In light of the tragic events that transpired at the US Capitol building on January 6, 2021, it is imperative to examine the effect of Donald Trump’s rhetoric on his supporters. The 2016 Trump campaign held more than 300 rallies, with attendance ranging from a few hundred supporters to tens of thousands of devotees. Fiery speeches at these rallies often used contentious language that sometimes closely mirrored narratives of prominent US white nationalists and far-right extremist groups. White nationalists frequently celebrated Trump’s candidacy and election to the presidency, including Richard Spencer, who addressed an audience of far-right supporters in Washington, DC, just days after Trump’s election, saying, “Hail Trump, hail our people, hail victory” (Lombroso and Appelbaum Reference Lombroso and Appelbaum2016).
Although organizations (e.g., Anti-Defamation League Reference Anti-Defamation2018) postulated that the substantial rise in reported hate incidents across the United States since his election in part are the result of Trump-related white-nationalist enthusiasm, empirical analysis evaluating this link has been minimal. Our research examines whether Trump’s rallies resulted in increased reported hate incidents and is more prudent than ever before for both scholars and security-oriented practitioners. Specifically, we contend that Trump’s inflammatory and divisive rhetoric, frequently used at rallies during his presidential campaign, animated white supremacists and stimulated additional hate crime and bias incidents. We used reported hate activities from 2016 and a list of all Trump rallies during this period to analyze the role that these campaign rallies had on the variation of reported bias incidents across the United States. Our research found that counties that hosted a Trump campaign rally were significantly more likely to witness a spike in reported hate incidents in subsequent months.
LITERATURE REVIEW
Presidential rhetoric and its effects have been a critical research area in political science for decades (Cohen Reference Cohen1995; Eshbaugh-Soha and Peake Reference Eshbaugh-Soha and Peake2011; Wood Reference Wood2021; Zarefsky Reference Zarefsky2004). Whereas much of this research focuses on the role that presidential rhetoric has on altering public opinion and affecting agenda setting, a growing body of scholars—particularly following the election of Donald Trump—is interested in the effects that presidential rhetoric may have on broader political processes (Rubin Reference Rubin2020; Sanchez Reference Sanchez2018; Schneiker Reference Schneiker2020). Stuckey (Reference Stuckey2020, 386) argued that to study presidential power effectively, scholars must examine not only presidential speeches and tweets but also the uptake and circulation of that rhetoric. This article builds on this assertion, suggesting that presidents who frequently and publicly use toxic discourse—particularly language that undermines key political institutions, demeans and stereotypes marginalized groups, and underscores imminent threat—are likely to affect more than only the structure of our national conversations, also resulting in changes to constituent behavior (Edwards and Rushin Reference Edwards and Rushin2018).
Trump’s norm-shattering rhetoric also has led scholars to examine how it affects his support and electoral success. Scholars examined how this rhetoric contributed to Trump’s support by underscoring how “the threat of increasing diversity” may make certain whites more likely to favor candidates who promulgate an existential risk from outgroups (e.g., Major, Blodorn, and Blascovich Reference Major, Blodorn and Blascovich2018). Lajevardi and Oskooii’s (Reference Lajevardi and Oskooii2018) findings underscored a connection between “old-fashioned racism” and support for Trump. Other scholars suggest that this rhetorical mainstreaming of fringe ideas and conspiracy theories may result in violence and the organization of pro-Trump militias to defend against perceived enemies (e.g., Barkun Reference Barkun2017).
Social-psychological theories of identity and intergroup relations demonstrate how demographic, ideological, and political realities make outgroup threats more pronounced, leading to political positions that discriminate against outgroups (McGregor, Prentice, and Nash Reference McGregor, Prentice and Nash2009; Tajfel et al. Reference Tajfel, Turner, Austin and Worchel1979). For example, Stephan, Diaz-Loving, and Duran’s (Reference Stephan, Diaz-Loving and Duran2000) integrated threat theory uses a combination of tangible (i.e., competition over resources) and symbolic (i.e., destruction of values) threats posed by increasing American diversity to produce increased white prejudice toward immigrants. This aligns with findings that considerable numbers of white Americans have diminishing bias toward minorities and a consequent increase in bias toward whites, intimating that movement toward equality of all groups comes at a specific cost to whites. Thus, the promotion of racial- and ethnic-equality ideas and policies is a threat to some whites, heightening their group identity. According to De Jonge (Reference De Jonge2016), supporters of then-presidential candidate Trump frequently exhibited this belief.
Major, Blodorn, and Blascovich (Reference Major, Blodorn and Blascovich2018) found that whites high in ethnic identification were more likely to support Trump and his anti-immigrant policies and to oppose certain political-correctness norms. Furthermore, they found that increased group-status threat and support for Trump was specific to exposed white individuals with high ethnic identification; however, no change was perceived among exposed white individuals with low ethnic identification. Although identifying the tangible effects of political rhetoric on support for candidates who promise to address concerns regarding perceived threats is critical, questions remain regarding the material effects. Studies of minority targeting indicate that whereas bias-motivated incidents and hate crimes tend to occur in predominantly white areas, they are most common when those locations experience a rapid influx of minorities (Green, Glaser, and Rich Reference Green, Glaser and Rich1998; Lyons Reference Lyons2007), thereby increasing perceived minority threat to the status quo.
Although identifying the tangible effects of political rhetoric on support for candidates who promise to address concerns regarding perceived threats is critical, questions remain regarding the material effects.
In several instances, Trump evoked negative stereotypes of racial, ethnic, and religious minority groups, speaking to the potential threat that these groups pose to white Americans. Perhaps most infamously was Trump’s proposed “Muslim Ban” (Blumenfeld Reference Blumenfeld2016). Trump rationalized this decree as necessary for national security, invoking the notion that Islam as a religion and Muslims as a people pose a unique threat to the United States. Another key incident invoking undesirable minority stereotypes was Trump’s quote referring to Mexicans: “They are bringing drugs. They are bringing crime. They are rapists” (Sanchez Reference Sanchez2015). This language often is indistinguishable from the contemporary rhetoric of white nationalists.
Not only did Trump use political rhetoric explicitly claiming that minorities are a unique threat to America; he also regularly promoted violence toward his political opponents. For example, on February 1, 2016, Trump said during a speech, “If you see somebody getting ready to throw a tomato, knock the crap out of them, would you. Seriously. I will pay for the legal fees, I promise” (Banfield Reference Banfield2016). The advocation of violence against political enemies or the ex-post justification is commonplace within hate groups, which seek nonlegal means to respond to perceived threats posed by outgroups.
Another feature of Trump’s political rhetoric was his heavy reliance on conspiracy theories, which were used to underscore the severity of the threat posed by his political opponents or groups that interfered with his version of and vision for “American Greatness.” For example, at a rally in West Palm Beach, Florida, Trump said, “Hillary Clinton meets in secret with international banks to plot the destruction of US sovereignty in order to enrich these global financial powers, her special-interest friends, and her donors” (Chokshi Reference Chokshi2016). As a chief executive of the Anti-Defamation League (ADL) observed following the rally, “Mr. Trump focuses on the very issues and themes that obsess conspiratorial anti-Semites: they believe that there is an elite group of Jews who control the media, the government, and banking and who are trying to destroy white America” (Chokshi Reference Chokshi2016). Similar to the consequences of the promotion of violence, Trump’s espousal of “deep-state” or global conspiracies to constituents purposefully undermined the use of acceptable political behavior and civic activities designed to address social concerns. When the means of the state can no longer be relied on to rebuff threats, extralegal actions such as bias incidents and hate crimes can become more justifiable, closely resembling Black’s (Reference Black1983) theory of crime as a means of justice seeking when turning to law enforcement or government is unreasonable (King and Sutton Reference King and Sutton2013, 873).
Critical to our analysis is the contention that political rhetoric—which heightens and underscores a threat to group identity—encourages violence and justifies extralegal activities with tangible behavioral consequences. Using political rhetoric to encourage outgroup hostilities is well documented, especially as it relates to economic threats posed by the outgroups (Green, Glaser, and Rich Reference Green, Glaser and Rich1998; Horowitz Reference Horowitz1985). An externalized threat to the status quo, economic or otherwise, can lead to instability and contribute to hate crimes targeting those perceived as responsible for the societal strain (Edwards and Rushin Reference Edwards and Rushin2018, 5; Hall Reference Hall, Hall, Corb, Giannasi and John2014). Karapin (Reference Karapin1999), in particular, identified the role that threat-invoking political rhetoric has in targeting immigrants in Western Europe.
Critical to our analysis is the contention that political rhetoric—which heightens and underscores a threat to group identity—encourages violence and justifies extralegal activities with tangible behavioral consequences.
Another research area vital to our study explores the link between antecedent events and hate crime and bias-incident variation. Lickel et al. (Reference Lickel, Miller, Stenstrom, Denson and Schmader2006) found that hate crimes can be triggered by specific events that would motivate the need for retributive action by the victimized ingroup toward the responsible outgroup. King and Sutton (Reference King and Sutton2013) expanded the study of the temporal clustering of hate incidents and found that highly visible criminal trials with an interracial component and lethal domestic-terror attacks result in spikes of both religious and racial hate crime. Feinberg (Reference Feinberg2020a, Reference Feinberg2020b) confirmed that larger geopolitical events can affect reported bias incidents: America’s Jewish diaspora faced increased hate crimes during violent Israeli military conflicts.
THEORY
We contend that inflammatory political speech that uses racial, ethnic, and religious tropes excuses the use of violence; promotes conspiracy theories; represents a dangerous oratorical “cocktail” that emboldens ingroup threat; and encourages bias incidents, hate crimes, and extremism. The effect of this dangerous speech is most evident when ardent supporters of the speaker meet at a discrete time and location (e.g., a political rally). Consequently, we expected counties that hosted Trump rallies, which frequently used this incendiary rhetoric, were more likely than other counties that did not host Trump rallies to witness an increase in reported bias incidents.
Several causal reasons explain our expectation that counties that hosted a Trump rally would see heightened minority targeting; however, because of limitations in this study’s data, we cannot specifically evaluate each reason. First, Trump’s political rhetoric drew on perceived threats to whites, thus heightening white identity. The notion of whites losing out to minorities was already common among many Trump supporters (De Jong Reference De Jonge2016). Having their cultural and economic fears confirmed by Trump and a community of believers may have bolstered extralegal action to combat these perceived threats posed to whites by “responsible” outgroups. These actions were further encouraged by the frequent usage of conspiracy theories suggesting that the government itself is complicit in the threats facing white Americans, making legal political behavior such as voting or reliance on law enforcement unreliable or insufficient in rectifying their concerns. Furthermore, individuals and groups who are highly sympathetic or openly white supremacist perceived Trump’s political speech as a tool for mainstreaming their political identity and aims. They often interpret the adoration of his millions of supporters as confirmation that their tactics, including those that are extralegal in nature, are succeeding.
Physically attending a rally—as opposed to listening to a report about the event covered by the national media—produces a greater sense of legitimacy for the positions being espoused by the speaker because of the many supporters in attendance combined with a greater sense of excitement as a result of the party-like atmosphere at the rallies. Media reports revealed that Trump rallies tended to follow a similar format of “spectacle” more akin to a sporting event than a political meeting (Hall, Goldstein, and Ingram Reference Hall, Goldstein and Ingram2016), where the audience is encouraged to cheer and boo as Trump “voluminously belittled and mocked his enemies in politics, government, and the press” (Parks Reference Parks2019). Furthermore, many rallies featured specific calls to action by Trump to attendees. Consequently, we expected counties that hosted Trump rallies, which frequently used incendiary rhetoric, were more likely than other counties to witness spikes in reported bias incidents. We contend that these ingredients resulted in Trump rallies serving as particularly explosive events that led to the increased reporting of hate and bias incidents. For these reasons, we hypothesized the following:
Hypothesis 1: In the months following a Trump rally, the hosting county was more likely to report an increase in hate incidents.
DATA AND METHOD
To examine whether the 2016 Trump political ralliesFootnote 1 influenced incidents targeting minorities, we used two pieces of data (Feinberg, Branton, and Martinez-Ebers Reference Feinberg, Branton and Martinez-Ebers2022). First, we used data compiled by the ADL Center on Extremism regarding hate incidents reported between January 1 and December 31, 2016. These data provide information on the city and date of incidents reported.Footnote 2 Second, we compiled information regarding the city and date of each rally that Trump held between January 1 and December 31, 2016.Footnote 3
We aggregated the incident data and Trump-rally data to the county level; thus, the unit of analysis was county–month. The aggregation of the data was intended to recognize that exposure to information and the impact of a rally likely extended beyond a singular city. To identify the county in which each Trump rally and hate incidents occurred, we used GIS to locate the county in which each city of a rally and an incident are located. Figure 1 illustrates the wide distribution of Trump-held rallies across the United States. The counties in which Trump held a rally are highlighted in blue.
Figure 1 Trump Rally Location
Note: This illustrates Trump-held rallies across the United States.
A logit analysis indicated that the Trump campaign was significantly more likely to hold a rally in an urban area, in the Northeast or Midwest, and in more affluent counties. (This auxiliary analysis is included in the online appendix, table B.) Notably, the findings indicate that county-level hate incidents and state-level hate groups were not significantly associated with the likelihood of the Trump campaign holding a rally. In essence, the urbanness and the regional location influenced the campaign’s decision to hold a rally.
The dependent variable is a count of the hate incidents that occurred in each county by month (i.e., the county–month unit of analysis) in 2016. The number of county–month incidents ranged from zero to 16, with a mean of 0.03.Footnote 4 Given the nature of the dependent variable, we used negative binomial regression to estimate the impact of Trump rallies on hate incidents.Footnote 5 The map in figure 2 demonstrates the variation in and number of hate incidents. Darker colors reflect a greater number of hate incidents. Substantively, this figure highlights the vast variability in hate incidents across US counties.
Figure 2 County-Level Hate Incidents
Note: This depicts the number of hate incidents.
The key independent variable is a dichotomous variable coded “1” if the Trump rally occurred in a county in a month and “0” if there was no rally. We used a one-month lead version of the Trump-rally variable. For example, Trump held a rally in Waukesha County, Wisconsin, in September 2016; therefore, the variable for Waukesha County was coded “1” for October and “0” for all other months. In some instances, Trump held more than one rally in a county—for example, in Fulton County, Georgia, in February and June of 2016. The rally variable for Fulton County was coded “1” for March and July and “0” for all other months.Footnote 6 The structure of the coding of the rally variable accounts for repeated rallies in the same county.
The histograms in figure 3 illustrate the distribution of reported per capita hate incidents. The top-left panel presents the histogram of hate incidents in counties that hosted a rally in the month before the rally; the top-right panel illustrates hate incidents in those same counties in the month after the rally; and the bottom-left panel illustrates hate incidents in counties with no rally. These three histograms demonstrate that the distributions are heavily right-skewed, meaning that in most months, there were zero reported hate incidents. However, a closer comparison of the panels reveals that the distributions are different. The most notable difference is in the greater presence of hate incidents in the top-right panel (i.e., rally counties, post-rally) compared with other panels. Additionally, reporting of no hate incidents is lowest in the top-right panel compared to other panels.
Figure 3 Histograms of Number of Hate Crimes
Note: This illustrates the distribution of reported per capita hate incidents.
The analysis includes several control variables: one state-level variable (i.e., hate groups) and six county-level variables (i.e., % urban, Jewish population, violent crime, property crime, partisan context, and regional dummies). We also included a state-level measure of the number of hate groups in 2016 as reported by the Southern Poverty Law Center (SPLC) (2018).Footnote 7 Because of the lack of systematic county-level data on hate groups, we used a state-level measure. The number of hate groups ranged from zero to 79, with a mean of 24.6. As the number of hate groups increased within a state, it is likely that the number of incidents targeting minority groups also increased. The analysis included a county-level measure of the percentage of the population that lives in an urban area. We controlled for the percentage urban to account for potential differences in hate incidents in urban compared to rural counties. We included a county-level measure of the estimated per capita Jewish population from the 2011 North American Jewish Data Bank and the 2011 US Census American Community Survey (Comenetz Reference Comenetz2012). The per capita Jewish population ranged from zero to 28,903.7, with a mean of 315.7. This variable was included based on the expectation that counties with a sizeable Jewish population may experience higher rates of anti-Semitic incidents. To control for incidents more generally, we included a measure of county-level per capita violent crime and per capita property crime.Footnote 8 Both crime measures were provided by the FBI’s 2015 Unified Crime Report (Federal Bureau of Investigation 2015).
To account for partisanship, we used county-level vote share received by the Republican presidential candidate in the 2012 election. The partisan measure, % Republican 2012, ranged from 7.1% to 95.9%, with a mean of 59.7%. Next, to create a measure of socioeconomic context, we used the 2015 American Community Survey population estimates for college-educated people (US Census Bureau 2015). The measure, % college, represents the percentage of the county-level population older than age 25 that is college educated, which ranged from 3 to 80.2, with a mean of 20.8. To account for regional differences in the dependent variable, the model included three regional dummy variables: South, Northeast, and Midwest (the West is the baseline category).Footnote 9 Finally, the analysis included a dummy variable for each month to control for unmeasured events that may influence hate incidents (November is the baseline category).Footnote 10
PREDICTING EXTREMIST AND ANTI-SEMITIC INCIDENTS
Table 1 presents results for the number of hate incidents that occurred by month across US counties. The negative binomial coefficients for the full sample are listed in the first column, the corresponding standard errors are in the second column, and the predicted change in incident ratio rates (IRRs) is in the third column. Note that for a robustness test, we also used coarsened exact matching (CEM) to match counties that held a Trump rally and those that did not based on county-level demographic characteristics.Footnote 11 We estimated a second model using the matched observations. The negative binomial coefficients for the matched sample are listed in the fourth column, the corresponding standard errors are in the fifth column, and the predicted change in IRRs is in the sixth column. The results based on the matched cases are substantively and statistically consistent with those for the full sample.
Table 1 Number of Hate Incidents
Notes: Standard errors are in parentheses. ∗p<0.05, ∗∗p<0.01, ∗∗∗p<0.001.
We began by examining whether the occurrence of a Trump rally affected hate incidents at the county level. The results indicated that the month after Trump held a political rally within a county, the number of hate incidents significantly increased. The predicted IRR was statistically significant and positive. Furthermore, the results indicated that there were heightened incidents in counties that hosted a rally compared to those that did not. Indeed, the results of the full and matched samples consistently indicated that counties that hosted a Trump rally experienced significantly higher rates of hate crimes when compared to counties that did not host a rally.Footnote 12
The results indicated that the month after Trump held a political rally within a county, the number of hate incidents significantly increased.
To facilitate the discussion about the impact of a Trump rally on county-level hate incidents, figure 4 presents the predicted IRRs of the lagged-rally variable for both the full and the matched samples.Footnote 13 , Footnote 14 The figure demonstrates the positive impact of Trump rallies on county-level hate incidents. Indeed, for the full sample, counties that hosted a rally were predicted to have an IRR 2.340 times higher (a 134% increase) than counties that did not host a rally. Furthermore, the predicted number of hate incidents in a county that hosted a Trump rally was 0.007 and the predicted number of hate incidents in a county that did not host a rally was 0.002 (significant difference at the 0.05 level). Regarding the matched sample, counties that hosted a rally were predicted to have an IRR 1.701 times higher (a 70.1% increase) than counties that did not host a rally. The matched results indicated that the predicted number of hate incidents in a county that hosted a Trump rally was 0.011 and that the predicted number of hate incidents in a county that did not host a rally was 0.006 (significant difference at the 0.05 level). Taken together, the full and matched sample findings supported our expectation defined in Hypothesis H1. Although hate incidents in general are a rare event, these findings support our expectation that counties that hosted a Trump rally experienced heightened levels of hate incidents.
Figure 4 Hate Crime: Incident Rate Ratios
Note: This presents the predicted incident rate ratios for the lagged rally variable for both the full and matched samples.
CONCLUSIONS
Does what a political leader says to supporters at rallies matter? This analysis shows that charged rhetoric at political rallies can produce additional hate-motivated incidents. Specifically, we established that the words of Donald Trump, as measured by the occurrence and location of his campaign rallies, significantly increased the level of hateful actions directed toward marginalized groups in the counties where his rallies were held. We contend that Trump’s political message activated attentive whites’ sense of threat and prejudice toward racial, ethnic, and religious minorities and encouraged a number of people to act on that threat. We could argue that the overall increase in hate incidents during this short period in the campaign of a presidential candidate was minimal and not worth our attention. However, we contend that the implications of our findings are considerable as the political language used by Trump at rallies becomes increasingly mainstream. Furthermore, former President Trump continues to use the same charged rhetoric even after his removal from office not only to consolidate his political influence but also to fundraise for a possible presidential run in 2024.
DATA AVAILABILITY STATEMENT
Research documentation and data that support the findings of this study are openly available at the PS: Political Science & Politics Dataverse at https://doi.org/10.7910/DVN/K2LOJN.
Supplementary Materials
To view supplementary material for this article, please visit http://doi.org/10.1017/S1049096521001621.
