack thereof) will play a key role in a successful November election”¹⁸⁸.

 

Post-Election Litigation

The D3P team also tracked election-related litigation filed between November 3rd, the end of voting, and December 14th, the date on which Electors in each state met to cast their votes, formalizing the official results of the elections held by the 50 states. In some instances, this included continued monitoring of the impact of lawsuits about post election processes that had been filed before November 3rd, and were still active after Election Day. While pre-election litigation had focused on voting mechanics and processes, post-election litigation focused on post-election processes, such as counting ballots and certifying election results.

Of the over 60 cases filed during this period, 58 were brought by President Donald J. Trump’s campaign and related Republican entities. These cases broadly argued that election officials failed to follow proper election procedures, that there were significant election irregularities, and, as a consequence, the election results should be amended in various ways. One of these cases, Arizona Republican Party v. Fontes, asked only for an audit of, rather than some sort of change in, the results, and therefore was excluded from further analysis below. These cases followed a distinctive, phased pattern:

• Phase 1: Stop the Count—(11/3/2020-11/7/2020)
• Phase 2: Target and Disqualify Certain Ballots (11/5/2020-12/1/2020)
• Phase 3: Stop the Certification (11/16/2020-12/14/2020)

 

➠ Key Point: Post-election litigation was overwhelmingly failed in the courts, in no instances proving fraud or significant election irregularities, or in changing the results of the official tabulation processes.

A common trend across the three phases was the failure of these lawsuits: the 57 post-election cases filed by the Trump campaign and its affiliates to influence the election results through the courts were overwhelmingly unsuccessful. A majority (34 cases) were denied by the court. Interestingly, ten were voluntarily dismissed by the plaintiffs, and six were dismissed with prejudice by the court, a determination on the merits of the case that prevents further appeals on the matter¹⁸⁹. Only one case, Donald J. Trump for President, Inc. and the Republican National Committee v. Boockvar, filed in Pennsylvania in the first phase, was decided for the Trump campaign¹⁹⁰. (Table 14)

 

Table 14: Summary of the outcomes of the tracked lawsuits across the three post-election litigation phases.

Described in detail below, differences among these phases is in part a reflection of a shifting strategy amongst the litigants, and in part a reflection of the continued progression states made through the post-election processes towards certification.

 

Phase 1: “Stop the Count” (11/3/2020-11/7/2020)

In the first phase of post-election litigation, before any of the major news sources projected then candidate Joseph R. Biden as the ultimate winner on November 7, new cases or appeals of earlier cases were filed largely by the Trump campaign or Republican Party affiliates as officials across the country continued to count and report vote tallies. None of these cases made any allegations of voter fraud, instead bringing challenges regarding procedural issues related to ballot counting. These cases generally sought to stop, pause, or limit state vote tabulation processes by arguing that states were not complying with their own laws regarding these processes. (Table 15)

 

Table 15: Phase 1: Stop the Count Cases

Note: A single star (*) preceding a case name indicates that the plaintiff in the case was the Trump campaign; a double star (**) preceding a case name indicates that the plaintiff in the case was affiliated with the Republican Party, but not explicitly with the Trump campaign; three stars (***) indicates that the Trump campaign and the Republican party joined the plaintiff as an intervenor.

 

➠ Key Point: Litigation filed in this phase was not motivated by election integrity concerns about large numbers of outstanding, uncounted mail-in ballots post Election Day, or about potentially greater election integrity vulnerabilities in states with tight races. Rather, litigation focused solely on states for which no projection of an unofficial winner had been made, and, thus, were seen as “up-for-grabs” and determinative of the winner of the Presidency.

States targeted in this phase did not include all states with large numbers of outstanding, uncounted mail ballots after November 3rd, nor all states with close results that involved some number of outstanding mail ballots, where relatively smaller election integrity issues could materially affect the outcome. In other words, the pattern of litigation does not suggest general concern over the integrity of election results as a result of large increases in mail in voting. For instance, as of November 6th, California had an extraordinary number of ballots—4.5 million—remaining to be counted¹⁹¹. In Alaska, only 50% of its votes had been counted by November 5th, and the state did not plan to release the results of any absentee ballot until November 10th¹⁹². Yet no cases were filed in either state. Nor does the pattern suggest that the motivation came from concern around the integrity of the results across all swing states, where tight races mean that a proportionally lesser amount of manipulation could swing election results. For instance, as of November 4th, Ohio, a critical swing state, had over 286,000 outstanding mail-in and provisional ballots to be counted¹⁹³. Similarly, in Florida, 4% of the expected vote remained uncounted by the evening of November 3rd¹⁹⁴. But neither in Ohio nor in Florida were cases filed.

Rather, the pattern suggests that legal efforts targeted only states where the races were either too close or too early to project an unofficial winner by the evening of November 3rd, and the number of electoral votes assigned to a state would be consequential in reaching the 270 threshold to win the presidency¹⁹⁵. In other words, litigation targeted only states that appeared critical and determinative to news media calculations to project an unofficial winner. All of these characteristics were true of Pennsylvania, for instance, which saw seven cases filed in this first phase of litigation alone. The states with the next highest case count in this phase were Arizona, Minnesota, and Nevada, all of which had two case filings. It is perhaps not a coincidence that Pennsylvania, of the states that did not have a projected winner for the presidential race by the evening of November 3rd, holds the most electoral votes (20), four more than the next highest number of electoral votes (16), held by Georgia.

➠ Key Point: Litigation in this phase focused particularly on the states where Trump maintained an apparent lead on November 3rd, but which Biden eventually won, and ignored states where, conversely, Biden’s initial lead was eventually overcome by Trump.

The states targeted for litigation in this stage included all three states—Georgia, Michigan, and Pennsylvania—where Trump maintained an apparent, interim lead 12 hours after polls closing, but which Biden eventually won¹⁹⁶. The most cases were filed in Pennsylvania, where Trump’s initial lead 12 hours after the close of polling was approximately 500,000 votes, the largest initial lead among states that eventually flipped to the other candidate¹⁹⁷. Such interim leads emerged based on the order in which election officials tabulated ballots, and the fact that Democrats were more likely to vote by mail, while Republicans were more likely to vote in person on Election Day, in part because Trump himself denounced mail-in voting repeatedly in the run up to the election¹⁹⁸. Consequently, it was expected that, in a state like Pennsylvania, where mail ballots were counted last, per state laws that did not allow any pre-processing of mail ballots received before election day¹⁹⁹, Biden’s lead would only emerge late in the counting process. Conversely, in Ohio, where officials tabulated mail ballots first²⁰⁰, Biden held an initial lead after the close of polls. In spite of these predictable and normal features of the tabulation process, litigation targeting these four states, and Pennsylvania particularly, appeared to try to take advantage of the change in the leader of the race to argue that such a change was evidence enough of malfeasance.

But litigation was not consistent in this approach, instead focusing only on states where it was Trump’s, and not Biden’s initial lead that was eroded as ballot counting continued. For instance, as mentioned above, Biden had maintained an initial lead in Ohio for a few hours after the close of polls. In North Carolina, an even closer analogue to how the results evolved in Georgia, Michigan, and Pennsylvania, Biden still held an initial lead in North Carolina 12 hours after the close of voting, after which point Trump pulled ahead. Yet no cases seeking to stop ballot counting were filed in either Ohio or North Carolina. This further suggests that litigation was not generally concerned with election integrity issues, but on contesting the results in only states for which there was a political expedient: preventing a Biden win and ensuring victory for Trump.

Ultimately, if counting had indeed stopped on the evening of November 3rd, and none of the legally-cast ballots that were as-yet uncounted were tabulated, Trump, by that measure, would have “won” Georgia, Michigan, and Pennsylvania. With their 52 electoral votes, he would have gained more than the 270 vote threshold needed to win the presidency. The pattern of litigation filed during this “Stop the Count” phase suggests that striving for this outcome was part of the Trump campaign’s strategy. Trump’s public statements appear to support this conclusion.

➠ Key Point: Trump’s Tweets are consistent with the conclusion that legal efforts were not concerned with election integrity generally, but, rather, strategically sought to take advantage of Trump’s initial leads in certain states to pause ballot counting at procedurally arbitrary, but politically expedient, points in order to guarantee a Trump victory.

From November 3rd to 7th, while these cases were being filed, Trump publicly claimed to have won the election “by a lot,” and called for states to stop counting ballots (Image 8).

 

Image 8: Stop the Count related tweets from President Trump

 

Trump referenced the initial leads he had held among some of the states continuing to tabulate ballots, and suggested that the “miraculous” erosion of those leads was something that was nefarious and would therefore be remedied through litigation. On the morning of November 5th, at a procedurally arbitrary, but politically expedient, point—Michigan and Wisconsin had been projected for Biden on the 4th, and Trump’s lead was continually shrinking as results were updated in Pennsylvania and Georgia—President Trump proclaimed that election officials should stop counting ballots. Further, in a Tweet on the afternoon of November 4th, Trump singled out Pennsylvania, Georgia, and North Carolina specifically, “claiming” them for “Electoral vote purposes” because, at that moment in time, “each one of which has a BIG Trump lead” (Image 6, in Section 3).

This, of course, ignored the number of outstanding legal ballots remaining to be counted. In that same tweet, he further stated that “Additionally, we hereby claim the State of Michigan if, in fact....... there was a large number of secretly dumped ballots as has been widely reported!” Thus, Trump further reinforced the assumption that continued ballot counting, if the results were detrimental to his lead, were evidence of illegal actions²⁰¹. In singling out the states where he had built ostensible leads that were continually shrinking, Trump explicitly pointed to a strategy that focused on halting further ballot counting in critical swing states without projected winners in order to preserve and ensure the apparent, initial leads that he had held.

➠ Key Point: Some of Trump’s tweets in this period made specific reference to, and was consistent with, the lawsuits being filed; others went far beyond the claims made, and evidence brought, in these lawsuits.

Some of Trump’s tweets in this phase referenced specific elements of the cases that had been filed in the phase. For instance, his November 5th Tweet claiming that “ANY VOTE THAT CAME IN AFTER ELECTION DAY WILL NOT BE COUNTED!” (Image 8) was consistent with litigation that sought to prevent any ballots received after November 3rd from being included in vote counts.

In fact, all states required ballots to be in the mail or dropped off before Election Day; but, many states’ laws consider ballots to have been legally cast even if they arrive at election offices after election day (but within a specified period), as long as they were postmarked by Election Day. This issue was highly visible in Pennsylvania: before and after the election, multiple lawsuits were filed in state and federal courts to challenge election officials’ extension of the vote-by-mail deadline to allow for ballots postmarked by November 3rd to be counted as long as they were received by November 6th²⁰².

However, other tweets went well beyond the specific claims being made in legal proceedings. For instance, Trump’s November 4th Tweet referencing that “a large number of secretly dumped ballots as has been widely reported!” represented a relatively specific allegation of fraud (Image 7). Yet, none of the lawsuits filed within this same period referenced any such voter fraud, exemplifying how Trump’s public statements often made assertions beyond what his legal team believed they could prove in court (Table 14).

 

Phase 2: Target and Disqualify Certain Ballots (11/05/2020-12/03/2020)

In the second phase of post-election litigation, lawsuits targeted ballot categories on the basis of certain characteristics (e.g., ballots with missing witness signatures or partial dates, or, simply, all absentee ballots) or on the basis of where they were cast (e.g., all ballots from a certain county or city). These lawsuits asked the courts either to exclude these categories of ballots from official counts, or, even more extreme, to prevent the certification of official state results, an action that would effectively nullify all votes within a state. These cases were filed largely after the media had projected Biden as the winner, but before states had finalized official results through the certification process. Lawsuits targeting certain ballot characteristics focused on elements such as mail-in ballots returned with partial dates or addresses, or mail-in ballots with missing secrecy envelopes; in some cases, they simply targeted all mail-in ballots (Table 16). Other lawsuits targeted ballots from specific locations, alleging that these locations had high rates of voter fraud (Table 17). And some lawsuits targeted ballots with certain characteristics—such as ballots with partial dates—but only from certain local jurisdictions (Table 18).

 

Table 16: Phase 2 cases targeting ballots with certain characteristics

Source: D3P Election Dataset.

 

Table 17: Phase 2 cases targeting ballots from certain locations

Source: D3P Election Dataset.

able 18: Phase 2 cases targeting certain ballot categories also from specific locations

Source: D3P Election Dataset.

 

➠ Key Point: Lawsuits in this phase were filed in the same states as in the Stop the Co to target only states where a change in result would be favorable to Trump. However, rather than seeking to stop ballot counting, these cases sought to disqualify certain categories of ballots, reflecting the fact that cases from the previous phase had been unsuccessful in disrupting progress on the normal course of post-election processes.

The states in which cases were filed in this phase of litigation, with the exception of the two cases in Wisconsin, were all also the subject of lawsuits in the first phase of litigation. This suggests that the legal strategy of focusing on critical swing states with uncertain outcomes on the evening of November 3rd remained in place in the second phase of litigation. In this period, nine cases were filed in Pennsylvania alone, as opposed to four in Michigan and two in each of Nevada, Georgia, and Wisconsin, indicating that Pennsylvania, with both the largest number of electoral votes and a counting order in which Biden’s lead only emerged after some days, was key to this strategy.

But both the subjects of these cases and the remedies sought reflected the fact that cases in the first phase had been unsuccessful at halting the normal course of post-election processes, and that election officials were proceeding with moving from the initial tabulation of ballots to auditing, canvassing, and certifying the results. While in the first phase, cases aimed to stop ballot counting, cases in this second phase aimed to exclude specific categories of ballots, largely after initial counts had been completed, or to prevent certification, which occurs only after the initial count and a full canvass of the vote. These updates indicate that the Trump campaign and its affiliates, again the primary parties filing these suits, intended to stymie and block post election processes at all turns, whether election officials were in the counting phase, auditing phase, canvassing phase, or certification phase.

➠ Key Point: The ballot categories targeted for disqualification in this phase were categories that appeared to be composed mostly of likely Biden voters. Less than half of these cases made any allegations of voter fraud, all 23 were unsuccessful, and seven were voluntarily withdrawn by the plaintiffs.

In disputing mail-in ballots, or ballots from largely urban, suburban, or predominantly minority counties, these cases centered on voting blocks that contained proportionately more votes for Biden than Trump²⁰⁴,²⁰⁵. In seeking to disqualify these specific ballot categories, or to block certification due to allegations regarding these ballot categories, the lawsuits appear to be less focused on rooting out potential fraud everywhere, than on blocking ballot categories that were likely to turn state results to Biden²⁰⁶. In fact, less than half of these cases (ten out of twenty-three) included allegations of voter fraud.

As in the first phase, the 23 cases filed in this period were unsuccessful. Eleven cases were denied by the court, three were dismissed with prejudice, one was ruled moot, and one was a split decision. Strikingly, seven of the cases in this period were voluntarily dismissed by the plaintiffs. In fact, cases that sought to disqualify ballots solely on the basis of the location where ballots were cast—an action that would surely disqualify large numbers of legally cast ballots—were uniformly either dismissed with prejudice or voluntarily withdrawn, with the exception of the case that was ruled moot. In dismissing cases with prejudice, these judges ruled that they were completely without merit, and issued a decision that would prevent further appeals on that case. Moreover, the fact that so many cases were voluntarily dismissed shows that even the litigants did not have confidence in their arguments and evidence to see the cases through to a ruling.

➠ Key Point: In public communications and tweets, Trump often referenced ongoing litigation in this period, but frequently made claims that went far beyond the arguments made in the courts and ignored the fact that no judges, in examining these cases on their merits, found any evidence of voter fraud.

Throughout this period, Trump continued to invoke fraud and election rigging on Twitter, frequently referencing specific states and cities where relevant litigation was pending. However, in many cases, the allegations he made on Twitter did not match what was actually occurring in the courts. For instance, no judges, in examining the cases on their merits, found any evidence of voter fraud or violations of election processes, deciding in no instances to discard any regularly cast, non-provisional ballots (and, even then, only 213 provisional ballots were disqualified in a split decision in Ziccarelli v. Westmoreland County Board of Elections). Yet, Trump’s public communication and tweets repeatedly claimed that voter fraud had been proven or was rampant. He also continued to claim he had won the swing states under dispute. (Image 9)

 

Image 9. Target and Disqualify Certain Ballots example Tweets from President Trump

 

 

In alleging that he would win various state contests if certain ballots were excluded from the count, or if a recount or audit was conducted, many of Trump’s tweets at this time were consistent with the lawsuits in this phase that sought to disqualify certain ballot categories or that asked for audits (Image 10).

 

Image 10: Target and Disqualify Certain Ballots example Tweets from President Trump continued

 

 

Again, we see that the details that Trump employs were not necessarily in line with what the lawsuits argued. For instance, no case alleged that 700,000 additional ballots were not counted that should have been; on the contrary, all cases in Pennsylvania sought to disqualify certain ballots, rather than to ensure that more ballots were counted. Moreover, in court, Trump’s lawyers admitted that there were, in fact, a “nonzero” number of Repbulican poll observers in Philadelphia²⁰⁷. Nor were these tweets in line with the failure of all these cases to prove fraud or procedural issues had occurred.

In examining Trump’s public statements on Twitter, it appears that he used references to the existence of legal cases to enhance the credibility of his claims of fraud or other wrongdoing, which aimed to decrease the perceptions of the credibility of the election. However, his public statements often went far beyond the actual allegations contained in these lawsuits, and completely ignored the judgements that repeatedly find no merit in these allegations.

 

Phase 3: Stop the Certification (11/16/2020-12/14/2020)

The third phase of cases occurred as states neared the point at which they were ready to certify official election results. These cases focused on attempting to reverse, pause, or otherwise prevent certification in states won by Biden, in some cases asking the court to order a whole new election or to allow the state legislature (where the legislatures were controlled by a Republican majority) to appoint electors to the Electoral College. The first case of this kind was filed in Michigan on November 16, and these cases continued to be filed over a month after Election Day²⁰⁸. (Table 19)

 

Table 19: Phase 3: Stop the Certification cases

 

➠ Key Point: Litigation in Phase 3 continued to target states in which no winner had been projected by the end of November 3rd, and only those states which Biden eventually won. Cases in this phase sought to prevent or reverse the certification of state results, which would have effectively disenfranchised tens of millions of legally cast votes, but were entirely unsuccessful.

As in the prior phases, litigation in this phase was concentrated in states whose results were uncertain on the evening of November 3rd, would determine the winner of the presidency, and which Biden eventually won. In focusing on reversing or preventing the certification of results only where those results would formalize Biden as the President-Elect, the pattern of these lawsuits continued to indicate that the Trump campaign and its affiliates were only interested in alleging and examining election irregularities where it would serve them politically, rather than objectively pursuing irregularities across states, regardless of how it would impact results. At a minimum, it appears these lawsuits would allow them to exploit the apparent “delay” in projected results to augment doubts about the credibility of Biden’s win, or, at a maximum, to flip the results in those states to Trump.

Since legal cases in Phase 2 failed in seeking to affect post-election processes, election officials continued working through the procedures of auditing, canvassing, recounting, and certifying results, with each state certifying their official results before the December 14th deadline. In Phase 3, cases got less, rather than more, specific in their allegations. Cases argued that widespread fraud and/or procedural failures had occurred, rather than fraud or procedural failures limited to various ballot categories. Plaintiffs sought to reverse or prevent certification of all state results, rather than to exclude certain ballot categories. It seems an unlikely strategy: If litigants had not succeeded in convincing judges to throw out entire categories of ballots on the basis of unsubstantiated claims of fraud or procedural missteps, they were not likely to succeed in convincing judges to disenfranchise all of a state’s voters without concrete proof of fraud or procedural missteps. Indeed, none of these cases were successful. Fourteen were denied by the court, two were voluntarily dismissed by the plaintiffs, one dismissed for lack of jurisdiction, and three were dismissed with prejudice.

➠ Key Point: In this phase, Trump continued to question and sow doubt in the legitimacy of elections in the targeted states with explicit and implicit references to ongoing litigation. Without ever acknowledging the overwhelming failure of these cases in the courts, the mere existence appeared to serve as the continuing amplification of claims of fraud.

During this phase, Trump’s tweets repeated two main themes: first, that the election was fraudulent, and, second, that he could still win. In many instances, Trump explicitly referenced ongoing litigation procedures, such as hearings that were occurring or “big things happening” in the cases (Image 11), though, in fact, no bombshell evidence ever materialized in the courts and none of judgements in these cases found any instances where fraud occurred.

 

Image 11: President Trump’s tweets referencing litigation in Phase 3

 

In at least one tweet, Trump, referencing a statement from Sidney Powell, a member of Trump’s legal team whom he later disavowed, directly warned states against certifying election results:

 

 

In other tweets, without specifically mentioning legal proceedings, Trump referenced cities that were targeted in the second phase of litigation, where lawsuits often honed in on specific, local jurisdictions; in this third phase, while litigation efforts broadened to focus on state-wide results, the states targeted corresponded with the cities targeted in the second phase (Image 12).

 

Image 12: President Trump’s tweets highlighting cities targeted in litigation

 

 

In continually mentioning, both explicitly and implicitly, ongoing litigation, Trump continued to question and sow doubt in the legitimacy of elections in the relevant states. However, never did Trump recognize or concede the overwhelming failure of these cases to prove any substantial cases of fraud or procedural malfeasance. The mere existence of the cases, in spite of how they were decided, served as a method to continue to amplify claims of fraud.

 

Section Conclusion

Pre-election litigation, particularly given the relatively high volume of cases still under review in the month leading up to Election Day, added another layer of complexity to the 2020 election cycle. Both voters and election administrators had to deal with the uncertainty that resulted from decisions pending in courts that would affect election administration, almost up until the last possible minute. In this light, the ability of election officials in pulling off a smooth general election, particularly when compared to the difficulties that emerged in primary elections held after the advent of COVID-19, is an even greater success. The high volume of pre-election litigation also speaks to the design feature that state legislatures may not have the ability to rapidly and fulsomely respond to the extreme, exogenous events that can impact election administration, such as COVID-19. Updates to election laws should take into account the possibility of extreme, adverse events and update election laws to better prepare for such possibilities, including by clarifying the parameters for election administrators in responding to these events.

Post-election litigation efforts on the part of the Trump campaign and its affiliates was overwhelmingly unsuccessful: not only did all but one one of the 57 cases filed fail in the courts, but none of them proved that election fraud had occurred. Some of these cases were voluntarily withdrawn by the plaintiffs themselves, many did not make any allegations of fraud but focused on election process issues—such as whether enough observers had been allowed in tabulation rooms. The unprecedented number of post-election cases represented a fulsome examination of many claims of election irregularities, and no judge—many of whom had been appointed by Trump—found merit in any of these claims.

The targeting pattern and arguments of cases suggest that political expediency, rather than general concern over election fraud, motivated this post election litigation. Cases targeted states in which no winner had been projected by the end of Election Day, and increasingly narrowed in only on states where a Biden win looked increasingly likely as counting continued. In other words, cases targeted states where the absence of a projected winner on Election day could be used as evidence of problems with the election administration in that state, and where a reversal in results would favor Trump. The strategies for arguing this evolved as election officials moved through the usual course of post election processes, first trying to stop further counting, then targeting and trying to disqualify certain ballot categories that had proportionately more Biden votes, and then trying to prevent or reverse the certification of statewide results.

These cases failed to prove any instances of voter fraud or election irregularities had compromised the integrity and security of the election results. However, their mere existence might have served to bolster Trump’s public statements about election fraud. Trump often made direct references to these legal cases, as well as oblique references to the locations targeted in these cases or the allegations made in them. However, his public statements often went well beyond the scope of what was being argued in the cases; for instance, he asserted that widespread fraud had occurred in the middle of the Stop the Count Phase, when most cases filed by his campaign and its affiliates made no allegations of fraud at all. The references to the mere existence of these court battles, regardless of their outcomes, might have bolstered the claims of fraud.

 

Section 5: Prevailing narratives about the U.S. election on Weibo and in Chinese traditional media outlets

Section Author: Yi-Ting Lien

Key Data Points

➠ In the days after the November 3rd, several U.S. elections-related topics trended on Weibo, which is also known as the “Chinese Twitter.” These trending topics, as well as posts hashtagged with “Trump,” “Biden,” or “U.S. elections” expressed no preference for either U.S. presidential candidate, and often portrayed the U.S. elections as chaotic, polarized, and inefficient.

➠ In leaked memos from November 3rd, the Communist Party of China ordered traditional media companies and interactive platforms to “strictly follow unified reporting” on topics related to the U.S. elections. Apparently consistent with this guidance was the fact that many of the trending topics on Weibo related to the U.S. election were supported by posts from Chinese state-backed organizations.

➠ Chinese language posts on Twitter differed in content from posts on Weibo about the U.S. elections in the days after November 3rd. On Twitter, a company banned from China and therefore not subject to the same governmental control as Weibo, posts hashtagged with “Trump,” “Biden,” or “U.S. elections” were not consistnetly candidate-agnostic, often expressing pro-Trump messages.

➠ In the days after November 3rd, of two leading state-backed traditional media sources in China, the People’s Daily did not report on the U.S. elections at all, while the Xinhua News Agency devoted more U.S.-related coverage to highlighting the devastation of the U.S. COVID outbreak than reporting on the U.S. elections.

➠ In contrast, coverage of the U.S. elections in Taiwanese traditional media outlets differed notably: Taiwanese outlets dedicated significant frontpage coverage to the U.S. election in the days after November 3rd, and immediately recognized then candidate, now President Joseph R. Biden as President-Elect after the major U.S. news outlets projected he had won the election, unlike Chinese outlets.

 

Overview

Foreign influence operations, including those directed by state-affiliated actors, have increasingly become a feature of U.S. elections since 2016, and 2020 was no different. Actors from Russia, Iran, and China mounted both covert and overt influence operations in this general election cycle, seeking different outcomes²⁰⁹. U.S. election-related influence operations (IO) coming out of the People’s Republic of China (PRC) are particularly interesting. The PRC has increasingly sought to broadly promote the idea that the “China option” (中国方案) is a more efficient development paradigm than that of liberal democracy through cultural and information-based IO beyond its borders²¹⁰. As a leader of the liberal democratic order, the United States’s model is the implicit counter-option, and international perceptions of its elections, the bedrock of its governance system, impact preferences for this model.

While covert IO efforts from the PRC in the 2020 election cycle were both broadly unsuccessful and strove to engage U.S. citizens in seeking to cast the U.S. in a negative light, the state’s overt IO efforts were driven by public comments from state-backed organizations, focusing on PRC citizens as one key target group²¹¹. The focus of this analysis is on these overt IO efforts in the PRC. What narratives about the U.S. election in China trended on social media and prevailed in traditional media? How congruent were those narratives? What types of actors supported them? These are the questions we seek to answer.

In the days following November 3rd, we monitored discussions about the U.S. elections on the Chinese social media platform, Weibo, which included coverage from state-backed media sources in the PRC. Weibo, also known as “Chinese Twitter,” is one of the most popular social media platforms in mainland China, where the company is headquartered and where the bulk of its user base is located²¹². Like other media companies in the PRC, Weibo is subject to government content control measures²¹³. Leaked guidance from the Communist Party of China (CCP) on November 3rd shows that the government was particularly focused on messaging regarding the U.S. presidential election, as the guidance ordered media companies and interactive platforms to “strictly follow unified reporting” on the topic²¹⁴. Thus, an examination of trending conversations about the U.S. elections on Weibo would reflect at least some measure of CCP-controlled messages. As a point of comparison, D3P also monitored discussions about the U.S. elections in Mandarin-language Twitter posts, a rough proxy for Chinese opinions outside of PRC control²¹⁵, and in traditional media coverage in the democratic Republic of China (ROC), often known as Taiwan²¹⁶.

 

Data and Analysis

Trending Topics Related to the U.S. election on Weibo

From November 3-8th, we monitored Weibo’s daily list of top trending topics amongst its 241 million daily active users (DAU)²¹⁷ to determine the degree to which U.S. elections were a significant topic, and how the topic was being discussed in China. For each trending topic, we also documented example posts that supported the virality of that topic category (see Appendix 4).

From November 4th-8th, topics related to the U.S. elections were consistently trending among Weibo’s top 15 trending topics. On November 4th, two of the top 15 trending topics on Weibo were related to the U.S. election, on the morning of the 5th, four of the top 15 trending topics were related to the U.S. election, after which point the number of top trending topics related to the U.S. election decreased. (Image 13)

 

Image 13: Trending topics lists on Weibo November 4th-8th

 

 

 

➠ Key Point: Posts from Chinese state-backed organizations often supported and amplified the trending narratives on Weibo. Across example posts and trending topics, no preference was expressed for either U.S. presidential candidate, though many trending narratives portrayed the U.S. elections as chaotic, polarized, and inefficient.

Examining specific posts related to each of the above top trending topics provides greater detail on the accounts supporting these posts and the messaging they promoted. Often, the trending topics were supported by posts from Chinese state-backed organizations, and, in terms of messaging, some posts focused on factual descriptions regarding the state of the U.S. elections, while others involved pejorative characterizations of U.S. elections.

For instance, on November 4th, U.S. elections-related subjects were the eleventh and fifteenth top trending topics on Weibo. The eleventh-most trending topic was about a shooting in Nevada on November 3rd. One example post supporting this topic came from the Beijing News, a state-backed media company, emphasizing that the shooting had occurred on Election Day. The fifteenth -most trending topic was about the U.S. election generally. An example post supporting this topic came from CCTV, another state-backed media company, stating that supporters of then President Donald J. Trump and then candidate, now President Joseph R. Biden had clashed outside of the White House and two people had been arrested. Other posts supporting this topic came from verified Weibo accounts that had messages similar to those from state-backed media sources. For instance, users remarked, “Stunning! While the votes are still being counted, Trump has declared victory in the East Wing,” and, “As Trump wins Florida and Texas, votes for him are now far more than what the U.S. mainstream media polls predicted. The BLM (Black Lives Matter) movement in DC is ready to make trouble.” The congruence between posts from verified and state-run accounts is unsurprising, as evidence suggests that, at least since 2013, the Chinese government has exercised increasing influence over verified Weibo users, effectively ensuring that posts from these users reinforce government-driven messaging^218,219. (See Appendix 4 for screen shots of these posts and all posts referenced below.)

On the morning of November 5th U.S. elections-related subjects were the first, second, forth, and fifth top trending topics. The first trending topic focused on the news that Biden had won the Wisconsin race. An example post supporting this topic came from IFENG.com, a state-backed media company, stating that Biden had won Wisconsin therefore held 248 electoral votes. The second trending topic focused on Trump’s request for a recount in Wisconsin. An example post related to this topic came from the China News Service, also a state-backed media company, reporting that Trump had requested a recount due to election irregularities. The fourth trending topic of that morning was related to a Trump campaign lawsuit in Pennsylvania. One example post shows CCTV, again, a state-backed media source, posting that “Rudy Giuliani claims that the votes for Biden might be from Mars” and that Giuliani might file federal lawsuits on behalf of the Trump campaign to root out alleged malfeasance on the part of the Democrats. The fifth trending topic of that morning emphasized that many Americans were seeking to immigrate to Canada; CCTV was again one of the accounts supporting this trend, reporting that online searches for “immigrating to Canada” had surged in the U.S. on election day.

By the evening of November 5th, U.S. elections-related subjects were trending in the second, sixth, and eight spots. The second most trending topic was about the money spent on the U.S. Election. CCTV again was one of the accounts posting reports about record-high election spending in the U.S., as was the Communist Youth League (共产主义青年团), the youth organization of the CCP. An additional example post from a verified user stated that “The U.S. spent 14 billion U.S. dollars for an election, and China spent 1,169 billion to save its people [from COVID-19] at any cost!” The sixth most trending topic was about Nevada’s six electoral votes, and how slow the vote counting process was in that state. The eighth most trending topic was about Trump’s demand to stop vote counting. The Global Times, a state-backed media company, was one of the accounts posting about this topic.

On November 6th, the number of trending topics related to the U.S. election had decreased to two. The first was about the U.S. election generally, and was the sixth most trending topic at that time. One example post from a verified account reporting Trump’s statements that if only legal votes were counted, he would have won the election. The second was the ninth most trending topic at that time, and focused on two states dismissing Trump campaign lawsuits. One example post came from CCTV, which reported that the Trump campaign was pursuing lawsuits in several states, that Michigan and Georgia dismissed the Trump campaign lawsuits filed there, and finished by stating that an Arizona vote counting site was surrounded by Trump supporters, “many of them carrying guns.”

By November 7th, only one top trending topic was related to the U.S. election—though it was the second most trending topic—focusing on a lawsuit filed by the Trump campaign in Nevada that was dismissed by the court. An example post supporting this trend came from Ifeng.com, again, a state-backed media organization, matter-of-factly reporting that the Trump campaign had filed a lawsuit in Nevada on November 5th, which had already been dismissed.

On November 8th, the only trending topic about the U.S. elections—the ninth most trending topic at that time—was about protests and celebrations at the same time in the U.S. This was an oblique reference to the fact that Biden had been projected the winner at this point, but was reflected the diplomatic complication that it was still weeks before China’s President Xi Jinping and his government would recognize and congratulate Biden as the winner²²⁰. One example post came from CCTV, stating that “police stood by as Trump supporters protested, and Biden supporters celebrated on the streets.” This notably did not focus on Biden as the winner, in keeping with Xi’s refusal to acknowledge Biden as the winner at that time.

The prominence of state-backed organizations, particularly CCTV, in authoring posts related to trending Weibo topics evidences the importance of their role in shaping narratives on social media. It is consistent with the leaked memos identifying the U.S. election as a sensitive topic around which the CCP intended to create “unified reporting.” The congruence of messages between verified and state-run accounts further indicates a coordinated messaging campaign, given evidence of the level of state influence over verified accounts. These messages did not indicate a preferred candidate. Instead, they focused on the occurrence of specific events, or, in many cases, portrayed U.S. elections as chaotic and expensive, and the U.S. population as polarized, pugnacious, and disorderly, even indicating that many American citizens were seeking to emigrate to Canada, in one instance. In other words, an analysis of trending topics of Weibo from November 3th-8th indicates that, under the influence of state-backed organizations, social media conversations in the PRC emphasized a negative view of the U.S. election, implicitly supporting the comparative appeal of the “China option” (中国方案).

 

Weibo Posts Hashtagged with Trump, Biden, or U.S. Elections

Hashtags are traditionally used in social media to help make messages that use hashtags searchable across the internet, and for users to respond to common threads. Searching for and following certain relevant hashtags provided another way, beyond Weibo’s trending topics, to monitor conversations on Weibo about the U.S. election. From November 3rd to 6th, we also collected all 23,913 posts that hashtagged #川普/#特朗普 (Trump), #拜登 (Biden), or #美國大選 (U.S. elections) in both simplified and traditional Chinese characters on Weibo. Many of these posts were not related to the U.S. elections, focusing, rather, on topics such as the U.S.’s response to the COVID-19 pandemic, history between the U.S. and China, or the possibility of an invasion of Taiwan²²¹. Of the 3,236 posts that dealt directly with the U.S. election, we analyzed patterns and identified the most common topic areas, as summarized in Table 20.

 

Table 20²²²: Common topics on Weibo among posts with #Trump, #Biden, or #USelections

 

➠ Key Point: The pattern of posts on Weibo about the U.S. election that hashtagged Trump, Biden, or U.S. elections was similar to that of trending Weibo topics.

The relatively low number of posts related to these topic summaries evidence how focusing data collection on the three hashtags (#Trump, #Biden, and #USelections) limited the number of posts collected. Nonetheless, the focus on these simple, relevant hashtags provide another view into the prevailing narratives in China about the U.S. elections. The pattern of messages collected with these hashtags is similar to the pattern of the trending Weibo topics related to the U.S. elections: the most prominent messages were candidate-agnostic, event focused, and often emphasized negative characterizations of the U.S. elections.

 

Twitter Posts Hashtagged with Trump, Biden, or U.S. Elections

In contrast to Weibo’s 240 million DAU, Twitter has 186 million DAU worldwide. While the CCP has blocked the service since 2009, there are an estimated 10 million active users in mainland China as of 2016 who access the site through a virtual private network²²⁴. From November 3-6th,we tracked Mandarin-language posts across Twitter, documenting all 4,187 tweets that hashtagged #川普/#特朗普 (Trump), #拜登 (Biden), or #美國大選 (U.S. elections) in either simplified or traditional Chinese characters. Among these, we identified the most prevalent topic areas, categorized the 939 posts relevant to these areas into the appropriate topic summary, and calculated the retweets generated in each grouping (Table 21).

 

Table 21: Common topics on Twitter among Mandarin-language posts with #Trump, #Biden, or #USelections

 

➠ Key Point: The pattern of posts on Twitter about the U.S. election that hashtagged Trump, Biden, or U.S. elections differed from the pattern on Weibo; posts on Twitter were not necessarily candidate-agnostic, often expressing pro-Trump messages.

Across the data collection period, 4,187 tweets were posted that met the language and hashtag criteria. This relatively small number of tweets would not ordinarily indicate a significant trending topic on Twitter. However, the sample nonetheless provides an important comparison point to posts on Weibo with the same hashtag and language constraints, as Twitter is less controlled by, if not free from, CCP government control. Interestingly, as of January 10th, 2021, 6% (238) of the 4,187 tweets we documented were no longer available on Twitter. While some posts had been removed by the users, others appeared to have come from accounts that had been suspended. The most common reason for which Twitter suspends accounts is that they are “spammy” or “”just plain fake,” but the company also suspends accounts if it believes an account might have been “hacked or compromised,” or if it violates Twitter’s rules about abuse²²⁸.

In contrast to the generally candidate-agnostic posts on Weibo, more posts on Twitter expressed pro-Trump messages, such as wishing Trump a swift victory, questioning the accuracy of the vote counts in states that Biden led, mentioning Biden’s family scandal. One possible explanation for the pro-Trump leaning on Twitter comes from Sile Zhao, a Chinese journalist who gained fame as an activist for women’s rights and by writing books about Chinese NGOs; in her book Her Battles, (2017), Zhao suggested that Twitter has been widely used by Chinese dissidents as a platform to share their views about the CCP government even before Weibo was launched in 2009. Thus, Mandarin-language Twitter users might, in general, support Trump because they favor his aggressive stance towards the PRC.

 

Chinese news outlets and Taiwanese news outlets covered the 2020 U.S. election

The D3P team also monitored traditional news coverage about the U.S. in both communist China and democratic Taiwan from November 3rd to 9th. Focusing on top frontpage coverage, we documented all lead stories in that time period that focused on the United States.

 

Traditional media coverage of U.S. elections in China

In China, we monitored coverage in two most prominent news outlets, focusing on frontpage coverage in the People’s Daily, the official newspaper of the Central Committee of the CCP, and on all headlines in the Xinhua News Agency, a state-run news agency subject to the CCP.

➠ Key Point: Of two leading Chinese traditional media sources, the People’s Daily did not report on the U.S. elections at all, while the Xinhua News Agency devoted more coverage to highlighting the devastations of the U.S. COVID outbreak than U.S. elections.

These two state-backed publications took divergent approaches on reporting about the U.S. election. The People’s Daily, on the one hand, did not cover topics related either to the U.S. generally or to U.S. elections specifically. Instead, top headlines focused on CCP-related subjects, such as a speech given by Xi Jinping, policy announcements, or political achievements. On the other hand, the Xinhua News Agency did cover topics related to the U.S. elections, but, more of its U.S.-related coverage over that period focused on how severe the COVID-19 pandemic outbreak had become in the U.S. (Image 14).

 

Image 14: Focus of Xinhua News Agency lead articles about the U.S. from 11/3-11/9

 

 

Beyond devoting more coverage to stories about COVID in the U.S. than elections in the U.S., the content focus of the U.S. COVID outbreak was negative. Stories emphasized rising and record case numbers, and adverse economic impacts on specific industries, the unemployment rate, and the economy as a whole. Meanwhile, consistent with state-backed posts on Weibo, coverage of the election remained candidate-neutral. One notable narrative, reflected a Xinhua article from November 9th, was the fact that the PRC had not yet recognized Biden as president-elect, though it had been two days since all major U.S. media sources projected Biden as the winner and most world leaders congratulated Biden as president-elect²²⁹. (Table 22)

 

Table 22: Xinhua News Agency coverage of U.S.-related topics from 11/3-11/9

 

 

Traditional media coverage of U.S. elections in Taiwan

In Taiwan, the D3P team monitored coverage in four prominent outlets, the United Daily News, the Apple Daily, the Liberty Times, and the China Times (Table 23).

 

Table 23: Taiwanese traditional media coverage of U.S.-related topics from 11/3-11/9

 

 

➠ Key Point: Coverage of the U.S. elections in Taiwanese traditional media outlets differed notably from coverage in Chinese traditional media outlets: Taiwanese outlets dedicated significant frontpage coverage to the U.S. elections in the days after November 3rd, and immediately recognized Biden as President-Elect after the major U.S. news outlets projected he had won the election, unlike Chinese outlets.

In contrast to coverage in leading Chinese traditional media outlets, traditional media coverage among Taiwanese outlets consistently highlighted U.S. election stories, and never led with stories solely about the pandemic in the U.S. (the one lead story that did mention COVID, printed in the Apple Daily after the race had been called for Biden, reported that “Biden will take office with a mandate to fight COVID-19”). Coverage about the U.S. election was generally positive, such as an Apple Daily report on November 4th about the record voter turnout in the United States, and a United Daily News article on November 9th about Biden’s call for unity; such narratives did not appear in the media scan of the two Chinese publications, or in trending topics on Weibo. Similar to coverage within China, Taiwanese outlets highlighted the point when Biden was six electoral votes from winning the Electoral College, and specifically called out Nevada as the possible tipping point state. Finally, another difference with Taiwanese coverage was that all four publications acknowledged Biden as the winner and president elect in their coverage by November 9th, shortly after the U.S. news outlets projected Biden as the winner on November 7th in U.S. time, but November 8th in Taiwan. The Liberty Times lead with a story that focused on Taiwanese President Tsai Ing-wen congratulating Biden on his win. In contrast, China’s Foreign Ministry only extended congratulations to Biden on November 13th²³⁰, almost a week after Biden was projected to win, while Secretary General Xi extended congratulations on November 25th²³¹, over two weeks after Biden was recognized as the winner by U.S. media.

 

Section Conclusion

Across Weibo, one of the PRC’s most prominent social media platforms, and traditional news organization the Xinhua News Agency, a state-backed media outlet, prevailing narratives in the PRC about the U.S. elections implicitly supported the conception that the “China option” is superior to the U.S.’s liberal democratic model. Weibo trending topics and posts with election-related hashtags did not express a preference for either candidate, but emphasized chaos, disorder, and political polarization in the U.S. elections, implying that liberal democracy itself is chaotic, disorderly, and discordant. This is consistent, for instance, with a December statement from China Education News, a state-run organization, which said, “Time to wake up from blind faith in the Western system. Vicious partisan fighting has worsened in certain Western countries, social fissures have deepened, and a severe social crisis is brewing²³².”

In the week after Election Day, the Xinhua News Agency devoted more articles to the impacts of the COVID-19 pandemic on the U.S. than on updates regarding the U.S. election. This was striking because the COVID pandemic had been spreading for months, while the U.S. was facing a unique moment in that week when the election was over but the next president of the U.S. was still unknown. The focus on negative stories about the U.S. COVID response is consistent with the narrative across Chinese papers, broadcasters, and “netizens” that China’s response to COVID was far more successful than the U.S.’s response, and that this is proof that the Chinese model is more effective than the U.S. model²³³.

The comparison points provided by examining Chinese language Twitter posts with the same election related hashtags, as well as Taiwanese traditional media coverage in the week after the election, showed that prevailing narratives across those media did not fall into the pro China option pattern. Neither Twitter nor Tawianese traditional media outlets are subject to the same government controls as Weibo and the Xinhua News Agency. On Weibo, the data show that posts from state-backed organizations frequently supported the trending topics on U.S. elections, while Xinhua News Agency is, itself, a state-backed organization. The report of the leaked CCP memo designating the U.S. election as a sensitive topic and ordering media companies to follow the party’s lead in their discussion of the topic indicates that the U.S. election received significant attention. This suggests that the observed narratives in China, portraying the U.S. elections as anarchic and, in turn, implicitly emphasizing the superiority of the China option over that of liberal democracy, were directed and influenced by the Chinese state.

 

Appendices

Appendix 1

D3P monitored the media outlets listed below from November 3-5th:

• ABC News
• AP News
• Arizona Republic
• Atlanta Journal-Constitution
• Breitbart
• CBS News
• Cleveland Plain Dealer
• CNBC
• CNN
• Detroit Free Press
• Fox News
• Houston Chronicle
• Huffington Post
• Miami Herald
• Minneapolis Star Tribune
• NBC News
• New York Post
• New York Times
• Politico
• Philadelphia Inquirer
• Reuters
• Tampa Bay Times
• USA Today
• Wall Street Journal
• Washington Post
• Yahoo News

 

Appendix 2

 

 

 

 

Appendix 3

*Note that the number of outlets reviewed were reduced during the final media grab.

Appendix 4