Abstract
This article is a piece of a larger line of research supported by the Democracy Fund studying how to communicate about threats to elections in ways that do not dampen people’s desire to vote or make them question the integrity of electoral outcomes. It reports findings from a computerized text analysis of 2,970 open-ended survey responses in the field during the fall of 2018 to the prompt “when people say that elections are rigged, what do you think they mean?” Four key themes emerged in the data: (1) Democrats and Republicans were equally likely to regard electoral outcomes as predetermined, (2) Republicans were twice as likely to be concerned about illegal voting than Democrats, (3) Democrats were slightly more likely to be upset about money in politics than Republicans, and (4) Democrats were twice as likely to be preoccupied with Russian meddling than Republicans. A qualitative analysis of the first theme revealed both similarities across partisans as well as how Democrats focus on how threats to elections benefit people already in power, whereas Republicans worry that elections are threatened by ordinary people cheating. These findings, and the nuances contributing to them, raise new paths for research on communicating about elections without decreasing people’s faith in them.
Keywords
On October 15, 2016, Donald Trump alleged that the presidential election was rigged by the “dishonest and distorted media” and “at many polling places” (Martin & Burns, 2016). Even though other Republican officials pronounced that elections officials have made it “easy to vote and hard to cheat” (Martin & Burns, 2016), Trump stood by his claim. After his presidential victory, he tweeted that the media were “ignoring serious voter fraud in Virginia, New Hampshire, and California” and set up a task force to investigate such issues (Graham, 2019). He continued to make charges of “rigged” elections throughout his first year in office, in the days leading up to the 2018 midterms, and at his rallies in 2019 (Parker, 2019). In the weeks before the 2020 election, he further shared such concerns, repeating “the only way we’re going to lose this election is if the election is rigged” (Foley, 2020).
As the president of the United States has routinely cast doubt on the integrity of elections, how have others responded? Policy researchers at the Brennan Center for Justice conduct rigorous work disputing such claims and turn to their think tank, advocacy, and communications teams to correct misinformation about threats to elections. Political scientists show how “doubts about electoral integrity undermine general satisfaction with how democracy works” (Norris, 2019, p. 5) and connect it to lower levels of reported voter turnout (Norris, 2017, p. 43). Technology experts warn public officials, particularly Democrats, not to overreact to Trump’s charges lest they unknowingly dismiss concerns about election mechanics (Stokes, 2016). Journalists cast these accusations as an enduring Republican political strategy (Stern, 2016).
To date, much of the research on how the public interprets allegations surrounding threats to elections has employed closed–ended measures in surveys and experiments (Karp et al., 2018; Lehoucq, 2003; Norris, 2017, 2019; Norris et al., 2018). The benefits of conducting research with these “pre-established categories” (Roberts et al., 2014, p. 1064) include that they are efficient to analyze, allow scholars to employ similar metrics, offer participants a frame of reference, and do not discriminate against peoples’ verbal skills (Geer, 1988; Schuman, 1966). Yet research shows that offering people opportunities to discuss elections in their own words reveals different findings than in closed–ended studies (Schedler, 2002a, 2002b)—particularly as the language used to refer to election integrity may have wider public and “narrower technical meanings” (Norris et al., 2018, p. 24). Furthermore, individuals may “react to cues” (Iyengar, 1996, p. 64) in closed–ended survey options consistent with what the elected officials from their party advocate, even though such concerns may or may not be important to them (Beaulieu, 2014).
Driven by calls to have a more comprehensive understanding of perceptions of threats to elections (Lehoucq, 2003; Norris, 2019, 2017)—especially in light of Trump’s statements—this study analyzes how people respond to the open-ended question “when people say that elections are rigged, what do you think they mean?” This approach offers a glimpse into what participants are thinking, what attitudes and feelings are salient and top of mind, and what people offer when “un-cued by potential causes and treatments” (Iyengar, 1996, p. 64; see also Krosnick, 1999; Lazarsfeld, 1944; RePass, 1971; Roberts et al., 2014). It also offers a different way of interrogating how Democrats and Republicans, men and women, and younger Americans view the integrity of elections in the United States.
Labels as Arguments
What does it mean to use the word “rigged” in connection to elections? How might people interpret this term? This study is motivated by research contending that the act of labeling objects, phenomena, and ideas is inherently argumentative.
Labels are essential and instructive. As Walton and Macagno (2009) assert, “reality must be named and linguistically organized in order to talk about it” (pp. 81-82). They further contend that labels can be “examined as having two argumentative aspects” as they are “grounded on definitions” and “often lead to evaluative inferences” (p. 89). Zarefsky (2004) further claims that labels provide “the basis for understanding” and determine “appropriate responses” (p. 611) because labels signal how people “should view things in a particular way” (Zarefsky, 2006, p. 404).
The notion that labels have an “objective meaning” is often taken for granted (Walton, 2001, p. 122). For Schiappa (2003), these meanings “are linguistic propositions, unavoidably depend(ing) on social interaction” (p. xii). Such processes make them dependent not only on some “truth” but also on culture and identity, functioning as arguments by association. For Walton and Macagno (2009), labels can create emotionally laden processes of seeing and knowing that can “enhance positive attitudes” and “conceal” negative ones (p. 85). Many scholars agree, though, that labels can mean different things over time as definitions are disputed, move from one realm to another, and call attention to some things and not others (see Jarvis, 2005; Jarvis & Han, 2018; Schiappa, 2003; Walton, 2001).
This article treats the term “rigged” as a label with an argumentative purpose. It is interpreted through culture, emotive connotations, and values. While policy researchers, political scientists, and technology experts have weighed in on how their disciplines make sense of an election that is labeled as “rigged,” this research takes a communication approach to listen to how people make sense of this label as an argument connected to the integrity of the American electoral process.
Stances to Threats to Elections
Research shows how partisans, men and women, and older and younger voters regard threats to elections in distinct ways. To begin, partisan screens influence how people respond to various types of threats (Beaulieu, 2014). Partisan news diets—which can signal heightened or dampened attention to topics such as voter fraud—might increase feelings that electoral outcomes are unfairly influenced (Karp et al., 2018; Udani & Kimball, 2018). Partisan electoral success, too, matters as “voting for the winner . . . not only influences how a voter judges whether his own vote was counted accurately, but it [sic] also strongly influences whether a voter judges that the votes of others were counted properly” (Sances & Stewart, 2015, pp. 183-184). In addition, recent evidence suggests conservatives may be more prone to question the integrity of elections than liberals (Lamberty et al., 2018; Udani & Kimball, 2018).
Men and women also see threats in different ways. Some data suggest women voters have lower levels of trust in elections than men, even after controlling for other socioeconomic and demographic factors (Layton, 2010). Given the contentious nature of the 2016 presidential election—featuring an unpopular woman and won by an unpopular man—such findings may merit deeper examination in the United States (Zhou, 2018). Additionally, the 2018 midterm campaign featured an unprecedented number of women candidates. It is unclear if and how more women running for office going forward might influence how men and women think and feel about threats to elections in the United States.
Age, as well, may influence perceptions of threats. Younger voters consume news differently than their older counterparts and are more likely to engage in fact-checking behaviors (Funke, 2018). Those younger than 29 years of age are more likely to get their news online, including from social media, rather from legacy media (Mitchell et al., 2016) and 57% of those who get their news from social media expect what they see there to be largely inaccurate (Matsa & Shearer, 2018). Younger voters, too, have more actively supported the Democratic party over the past few election cycles (Center for Information & Research on Civic Learning and Engagement, 2018). This predisposition, in combination with their news diets, may influence perceptions of threats. Consider these data comparing young voters who supported Hillary Clinton versus Donald Trump in 2016. Young Clinton supporters “were much more likely to say they were losing faith in American democracy than those who voted for Trump: 72% vs. 16%” with “youth who voted for Trump as much more likely to say that the United States government would improve over the next four years compared to those who voted for Clinton, 88% to 21%” (Center for Information & Research on Civic Learning and Engagement, 2019).
Method
This study uses computational analysis to identify dominant themes in open-ended responses to the question, “when people say that elections are rigged, what do you think they mean?” The question appeared in a population-based survey run by Survey Sampling Inc. (SSI) in the field from October 11 to 15, 2018. The median age was 47. Forty-six percentage of the sample were men and 53% were women. Thirty-two percentage identified as Republican, 33% as Democratic, 29% as Independent, and 6% as other. Among Republicans, 59% identified as strong and 41% identified as not strong. For Democrats, 59% identified as strong and 41% as not strong.
There are many computer programs available to conduct quantitative analyses of texts. For this project, we chose MALLET, a topical modeling program, due to its features, prominence, and open-source status (see http://mallet.cs.umass.edu/topics.php). Topic modeling is an algorithm-based procedure whereby a text is mined for themes expressed as groups of co-occurring words (Brett, 2013). These linguistic neighbors can provide insight into the most common contextual structures of the responses (Blei, 2013). By treating open-ended survey responses as a “collective system” rather than an amalgamation of disparate individuals (Moretti, 2003, p. 68), we can pull out recurrent patterns, interpreted as emergent themes.
All 2,970 open-ended responses were cleaned of spelling errors, abbreviations, or other such inconsistencies and saved as plain text (.txt) files. Files were created for the full set of responses and then specific subsets for Democrats, Republicans, men, women, and young voters (born between 1981 and 1996, see Pew Research Center, 2018). All files were imported to MALLET, during which sequencing is preserved and stop words such as “the” and “and” were deleted. In addition to creating “a cluster of words that frequently occur together,” MALLET employed context to “connect words with similar meanings and distinguish between uses of words with multiple meanings” (University of Massachusetts, Amherst, 2018).
The topic modeling program optimizes models and ensures a closer fit to the data by allowing topics to be weighted based on prominence (University of Massachusetts, Amherst, 2018). Through an iterative process, manually checking the distribution of topics across the data, the model was set to generate 15 topics. For parsimony, MALLET was configured to provide the 20 most frequent words in each topic. The program also generated weights of the most common co-occurring words. Four of the 15 topics, strings of words, emerged as dominant, accounting for 80% to 90% of the data across subgroups. The remaining themes each constituted 5% or less of the total or subdivided data. The MALLET generated data for the four main themes were imported into spreadsheets for readability so that the authors could make sense of each of the sets of 20 co-occurring words. After reading them closely, and going back into the open-ended responses for context, we identified four dominant themes (see Schmidt, 2013).
After reviewing the four key themes coming from the computational process, a qualitative analysis was conducted on the first theme as it was the most common for all groups in the analysis (Democrats, Republicans, men, women, and young voters). Responses that constitute the first theme were identified by selecting all those that contained more than one word from the 20-word list. The goal with the qualitative analysis was to read closely for nuances that were shared—or distinct—across the subsets of the sample.
Here, it is worth including a note on the subjectivity of the responses. The survey item asked “when people say that elections are rigged, what do you think they mean (italics added)?” The question indicates that other individuals, not the respondent themselves, are making the rigged argument and asks the respondent to interpret these claims through the intent of the speaker. Many reflect this construction without affective commentary by writing that “they believe there are forces (italics added)” or that “I think a lot of people equate rigged with (italics added).” Another group of respondents answer for the subject “they” in ways that assert how “they” is different or distinct from the participant, “I.” For example, one respondent wrote “I think they’re just repeating talking points that they’ve heard others say and they have no idea what it actually might mean (Personally, I think it means that there are . . . ).”
Not all such responses make the division so precise. Rather, some create opposition through emotion like the participant who wrote “they are bitter because their side didn’t win,” implying the writer is not bitter because their side did win. A third group of responses makes no reference to the other “they” or to the personal “I” and could be interpreted with either subjectivity. When a participant offers that “it means someone manipulated the system or the wrong person got the most votes,” it is unclear if this is the projected belief of another or the personal belief of the individual respondent. Yet another group explicitly responds from their perspective, writing with their belief in (“I question the validity of the vote count”) or rejection of (“I think that would be very difficult to do”) the idea that elections are rigged. Both the quantitative and qualitative findings that follow reflect these variations in response subjectivity. Again, the goal is to learn more about how people, in their own words, articulate how the label “rigged” serves as an argument in conversations about elections in the United States.
Findings From the Quantitative Analysis
Table 1 features the dominant themes from the quantitative analysis. As illustrated below, when asked to describe what they think people mean when they say an election is “rigged,” individuals discussed (1) predetermined electoral outcomes, (2) illegal voting, (3) money in politics, and (4) Russian meddling.
Thematic Comparisons Across Groups (in Percentages).
Note. MALLET will display the top words per topic in a quantity specified by the model creator. For the sake of parsimony, the 20 most frequently occurring words for each topic are listed. These themes compose 88.77% of the total corpus of survey responses and between 81.80% and 88.57% of subgroup responses, showing a clear dominance over the discourse.
The first theme addresses frustration that elections are predetermined. As Table 1 displays, it is the largest theme overall (and the largest theme for each subgroup). The words contributing to it, in order of dominance, are as follows: votes, people, win, don’t, elections, matter, candidate, outcome, person, ballots, count, results, make, means, decided, didn’t, electoral, set, and aren’t. Listen to how individuals in the study used these words. One respondent argued calling an election rigged means “that individual votes don’t really matter. The political parties will put whoever they want into office.” Another person explained, “someone set up the ballot for a certain candidate to win.” A third individual offered that “votes [are] not counted properly. Outcome [is] set before [the] election.” These responses use different combinations of the topic’s words, but all reflect a feeling of predetermination—signaling how electoral participation does not matter because outcomes are predestined, already decided, and set in advance. In a sense, this sentiment could be interpreted in line with Lehoucq’s (2003, p. 233) broad definition of “electoral fraud as clandestine efforts to shape election results.” Participants in this first theme offer a vision of U.S. elections where the votes of Americans do not matter because elections are not decided by ballot counts.
The predetermined theme is the only one in this study that does not include specific referents to the 2016, 2018, or 2020 elections. The absence of such language is notable. On one hand, it could connect to the nature of the question. Because the prompt was phrased, “when you hear people say that ‘elections are rigged,’ what do you think they mean?” many people may have taken the question to solicit a broad response—and they offered one. Alternatively, these responses could be related to political knowledge. Many Americans may not know much on this topic, and, absent the cue of a closed–ended prompt, they responded more broadly than would those with more detailed information about empirical threats to the mechanics of elections (Norris et al., 2018, p. 12). Furthermore, young voters were the most likely to appear in this theme. Research routinely shows younger Americans possess lower levels of political knowledge than their parents and grandparents (Kaid et al., 2007). Regardless of age, though, it is intriguing that the most common response to the prompt addressed a sense of power and not the politics surrounding or the safety of elections. Across all groups in Table 1, then, the dominant sense of what “rigged” means is that the power to decide elections resides with others.
The remaining themes have stronger connections to partisan differences and specific threats to elections in 2016, 2018, and potentially 2020. The second theme addresses concerns about illegal voting. The words contributing to it, in order of dominance, are as follows: vote, voting, counted, party, fixed, illegal, voters, outcome, voter, big, paid, lost, dead, fraud, illegals, voted, favor, illegally, parties, and outcomes. Qualitative examples of this theme attest how voters might be paid (“people are paid to vote for a certain person/party”), votes might be cast on behalf of the deceased, noncitizens or other undocumented people (“dead people and illegal aliens are voting”), and ballots could be purposefully lost (“the ballots are not counted properly and the votes for a party are lost or not counted”). This theme appeared in more Republican (29.1%) than Democratic (14.61%) or younger voter (12.7%) responses. These patterns are consistent with research on the demographics supportive of voter identification laws passed by Republican state legislatures to purportedly improve public confidence in elections (Bowler & Donovan, 2016; Doherty et al., 2018).
The third theme focuses on apprehensions about the role of money in politics. The words contributing to it, in order of dominance, are as follows: election, money, rigged, candidate, influence, results, cheating, predetermined, candidates, process, trump, gerrymandering, means, foreign, they’re, manipulated, system, political, things, and count. Here, the rank ordering of the words is important. Money appeared right after the term election (a word that was in our prompt and was repeated by many respondents in their open-ended answers). Qualitative examples of these answers tied money to unfair influence affecting results via responses like one claiming there is “too much influence from big money corporations and personal friends.” Others link money to other unjust electoral processes (e.g., gerrymandering) and wrote responses indicating “their [sic] is too much gerrymandering that has taken and their [sic] is too much influence from individuals and corporations with big wallets.” There are also responses tying money to various forms of cheating (e.g., foreign help to candidate Trump). Trump’s appearance in the theme is the only incidence of a specific person rising to prominence in the data.
A final theme in the data attends to distress over Russian meddling. The words contributing to it, in order of dominance, are as follows: vote, government, winner, it’s, voting, doesn’t, win, Russia, interference, republicans, power, changed, Russian, counts, counted, tampering, favor, popular, fair, and president. Despite being the theme of lowest salience in the overall model, it is the most straightforward. Responses include “the Russians [sic] interference to get Donald Trump elected POTUS” and “Russians and the president lying.” As other projects would predict, Democrats (17.9%) invoked this theme more than twice as often as Republicans (7.5%), who often did so in varying degrees of negation (e.g., “it means that they still think Russia hacked our election so Donald Trump won. THEY ARE STUPID BRATTY BASTARDS WHO NEED TO SHUT . . . UP!”; Tyson, 2018). Younger voters (20.4%) showed the greatest use of the topic among the subgroups. It is possible that younger voters’ use, knowledge, and informed skepticism about technology, online security, and social media ad buys—combined with their general support of Democratic candidates—contribute to notions of Russian interference (with mentions of how Republicans were denying such a possibility) being salient (Pew Research Center, 2018).
Findings From the Qualitative Analysis
Given that the “predetermined” theme was the most common concern across all groups, we conducted a qualitative analysis of responses contributing to it. The most notable findings appeared in the similarities and differences across the parties.
A central similarity shared by Democrats and Republican addressed how respondents believed that those calling elections “rigged” were expressing a concern that outcomes were decided long in advance. Indeed, members of both parties shared this worry in equal amounts and in similar words. One Democrat offered that “a candidate has already been predetermined to be the winner before any votes are cast.” Other Democrats wrote how “rigged” elections “are fixed to help a certain candidate win,” or that “results are decided ahead of time,” or that “generally, I think they mean ‘there’s no point in voting because my individual vote won’t make any difference to the outcome.’” Similarly, a Republican suggested that “the winner is picked before the election and no matter what the vote is it won’t matter.” Other Republicans explained “rigged” as meaning “the winner is determined before the votes are counted” because there would be a “set outcome,” or the “results given are not an accurate representation of votes cast.” Taken together, these responses point to a bipartisan concern that elections are “rigged” because the predetermined outcomes are outside of voters’ hands.
A second theme highlighted a difference between the parties. Namely, many Democrats expressed how a “rigged” election would benefit political elites, whereas Republicans felt that other citizens may have cheated—leading to an unjust outcome. Consider the examples of how Democrats voiced frustrations with “people in power” benefitting from “rigged” elections. They referenced the Electoral College, the Citizens United Supreme Court Case, the Republican Party, gerrymandering, voter suppression, or other purportedly undemocratic bodies or processes—linking such concerns to benefitting elites. Specifically, they addressed how “I think a lot of people equate rigged with their vote not mattering, especially in presidential elections . . . with gerrymandering and Citizens United,” how “the electoral college all have their hands in the elections to control their money! The top 1% control everything!” and how “Republicans make it hard for minorities and others to vote by coming up with all of these ridiculous regulations.” Comments that fit the “predetermined” theme for Democrats highlighted a sense that elections were out of people’s control because they were in the hands of elite actors with access to resources outside of the grasp of “everyday voters.”
In contrast, Republican responses placed agency with “everyday people” trying to game the system. If the Democrats, above, were beleaguered by systemic pressures they might not be able to surmount, the Republicans in this sample seemed to identify opponents who were cheating and whose (purportedly) illegal votes must be countered. Republicans defined “rigged” elections as when “people vote under different names,” or “illegal votes being counted from people that are not registered to vote like illegals [sic].” Sometimes, Republicans reported, “rigged” predetermination was possible because of flaws in the system. One Republican wrote that mostly I think that there are many phantom votes. Mail in votes for people who are dead or the votes are being done by an ineligible voter are the most common. Since we do not need an ID for voting many of these people are easily accepted as the person voting.
Republicans were also concerned about poll workers—unelected individuals supervising local polling places. These Republicans worried that “rigged” might mean “adding up false votes to elect a specific candidate” or switching or throwing “away ballots in order to make one candidate win.”
Conclusion
This research is part of a larger project examining how to communicate about threats to elections in ways that do not dampen people’s desire to vote or make them question the integrity of electoral outcomes. A concern connected to the 2018 and 2020 elections, for many, has been how to respond to President Trump’s charges that elections are rigged in the United States. This article employed an open-ended survey prompt in the fall of 2018 to ask 2,970 people what they think accusations of “rigged” elections mean. Knowing more about this topic opens the door for the construction and testing of ways to talk about this concern.
A topic modeling program identified four key themes in the responses. The most common theme related to the notion that electoral outcomes are predetermined. Democrats (32.1%) and Republicans (33.8%) voiced this concern at equal rates. This reaction was the most common for all groups examined and particularly salient for women (38.1%) and younger voters (40.9%). Future research can examine if and how this response connects with other political variables. It can also continue to explore what it means that the most top-of-mind response to our prompt has little to do with election mechanics, themselves, and more to do with a perceived lack of power, agency, and influence in the electorate. Knowing more about such matters will aid the development and testing of messages to advocate for greater trust in elections.
The other three themes are consistent with prior projects documenting partisan differences in perceptions of threats to elections. Republicans (29.1%) were twice as likely to be worried about illegal voting than were Democrats (14.6%). Democrats (22.7%) were more likely to be frustrated that “money in politics” may have been used to cheat in the 2016 election and help Donald Trump than were Republicans (15.1%). That “money in politics” was salient to men (21.2%) merits further investigation, especially given the men’s support for Trump in the 2016 election. Finally, Democrats (18.0%) were more upset about Russian meddling than were Republicans (7.5%).
Furthermore, in the qualitative analysis, both Democrats and Republicans mentioned notions of power in their responses that elections are “rigged” because they are predetermined. In one sense, members of both parties expressed a concern that electoral outcomes were out of their hands. In a more nuanced sense, there was also a difference in how partisans saw power operating. Returning to labels as arguments, our assessments of the qualitative data lead us to believe that when Democrats are asked to interpret what it means to say that elections are “rigged” because they are predetermined they see it as out of voters’ control, whereas when Republican think about what it means to call an elections “rigged,” it is a matter of other voters behaving irresponsibly and whose behaviors could be countered by more registered Republicans voting.
There are strengths and limitations to this type of project. Open-ended prompts offer a view into what people are really thinking and worrying about. They have the advantage of nonreactivity as they do not cue answers. They can also signal a sense of salience for the respondent. It is important to note, however, that such measures are influenced by people’s verbal abilities and willingness to respond to them. Individuals and groups advocating for the integrity of elections are well aware of our second, third, and fourth themes regarding partisan concerns with illegal voting, money in politics, and foreign interference.
A label like “rigged” can carry powerful arguments. These data suggest that the argument “rigged” holds most strongly is one of power. Both Democrats and Republicans described this word as signaling how “others” decide elections. Donald Trump made repeated claims about elections being rigged in the 3 years leading up to our data collection and continued to do so heading into the 2020 election. It is notable that most Americans, when answering an open-ended prompt, did not put a specific elite face—like Trump’s—on the term “rigged.” Even Republicans did not echo a sense that their President, specifically, might be targeted unfairly or that an election benefitting the other side would be “fake.”
That stated, the President has the power to define (Zarefsky, 2004) but often that definition, or label, is subject to shifting interpretations. It is crucial to be mindful that labels can change based on the voices that use them most regularly, strategically, and passionately (Jarvis, 2005). Going forward, scholars can keep an eye on what voices are discussing the term “rigged” and how it is recirculated in the American political conversation. Moreover, researchers can be mindful of the broader sense that members of both parties feel disempowered and may be vulnerable to future persuasive efforts targeted to their sense that “others” decide American elections.
Footnotes
Declaration of Conflicting Interests
The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.
Funding
The author(s) disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: This research was supported by the Democracy Fund. Study number 2017070082A, IRB 2018-09-0034, University of Texas at Austin.
