An exploration of Donald Trump's allegations of massive voter fraud in the 2016 General Election☆
Introduction
Regular and fair elections are the keystones of democratic governance (Lipset, 1959; Katz, 1997). These mechanisms translate voter preferences and opinions into elected officials, who ultimately make policy. Electoral fraud distorts the relationship between constituents and representatives, and for this reason alone the threat of voter fraud is inherently serious. Moreover, elections perceived as unfair can decrease electoral legitimacy (Norris, 2014), reduce governmental credibility (Magaloni, 2010), and undermine perceptions of voter efficacy (Elklit and Reynolds, 2002).
Insofar as it was repeatedly tarred by allegations of widespread voter fraud, the 2016 American General Election exemplifies these concerns. Despite a dearth of evidence that fraudulently cast ballots play an important role in American elections (e.g., Levitt, 2007, Minnite, 2010, Goel et al., 2016), as the Republican nominee for president Donald Trump claimed that he was at risk of losing the presidential contest to Democratic rival Hillary Clinton because of systematic voter fraud. Later as president-elect, Trump asserted that Clinton had received “millions” of improper votes, and he blamed his loss of the popular vote on illegal activity. And finally, as the 45th President of the United States, Trump asserted that voting in New Hampshire was tainted by fraud and that, in the absence of illegal Massachusetts voters, Trump would have won the Granite State's four electoral votes and then-United States Senator Kelly Ayotte, who lost a close election to former New Hampshire governor Maggie Hassan, would have been reelected.
Trump's expansive claims merit attention because of the role that elections play in democratic politics and on account of Trump's status—45th President of the United States. Moreover, assertions of voter fraud are a significant source of political division and conflict in American politics (Hasen, 2012, Bentele and O'Brien, 2013, Hicks et al., 2014), and they are believed by a non-trivial segment of the voting population (Ansolabehere and Persily, 2008, Stewart et al., 2016). Lastly, simply because there was little voter fraud prior to November, 2016, does not imply perforce that Trump's claims are necessarily vacuous; it is always possible that 2016 was the first year in which systematic voter fraud was a meaningful factor in a presidential contest. These points motivated us in mid-2016 to develop an election fraud research project premised on the question, what could we academics say about election fraud in the aftermath of the then-upcoming presidential election? Our concern as of the summer of 2016 was that Trump might suffer a close loss in his bid for the presidency and react by leveling widespread accusations of voter fraud that, in principle, could explain his defeat at the polls.
Given the tenor of the Clinton-Trump presidential contest at the time of the Republican and Democratic party conventions, we anticipated post-election fraud allegations that centered on illegal voters, in particular non-citizens. To prepare ourselves to scrutinize such allegations, we assembled a county-level dataset that included historical election returns, demographics, and economic indicators. We also contracted with the Associated Press so that we would be able to access their national database on county presidential election returns. Our plan was to begin work on fraud allegations on Election Day evening (November 8, 2016), and we were prepared for an intense post-election week or two.
Since its inception, our research project has evolved in reaction to two developments. First, Trump did not lose the 2016 presidential election; this relieved us of the pressure to investigate fraud allegations made in the aftermath of a close Trump loss. Second, and seemingly in spite of his victory, Trump continued to invoke the specter of widespread voter fraud. This latter development has spurred on our project, the result of which is this article.
We would like to draw particular attention to our use of the term, “widespread,” in the sense of what we are calling allegations of widespread voter fraud. Donald Trump, as candidate and then later as president, has not anchored his voter fraud claims on the likelihood of a person, here or there, voting illegally.2 Rather, Trump and key supporters have spoken literally of “millions” of illegal votes, as our introductory quote makes clear. With this as context, our research project, an attempt to introduce scientific rigor into a debate largely dominated by bombastic claims, is not aimed at ferreting out what one might argue are more minor instances of voter fraud. While all instances of voter fraud are troubling, not all frauds are pivotal and not all frauds are systematic and widespread. Our research focuses solely on the possibility of massive and systematic fraud because fraud of this type in principle had the potential to be pivotal to the 2016 presidential election and because this is precisely the type of fraud against which Trump and his supporters, both before and after November, 2016, have regularly inveighed.
One can think of the analysis that follows as the proverbial canary, one that is an appropriate yet far-from-final step on the path of testing for voter fraud in the 2016 General Election. Detailed, individual-level audits, conducted on random samples of voters across jurisdictions spanning the United States, might be the ideal method to test for instances of voter fraud. However, in the absence of such audits, our analysis of aggregate county voting represents a valuable start. As will be clear shortly, we leverage variation in election outcomes across thousands of counties and connect that variation to a litany of explanatory variables, including counts of non-citizens provided by the American Community Survey. In the absence of a very expensive—and possibly unfeasible—audit of voter lists in jurisdictions across the United States, we believe that our aggregate analysis provides a significant advance in testing claims of voter fraud.
One could argue that an alternative method for testing voter fraud allegations would be to leverage a large-scale survey that questions respondents about, say, citizenship status and voting history. Such a survey would have the benefit of assessing the eligibility of voters individually as opposed to in the aggregate. However, unlike an audit, a survey in this vein would depend on the accuracy of the information volunteered by its respondents. This dependence is exemplified by Richman et al. (2014), who use the Cooperative Congressional Election Study to analyze the voting behavior of self-identified non-citizens; drawing on survey data, they estimate that 1.2 million non-citizens voted in the 2008 General Election. Ansolabehere et al. (2015) show, however, that this estimate reflects respondent data errors. Our use of aggregate data in conjunction with a corresponding lack of dependence on surveys allow us to avoid the sort of response problems that confound Richman, Chattha and Earnest.
We consider three allegations of voter fraud in the 2016 General Election: participation across the United States by non-citizens who supported Hillary Clinton in her presidential bid; concerns about voting in three states, California, New Hampshire, and Virginia, with particular attention to the possibility that Massachusetts voters tampered en masse with the United States Senate election in New Hampshire; and, finally, a conspiracy of election officials who attempted to “rig” the presidential election against Trump. The voter fraud accusations that we examine here span both national (non-citizen voting) and state-specific (e.g., New Hampshire), and all are associated with Donald Trump.
Briefly, we find little evidence consistent with widespread and systematic fraud fomented by non-citizens. Our analysis of returns in California, New Hampshire, and Virginia likewise turns up no evidence of problems in the vein raised by Donald Trump. And, our closer look at New Hampshire also yields nothing concrete. Lastly, and keeping in mind that the concern about a “rigged” election is ambiguous—we operationalized this idea by considering patterns in the way that election returns were released starting on the evening of November 8, 2016—we find no suspicious patterns in result timing.
Our results do not imply that there was no fraud at all in the 2016 presidential contest; indeed, we already know that the rate of fraud in the 2016 presidential election was not literally zero.3 Nor do our results imply that the administration of the 2016 General Election was error-free. Nonetheless, they do strongly suggest that Trump's voter fraud allegations are not grounded in any observable features of the 2016 presidential election.
This article proceeds as follows. In the next section, we provide additional details on the motivation for and development of the research project whose results are described here. We then consider the aforementioned three sources of voter fraud, and we present results on them in sequence. Our final section concludes with suggestions for future research and how the academic community might want to consider studying voter fraud in upcoming elections.
Section snippets
Studying voter fraud: motivation and project development
Allegations of voter fraud in the 2016 General Election from the Trump camp were alarmingly common. Beyond those already mentioned, Donald Trump and officials allied with him have asserted that records of deceased individuals are regularly used in the commission of voter fraud and insinuated that some urban areas, specifically Chicago, Philadelphia, and St. Louis, are hotbeds of fraudulent voting.4
Results
Our results are in three sections. First, we offer a county-level analysis that addresses Donald Trump's claims about non-citizen voter fraud; allegations about non-citizens were promulgated pre-election, and we highlight the possibility of non-citizen voting in California, New Hampshire, and Virginia, three states which were mentioned explicitly by Trump post-election. Second, we continue our focus on states by analyzing a specific, post-election claim about New Hampshire. And third, we
Conclusion
We have presented here a number of analyses that speak to claims about widespread and systematic voter fraud in the 2016 United States General Election, claims promulgated by Donald Trump both before he became president and after his inauguration. These claims include allegations about non-citizen voter fraud, allegations specific to three states, and a national conspiracy of election officials. Our empirical results share a common theme: they are inconsistent with fraud allegations made by
References (31)
- et al.
The perils of cherry picking low frequency events in large sample surveys
Elect. Stud.
(2015) - et al.
Do non-citizens vote in US elections?
Elect. Stud.
(2014) - et al.
New Faces, New Voices: the Hispanic Electorate in America
(2010) - et al.
Vote fraud in the eye of the beholder: the role of public opinion in the challenge to voter identification requirements
Harv. Law Rev.
(2008) - et al.
Presidential vote models: a recount
PS Polit. Sci. Polit.
(2001) - et al.
What the numbers say: a digit-based test for election fraud
Polit. Anal.
(2012) - et al.
Jim Crow 2.0? Why states consider and adopt restrictive voter access policies
Perspect. Polit.
(2013) - et al.
Election day registration's effect on US voter turnout
Soc. Sci. Q.
(2001) Identifying irregularities in Mexican local elections
Am. J. Polit. Sci.
(2014)The impact of political geography on Wisconsin redistricting: an analysis of wisconsins Act 43 assembly districting plan
Election Law J.
(September 6, 2017)
Presidential voting and the local variability of economic hardship
Forum
Identifying election fraud using orphan and low propensity voters
Am. Polit. Res.
The impact of election administration on the legitimacy of emerging democracies: a new comparative politics research agenda
Commonw. Comp. Polit.
Predicting Presidential Elections and Other Things
One person, one vote: estimating the prevalence of double voting in U.S. Presidential elections. Working paper
Cited by (34)
The 2020 presidential election and beliefs about fraud: Continuity or change?
2021, Electoral StudiesCitation Excerpt :Given the patterns in Fig. 1, we suspect that while beliefs that the opposing candidate is cheating are largely stable, they may be impacted by elite cues and particular political circumstances. In 2012, voter fraud accusations were not levied by prominent elites; in 2016, however, then-candidate Trump repeatedly claimed that the election would be rigged, likely inflaming concerns among Republicans (Cottrell et al., 2018). However, Trump's competitor, Hillary Clinton, along with other Democratic elites, consistently argued that the election would not be fraudulent; this likely depressed fears of fraud among Democrats in 2016 (Stewart 2016).
Multiple meanings? The link between partisanship and definitions of voter fraud
2021, Electoral StudiesCitation Excerpt :Exploring the meaning of voter fraud has important implications for existing research on fraud perceptions. Many studies have only examined certain types of fraud, specifically defining fraud as a small set of illegal activities related to voting (Ansolabehere et al., 2015; Bowler and Donovan 2016; Cottrell et al. 2018; Udani and Kimball 2018). They also tend to focus on a narrow set of allegedly “bad actors,” namely non-citizens.
Confidence in an election authority and satisfaction with democracy: Evidence from a quasi-natural experiment of a failed election in Sweden
2020, Electoral StudiesCitation Excerpt :However, little attention was paid to how the rerun affected the voters’ confidence in the electoral process. In addition, the 2016 U.S. presidential election prompted investigations into electoral integrity after then President-elect Donald Trump claimed that massive voter fraud had been committed in order to thwart his nomination (Cottrell et al., 2018). Recent research on how the U.S. citizens viewed the 2016 election found that electoral integrity affected general satisfaction with democracy (Norris, 2019a).
Who Buys the Big Lie? White Racial Grievance and Confidence in the Fairness of American Elections
2023, Journal of Race, Ethnicity and PoliticsAccountability for the Local Economy at All Levels of Government in United States Elections
2020, American Political Science ReviewOne Person, One Vote: Estimating the Prevalence of Double Voting in U.S. Presidential Elections
2020, American Political Science Review
- ☆
The authors released an early version of this article on the Internet on November 29, 2016. We thank Hollye Swinehart and Morgan Waterman for research assistance; Rich Houseal of Research Services, Church of the Nazarene Global Ministry Center, for religious affiliation data; Dwayne Desaulniers of the Associated Press for assistance with 2016 election returns; three anonymous referees; and, seminar participants at the Freie Universität Berlin and the Hertie School of Governance for helpful comments. This academic research project is not affiliated with any of the candidates in the 2016 General Election, and none of the authors received external funding for the work described here. Herron thanks the Hertie School of Governance in Berlin for support while he was writing this article.