“I Am the Law and Order Candidate”: A Content Analysis of Donald Trump’s Race-Baiting Dog Whistles in the 2016 Presidential Campaign ()
1. Introduction
“Our cruelest adversaries are not those who contradict and try to convince us, but those who magnify or invent reports which may make us unhappy…and perhaps give us some slight regard for a party which they make a point of displaying to us, to complete our torment, as being at once terrible and triumphant.” —Marcel Proust, The Guermantes Way “The common thread linking the major Islamic terror attacks that have recently occurred on our soil…is that they have involved immigrants or the children of immigrants” —Donald Trump, Youngstown, Ohio, August 15, 2016
As we near the end of the presidency of Donald Trump, the reasons for his surprise victory over Democratic nominee Hillary Clinton and continued success attracting loyal supporters remain unsettled for many. In the aftermath of the 2016 election, explanations for Trump’s success usually aligned with several main themes. One of these themes was economic factors, such as Trump’s appeal to economically anxious White, blue-collar workers in the area known as the Rust Belt (Abernathy, 2017; Sargent, 2017; Saul, 2016; Schiller, 2016). Several explanations can be categorized as crediting shifts in voter turnout that favored Trump such as lower enthusiasm among Black Democrats and higher turnout among the less educated (Cohn, 2017; DeGroot, Emamdjomeh, Menezes, & Pesce, 2016; Fraga, McElwee, Rhodes, & Schaffner, 2017; Lauter, 2016; Silver, 2016). Finally, writers and television pundits often blamed Clinton’s loss on the failures of the Democratic party to appeal to Americans living outside ascendant, densely populated cities (Trende & Byler, 2017).
Those explanations, which either ignored the impact of race or minimized it, felt sterile in light of the series of controversial events at Trump rallies and campaign stops (Mathis-Lilley, 2016), including clashes between groups of protesters (Saunders, 2016), racist chants (DelReal & Sullivan, 2016), and physical violence like the Black man assaulted by a White Trump supporter (Parker, 2016). The seeming focus of public attention on the conventional aspects of Trump’s upset win, such as the economy, raises the question of how to account for the strong reactions Trump seemed to inspire in supporters and his opposition, as summarized above. One branch of analysis concludes that the voting shifts so crucial to Trump’s victory can be attributed to some form of racism (McElwee, 2016; Schaffner, MacWilliams, & Nteta, 2018; Valentino, Neuner, & Vandenbroek, 2018) and the related politics of status threat (Mutz, 2018). The status threat was defined by Richard Hofstadter (1965) as anxiety about perceived loss in status, such as White voters believing that they are losing ground in some way to minorities. It is worth questioning how Donald Trump was able to connect with voters higher in, for example, racial resentment (reviewed in McElwee, 2016 and studied extensively in Jardina, 2019) or those who feel they are losing ground to minority groups.
I propose that, during the 2016 campaign, Trump appealed to these voters by using specific language in his speeches to cue them that he was on their side. There is a lengthy history of coded race-baiting in politics (Hanley López, 2014; Mendelberg, 2001). Ian Hanley López calls these coded, racial appeals dog whistles. The term “dog whistle” implies that only those meant to understand and react to it (i.e. the “dogs”) do, while others do not. Such obfuscation is one goal of dog whistling, but it is in service to another goal: providing cover to the race-baiting politicians and their supporters (Hanley López, 2014). For the purposes of this paper, unless stated otherwise, a “dog whistle” is a race-baiting term or phrase that implies, but does not state, a connection to one or more minority groups.
Evidence that the winner of the 2016 election race baited in his campaign would also align with research showing the power of elite framing, especially uses of race-related terminology. Trump’s presidency has been concurrent with a deepening partisan divide during which “team-oriented” negative partisanship has become a powerful political motivator (Marietta & Barker, 2019). Such polarization, especially when focused on racial issues, tends to be emotionally driven and self-perpetuating (Ioanide, 2015; Phoenix, 2019)—especially when fed by some of Trump’s statements on race as president. Evidence of Trump’s dog whistling would help explain how Trump connected with the voters found to be vital to his win in 2016, thereby supplementing and enriching the findings of researchers who found that racism and status threat played a significant role in Trump’s election. We know that the racially resentful and those threatened by the rise of American minorities were more likely to support him (e.g., Jardina, 2019; Mutz, 2018, Schaffner et al., 2018; Valentino et al., 2018); the evidence in this paper shows what Trump did to activate that support.
A shift from direct race baiting in Trump’s candidacy announcement, which drew immediate scorn from politicians and potential voters (Allen, 2015; Gabbatt, 2015), to a dog-whistle approach would have provided valuable plausible deniability both to the campaign and the millions of Trump supporters—especially the White voters—who, presumably, would not consider themselves racists. This is necessary because White people tend to have negative emotional reactions to examining their racial beliefs and, often, to even discussing race (DiAngelo, 2018; Neville, Awad, Brooks, Flores, & Bluemel, 2013; Sue, 2011). The value in Trump’s dog whistling would be his visceral connection to some Americans while potentially avoiding the defensiveness those Americans might feel about their own racism if responding to an open racial appeal.
Certainly not all White people voted for Trump or are considered part of his “base” but, in the tradition of how race baiting operates, dog whistling targets White people. The documented influence of racism and status threat in 2016 (McElwee, 2016; Schaffner et al., 2018; Valentino et al., 2018), in conjunction with the history of race baiting, demonstrates that some White people will respond to emotion-stoking, racial appeals. Dog whistling politicians calculate that the risk of backlash, lessened by obfuscated terminology, is worth the potential reward of motivating enough voters to get elected. What is unique to this paper is the evidence of race baiting in the words of Donald Trump, words likely to provoke the kind of emotional, and sometimes violent, reactions to him during his campaign—words that played a role in building a loyal base of supporters and in eventually getting him elected president. I further demonstrate that Trump’s dog whistling is not typical of a modern politician; his use of coded racial appeals significantly and substantially exceeds that of comparable Republican presidential nominees.
There has been little formal analysis of Trump’s words during the 2016 campaign: Trump’s race baiting and dog whistling have been taken as settled (e.g. Capeheart, 2017; Cepeda, 2018; Gilchrist, 2018; Pandolfo, 2018), excepting limited analysis by Hanley López (2016). Often left unexamined are the following questions: What specific terms count as dog whistles, why is dog whistling effective, what proof (i.e. frequency, consistency of terminology) is there that Trump dog whistled during the campaign, and to what extent are we potentially mislabeling Republican talking points as dog whistles? I believe it is necessary to systematically examine the words Donald Trump used during the rallies so often shown on television throughout the 2016 presidential campaign. Only then will claims about Trump’s race baiting in 2016 be supported by comprehensive, empirical evidence.
This paper is a qualitative analysis of the content of Trump’s speeches. The paper is structured in the following manner. First, I present the rationale for the qualitative approach, which is rooted in the integrative model of content analysis (Neuendorf, 2002). That model requires several levels of analysis to be integrated with one another to present a full picture of the content, i.e. Trump’s message. In this case, the model requires analysis of Trump’s words, how those words align with Trump’s persona, and how his words and persona align with the type of voter Trump was trying to reach.
I argue that Trump used racial appeals to connect to his audience and motivate voter support. In support of this argument, I present a brief history of racial appeals and how they may be relevant to the current analysis. After the historical context is established, I address the specific racial terminology, classified as “dog whistles”, likely to be part of Trump’s message. Then I turn to examine the concurrent context: how dog whistling aligns with Trump’s persona and the characteristics of his supporters. Once the logical connections in these areas are established, I move on to the systematic content analysis of Trump’s campaign speeches to see whether he consistently pursued a dog whistling strategy. A discussion will follow that analysis.
2. Integrative Model of Content Analysis
The following study will be conducted using the guidelines for the integrative model of content analysis, as revised by Kimberly Neuendorf (2002). The integrative model, “calls for the collation of content analysis message-level data with other available empirical information” available on the source of the message, the receiver of the message, or other “contextual states” (Neuendorf, 2002: p. 61). The integrative model provides a guide for conducting holistic content analysis that connects the content to other, relevant data; the model also provides a system for evaluating content analyses. The following literature review will provide context with respect to the source (Donald Trump), the receivers (Trump voters), and the messages themselves, including offering empirical support for the history and operation of dog whistle terminology.
![]()
Figure 1. Application of integrative content analysis to the current study. Two-way arrows denote potential linkages among elements of analysis.
The receivers (Trump voters) of Trump’s messages will be discussed using what Neuendorf called a first-order linkage. In this case, a first-order linkage demonstrates a relationship between characteristics of the typical Trump voter, using information from post-election studies, and the characteristics of the messages being examined, the dog whistles in Trump’s speeches. The two are linked directly via two units of analysis: time, because all data were collected in 2016 and related to the 2016 election, and the concept of modern racism. (I am using modern racism in the general sense to describe post-Civil Rights Era racism). Specifically, the modern racism link is between the inherent racism of dog whistle messages and studies finding some form of racism in Trump voters—thus the appeal of dog whistles to these voters. Such a connection would qualify as a first-order Type B linkage, using Neuendorf’s terms. In addition, the source of the messages, Donald Trump, will be discussed using what Neuendorf referred to as a third-order linkage. This means that information provided in the review is intended to demonstrate a logical relationship between Trump’s characteristics (i.e., persona) and the race baiting inherent in his message.
The first section of the following literature review addresses the history of political elites using racism to their advantage by framing policies in racial terms, sometimes via coded appeals, thus establishing the effectiveness of dog whistling. The second section, on dog whistle terms, presents existing research establishing a set of terms considered dog whistles and, therefore, builds a foundation for the content analysis conducted in this study. Based on the integrative model of content analysis, the third section addresses existing research on relevant messenger and receiver characteristics that demonstrate why racial appeals would have been effective. Finally, the central hypotheses of the paper address the content of Trump’s speeches.
3. Racial Appeals
Brown (2016b: p. 328) explains that racially divisive appeals “encourage the spread of a moral panic” and are designed to motivate voters receptive to these messages. Banks and Bell (2013) showed that an “angry” implicit racial appeal resulted in increased opposition to racial policies for those with high symbolic racism scores. In fact, in a series of studies Banks (2014) showed that making White people angry calls to mind racial attitudes. The implication is that angry, populist appeals made by some politicians can make race more salient to their White audiences and use anger to drive voter behavior on racial issues. These discoveries align with a rich history of psychological research on the influence of emotion on behavior, including decision making (e.g., Banks, 2014; Cyders & Smith, 2008; Forgas, 2008; Greenberg, 2012; Haidt, 2006; Izard, 1977; Lazarus, 1991; Lent, Brown, & Hackett, 1994), and political decision making such as voting (e.g., Brader, 2006; Greene, 2013; Haidt, 2012; Lodge & Taber, 2013; Marcus & MacKuen, 1993; Neuman, Marcus, MacKuen, & Crigler, 2007; Pérez, 2016; Redlawsk, Civettini, & Lau, 2007).
To indirectly arouse the influential emotions concurrent with group-centrism —while deflecting accusations of racism—some political elites have used coded terminology in racial appeals. As Albertson (2006: p. 6) notes, “deniability is crucial” in racial appeals because of established norms of equality. Research has shown the strategic necessity of coded language (Bowler, Nicholson, & Segura, 2006; Domke, 2001; White, 2007) because the White participants only showed racialization in their opinions when race messages were coded, not when they were explicit. Or, as Republican campaign consultant Lee Atwater said regarding coded racial appeals,“All these things you’re talking about are totally economic things and a byproduct of them is, Blacks get hurt worse than Whites…‘We want to cut this’, is…a hell of a lot more abstract than ‘Nigger, nigger’”.
Coded terms and phrases used to evoke group-centric reactions are often called “dog whistles”. For the purposes of this paper, dog whistles are language used to obscure race-baiting, which primarily gives the speaker or hearer the ability to deny racist intent. As examined by Ian Hanley López (2014), dog whistles are terms that refer to a minority group, which in turn provokes and takes advantage of whatever emotional response the listener has toward that group, without mentioning them. The contentions of Hanley López (2014) are supported with evidence, such as former Chair of the Republican National Committee Michael Steele’s 2010 apology for the party’s courting of racial resentment. More broadly, the evidence in the Hanley López (2014: p. 5) book demonstrates a consistent pattern of particular language, examined in the following section, used by politicians to take advantage of racial resentment while allowing their supporters a “thin patina…to obscure from them the racial nature of their attitudes”.
Indeed it is this underlying set of attitudes—whether they can be called “racist”, “racial resentment”, or otherwise—that is more likely to result in the atypical behavior seen at Trump rallies than principled conservatism (see Sidanius, Pratto, & Bobo, 1996). The use of transference and countertransference in psychotherapy is founded on the idea that strong, emotional reactions require therapeutic investigation because the initial reason given by the person having the reaction is not often reliable (Chance & Glickauf-Hughes, 1995; Ellis, 1962; Gehlert, Pinke, & Segal, 2014) and may be substituting for a less socially-acceptable explanation (Freud, 1961; Rösch, Stanton, & Schultheiss, 2013; Yalom, 2005). Several studies cited by Lawrence Bobo (2017: p. 95) outline the negative emotional reactions White Americans have exhibited when their attention has been drawn to the demographic trend toward a “majority-minority population”. Based on this research, the courting of racial animus in campaign speeches would have activated White solidarity, causing the strong, emotional reactions at Trump rallies highlighted in the introductory paragraphs.
4. Dog Whistle Terms
The terms discussed in this section are considered dog whistles, according to criteria developed by Ian Hanley López (2014). A dog whistle is a term that works on two levels: the surface, policy level and the deeper, racial (and therefore emotional) level. The term “dog whistle” in this paper is used interchangeably with coded racial appeals; the dog whistles discussed herein may be able to be decoded, and some have been. What is more important is that the phrasing gives the hearer the gut reaction of a racial appeal and grants a sympathetic audience plausible deniability as to the term’s racialized nature.
Hanley López rooted his criteria for dog whistle terms in the history of those terms, the reactions caused by the terms, and the meaning of the terms, especially in post-Civil Rights Era usage dating back to the start of the “Southern Strategy”. The Southern Strategy, which Hanley López (2014: p. 48) calls “strategic racism”, was Richard Nixon’s attempt to use White solidarity around issues tinged with race such as school integration to make Republican inroads in what was once the Democratic South. While Alabama Governor George Wallace race baited, during the same 1968 campaign Nixon positioned himself as a more acceptable candidate by coding his appeals to White Southerners in language like “law and order”.
4.1. Law and Order
Richard Nixon’s appeals for “law and order” stand as one of the most well-known examples of a race-baiting dog whistle. Marable (2007: p. 124) explained the use of the “law and order” rhetoric as racial code to “instill reactionary anxieties among Whites”. Modern applications of the supposedly colorblind term “law and order” to Latinx people, in lieu of racist language and anti-Latinx stereotypes, run parallel to the “shift from explicit racism to institutional racism in the criminal justice system” (Lasch, 2016: p. 163). Ioanide’s (2015) analysis of coded racial appeals includes the prison expansion since the 1980s, which was supported by colorblind “law and order” appeals yet, “overwhelmingly associated the threats of crime with ‘hyperviolent’ Black and Latino men” (4).
4.2. School Choice
Discussions of “school choice” or school vouchers are rooted in Nixon’s Southern Strategy approach to school integration. Former Nixon aide John Ehrlichman (1982: p. 223) wrote that Nixon’s statements on schools contained a, “subliminal appeal to the anti-Black voter”. “School choice” has replaced the “busing” (sometimes called “forced busing”) of the Nixon era as an education issue, though the intent of the dog whistle is the same: maligning the policies and outcomes of public schools, which are required to accept all local students, regardless of race. In Abernathy’s (2005: pp. 37-38) analysis of the school choice debate, he cites evidence from the voucher system in Milwaukee in arguing that “private sector exit” has “destructive” effects for poor, minority communities.
Anti-busing or pro-voucher sentiment often serves as a populist, tax-driven cover story for racial anxiety over school desegregation (Hanley López, 2014: p. 213). It is worth noting that there is already de facto, though not de jure, segregation occurring in schools due to the persistent remnants of redlining, “White flight”, income-related migration patterns, and other housing policies (Rothstein, 2017). Roda and Wells (2013) summarized this connection in their study of affluent parents deciding on schools for their children. They concluded that, when school choice policies do not take into account and promote racial integration, they tend to reverse it (p. 262). Though school choice can be, and has been, used to address deficits in education for Black students (Fusarelli, 2003; Morken & Formicola, 1999), political appeals for school choice have been rooted, historically, in the fears and concerns of White parents (Abernathy, 2005).
4.3. Welfare
“Welfare” (Slocum, 2001), and “welfare dependency” (Brown, 2016a) have been so consistently linked to Black Americans and, therefore White voters’ feelings about them, that politicians have been able to use the terms—without explicitly mentioning Black people—to arouse an angry reaction to stereotypical images of lazy minorities supported by the voter’s tax dollars (Wells & Roda, 2016). Perhaps the best-known example is Ronald Reagan’s 1976 campaign using the term “welfare queen” while citing the welfare abuses of Linda Taylor, nee Martha Miller, who was identified as Black. Reagan’s usage maligned the potential, though uncommon, excesses of the welfare program and, by extension, Black recipients (Levin, 2013; Smith, 1987).
4.4. Voter Identification
Calls for stricter voter identification (ID) laws also function as a dog whistle. Voter ID proponents use supposed “common sense” to cloak legislation that would disproportionately affect groups that have trended toward voting Democrat like young people, Black people, and Latinx people (Bowler & Segura, 2012). Indeed, these groups were shown to be the victims of recent voter ID laws, according to reports from the 2016 election (Associated Press, 2017; de Vogue, 2016; Wines, 2016). Banks and Hicks (2016) found that inducing fear causes Whites high in implicit, but not explicit, racism to be more supportive of voter ID. It is worth noting the 2020 Trump campaign’s insistence that there is widespread voting fraud culminated with attempts to disqualify votes from predominantly Black counties in Michigan, Pennsylvania, and Georgia.
4.5. Illegal
Ian Hanley López (quoted in Desmond-Harris, 2014) notes that terminology such as “illegal alien”, “illegals”, and “criminal illegal alien” is an intentional dehumanization and criminalization of undocumented immigrants, and by proxy all Latinx people. It should be noted that “illegal” immigration is a blunter instrument than the preceding three dog whistles. This is due to racialized language evolving over time, as noted earlier, but this is an intentional evolution. The value of this dog whistle is that it provokes a response, not to law-breaking as the term itself denotes, but to all Latinx-Americans, legal immigrants or not. Even though the speaker does not mention where the immigrants are coming from, illegal has been paired enough with images of crime and Mexican-Americans (Chavez, 2001) that simply using the term “illegal” in the context of immigration is enough to invoke images of, and reactions to, Latinx people (Brown, 2016a).
Even if Americans decode this dog whistle as targeting Latinx-Americans it remains potent due to dog-whistler’s cover story that their real problem is with law breaking. This is despite Republican opposition to citizenship, which they term “amnesty”, for law-abiding and working undocumented immigrants and Republican-led Congressional opposition to immigration reform during the 2010s. The “illegal” usage against Latinx people was limited before the Hart-Celler Act of 1965 that restricted immigration from the Western hemisphere, but continues a lengthy history of often-coded immigrant scapegoating in the US (Ngai, 2013; Preston, 1963).
4.6. Islam
Dog whistles have been used to pit Christians (“us”) against Muslims (“them”), who are often conflated with terrorists in these messages (Brown, 2016b). Dog whistling about “the other” is meant to provoke the same emotional reaction found in prior work on the interaction of voting behavior and negative evaluations of outgroups (Cornielle, Yzerbyt, Rogier, & Buidin, 2001), specifically Muslims post-9/11 (Nisbet, Ostman, & Shanahan, 2008; Nisbet & Shanahan, 2004). Even before 9/11, Edward Said’s (1997: p. 48) analysis of Western media representations of Islam noted the logical conclusion of almost uniformly negative portrayals and othering of Islam is the necessity of a “confrontational response” towards it. Because of this linkage, the use of the word “Muslim” or “Islam” by political elites functions as a dog whistle, drawing attention to the differences between White (Christian) Americans and “an Arab Muslim enemy” (Hanley López, 2014: p. 120).
Much like the “illegal” dog whistle, “Islam” is a blunter instrument in that it appears easier to decode. The negative connotations with Islam, including radical, Anti-American terrorism, have been built up for decades (Said, 1997). These connotations resulted in a rapid rise of anti-Arab violence following the September 11, 2001 attacks. Notably, as it pertains to the analysis of Trump’s language, these hate crimes against Arab-Americans sharply rose during the 2016 campaign and rose again in the year following Trump’s inauguration (Council on American-Islamic Relations, 2018).
As discussed above, language evolves over time: some dog whistles have become less relevant as the public has decoded them (i.e., “busing”), others have maintained their relevance despite growing understanding of their meaning (i.e., “illegal”), and still others show potential to grow as reflective of proposed legislation (i.e., “voter identification”). Other dog whistles have fallen out of favor, such as “states’ rights”, as the words themselves have been less decoded than linked to an old-fashioned point of view, including sympathy for anti-integrationism and the slavery-defending Confederacy (Kendi, 2016). In the case of “welfare”, the public has become sensitive enough to this term’s use that it is not as acceptable to criticize welfare assistance in presidential campaign speeches, though other political elites continue to malign the social safety net as creating people dependent on government “handouts” (Hanley López, 2014; Thompson, 2018).
Dog whistles are relevant for the same reason that race baiting overall has been discouraged for the better part of a century: beyond the moral concern with individual racism, the reactions these words provoke can result in actual violence. Overall, the motivational influence of emotion-provoking racial appeals, when obfuscated by coded terminology, make dog whistles powerful weapons for political elites. I argue that the most important reason to analyze Trump’s use of dog whistles is to help explain the unusually emotional and sometimes violent reactions we witnessed during the Trump campaign.
In order to provide context to the analysis of dog whistles in Trump’s speeches, the following section is an examination of Trump’s political persona and the characteristics of his supporters, specifically in terms of whether it is likely their sympathies are aligned with the racism implied by dog whistles. Trump’s persona and audience are discussed here citing research from prior studies. The following section addresses two pieces of the integrative framework (Figure 1): characteristics of messenger and receiver. Once their alignment is established—that is, once it is demonstrated that characteristics of Trump’s public persona are suited to make racial appeals to an audience susceptible to them—I will move on to determining whether the message fits the racial appeal framework.
5. Donald Trump and His Audience
In a word analysis of the 2016 presidential debates, Krzywinski (2016) concluded that Trump stood out for his repetition and low number of independent concepts compared to the other three presidential candidates. A separate analysis of the “word data” from the primary debates by Zhong (2016: p. 8) found Trump to speak at a 4th grade level, which contributed to Trump’s effectiveness as a presidential candidate and his appeal to a, “low information audience”. In a study of Trump’s presentation style, a group of ethnographic researchers (Hall, Goldstein, & Ingram, 2016: p. 72) concluded that Trump gained politically by vicariously empowering a “rural White underclass” through anti-establishment behavior often deemed un-presidential. Recall that anger (Banks, 2014) is the emotion that primes racial attitudes in voters.
Research shows that this presentation aligned with his message and the voters he was trying to reach. Oliver and Rahn (2016: p. 199) found that Trump primary voters scored highest in, “mistrust of expertise, national affiliation, and nativism”. Research (Tesler, 2016a, Tesler, 2016b) has shown that racial attitudes mattered more in 2016 than in the prior two presidential elections—when a Black man was on the ballot—and that racial resentment and ethnocentrism were more closely linked to support of Trump than support for 2012 Republican nominee Mitt Romney.
Contrary to explanations linking economic hardship to support for Trump, Gallup researchers (Rothwell & Diego-Rosell, 2016: p. 14) found that one of the strongest predictors of Trump support was “racial and ethnic isolation” of White voters. They note in their conclusion that, “cultural views and social identity” are a more powerful influencer of political preferences than economic and most demographic factors (Rothwell & Diego-Rosell, 2016: p. 19). In a study using the Color Blind Racial Attitudes Scale (CoBRAS) as a measure of racism, Schaffner and colleagues (2017) found that “the effect of economic dissatisfaction is dwarfed” by the relationship between racism and voting for Trump (p. 16). Wood’s (2017) research using the American National Election Study found that moving from the 50th to the 75th percentile on the symbolic racism scale, “made someone 20 percent more likely to vote for Trump”.
In a study using Social Dominance Orientation (SDO) as a proxy for group threat, Mutz (2018) found status anxiety issues, a.k.a. status threat, to be determinant of voting behavior in the 2016 presidential election. Mutz (2018) concluded, “Those who felt that the hierarchy was being upended—with whites discriminated against more than Blacks, Christians discriminated against more than Muslims, and men discriminated against more than women—were most likely to support Trump” (p. 4338). Researchers (Fowler, Medenica, & Cohen, 2017) constructed a survey of White vulnerability (described above as status threat) in millennial voters. They found that fear of losing ground to non-White groups, which is driven “primarily” by racial resentment, has been identified as the largest predictor of voting for Trump (Fowler et al., 2017).
Dog whistles are an effective way to continue courting racist supporters while both appealing to the unspoken racial resentment of other voters and creating enough doubt about the candidate’s intended meaning to satisfy and provide plausible deniability to racially centrist Republicans and undecided voters. If Trump was race baiting in his speeches, there was a sympathetic audience to receive those messages; his campaign could have learned as much from reactions provoked by Trump’s Twitter account and the unwavering support he received from a core group of Republican voters following his candidacy announcement, in which he openly criticized Mexican immigrants. But there are reasons to believe that dog whistling would have been effective on more than Trump’s most loyal supporters. Racial appeals attract low-education (Silver, 2016) and low-information voters (Oliver & Rahn, 2016; Tesler, 2015) and dog whistles simplify complex issues (Hanley López, 2014; Krzywinski, 2016) by provoking an emotional response such as racial resentment (Potts, 2016; Silver, 2016; Wood, 2017) or nativist anger and anxiety (Oliver & Rahn, 2016). Even though most of the preceding research on Trump voters was published after the election, the reactions Trump received during the campaign are an indicator that Trump was communicating something meaningful, and often emotion-provoking, to voters.
The preceding review established the alignment of certain terminology as effective race baiting, Trump’s political persona, and the potential receptiveness among a portion of the electorate to race baiting. According to the Integrative Model of content analysis, these areas should be aligned. The history of racial appeals, including dog whistling, shows that they appeal to voters on an emotional, rather than a rational, level. The evidence presented in the preceding section established that Trump’s persona as an anti-establishment candidate was suited to taking advantage of the anger of potential voters. The research presented also indicates that voters feeling isolated, threatened, and angry were likely to support him. The “straight-talking” persona of the messenger and the vulnerability of the audience to emotional appeals have been established; what is left to be determined is whether Trump consistently delivered the type of message designed to take advantage of those factors.
In the following content analysis, the words of Donald Trump are examined to determine whether he consistently used dog whistles during the 2016 election. It is predicted in Hypothesis 1 that Trump consistently used dog whistles in his campaign. The standard of consistent usage is to ensure accurate representation of Trump’s messaging. To be considered consistent usage, a term must meet two criteria: appearance in at least 50% of the candidate’s speeches and overall usage averaging at least once per speech. The criteria were chosen to balance one another in terms of assessing usage of a term within and across speeches. Therefore, a dog whistle was consistently used if the message receiver was sufficiently likely to hear the term in any given speech1.
The emotionally charged and divisive 2016 election cycle indicated that there was something atypical occurring, yet it is worth considering whether these dog whistles were one of these atypical elements or merely ordinary policy-speak from a Republican presidential nominee. To answer this question, I subject the speeches of the preceding two Republican presidential nominees, Mitt Romney and John McCain, to the same analysis. The decision to include Romney and McCain is in line with previous literature assessing Trump’s appeal (Schaffner et al., 2017) and intended to address not only the possibility that the dog whistles in the study are merely US Republican candidate terminology but also how Trump’s use of the terms aligns with comparable, prior usage. If the terms are Republican policy-speak, Trump’s use of them should be in line with that of previous Republican presidential nominees. Hypothesis 2 is based on prior research positing that these terms are more than shorthand for policy: they are laden with additional, racialized connotations. Differences in dog whistle usage between Trump and the other candidates would place into context the uniqueness of the behavior seen at Trump rallies as well as the findings establishing the racist tendencies and status fears of some Trump supporters.
Mitt Romney is included not only because he was the most recent Republican presidential nominee but also because his candidacy was the first to follow the birth of the Republican-aligned Tea Party movement, which has been criticized for the racialized speeches given at some of its rallies (Bonilla-Silva, 2018; Hanley López, 2014). John McCain ran against the first Black, major party, presidential nominee in US history. McCain’s language in that contest holds value especially in comparison to Trump’s, considering Trump’s contribution to the racist Birther movement. Finally, given the prior discussion of the Southern Strategy, Trump’s speeches are also compared to those given by Richard Nixon as 1968 Republican presidential nominee, solely because of Nixon’s role as a pioneer of the modern dog whistling strategy and his usage of “law and order” to appeal to the “silent majority” who were disturbed by the pace of change, particularly the prevalence of antiracist protests and other demonstrations in the mid-1960s (Ehrlichman, 1982; Marable, 2007).
It is predicted in Hypothesis 2 that Donald Trump’s use of each dog whistle significantly and meaningfully exceeded Romney’s, McCain’s, and Nixon’s in t-test comparisons, with a minimum significance level of p < 0.05. Only in the case of the term “law and order”, is it expected that Nixon’s use exceeded Trump’s2. This prediction is based on the idea that something different was occurring in 2016, and Trump played a role in that difference. This is because supporters often take their cues from political elites (Mendelberg, 2008; Nelson & Kinder, 1996). It is predicted that Trump’s dog whistling is significantly greater due to the difference between Trump’s and the other candidates’ rallies; the difference was most noticeable in the behavior reported during Trump rallies, which are indicative of an emotional or automatic reaction to Trump’s words (DiAngelo, 2018; Freud, 1961; Rösch, Stanton, & Schultheiss, 2013; Yalom, 2005).
A delicate operationalization issue presents itself when dealing with coded language: intent. Donald Trump’s intent, or that of most of the other politicians included in the study, in dog whistling is not specifically examined. Intent cannot be concluded absent a statement such as those that, eventually, trickled out from operatives like Reagan’s Lee Atwater or Nixon’s Gordon Brownell (Lassiter, 2006). This study is interested in the historical thread that can be drawn using coded messages for which we know the meaning. Did Trump consistently use dog whistles, i.e. do his critics have solid ground to stand on when accusing his campaign of being racially divisive? In Hypothesis 1, I predict that Trump did use the following dog whistles consistently, as measured by appearance in at least half of his speeches and averaging at least once per speech, throughout his campaign as the Republican presidential nominee: law and order, illegal, Islam, school choice, voter identification, and welfare. And is Trump’s dog whistle usage particularly heavy-handed compared to that of his political predecessors? In Hypothesis 2, I predict that Trump’s average dog whistle rate is significantly greater than the other candidates’ in terms of each dog whistle (except law and order) and overall totals, as measured by t-test comparisons.
6. Method
The speeches Donald Trump made as Republican presidential nominee were downloaded from the American Presidency Project (APP) website. The APP website is non-partisan and hosted by University of California at Santa Barbara. Trump’s speeches as shown on the APP website were cross-checked with official speech scripts available at the official Trump campaign website (donaldjtrump.com). The APP website was also the resource for speeches by the comparison candidates; all the following candidate quotes are from said resource. The speeches were subjected to content analysis using the procedure detailed in the subsequent section.
6.1. Materials
The set of speeches analyzed were subject to an element of sampling convenience in that, in order to be readily accessible to the coders, the speeches needed to be available. The speeches selected were those made at official campaign stops as listed on the Trump campaign website. Press releases were not included, nor were interviews. This decision was made not only to reduce variance in the intent and format of the speeches but also to ensure that content and language in the speech was delivered by Trump himself directly to voters. The basis for this decision is reflected in the research on elite frames and racism, summarized in a preceding section. The political statements examined in these studies were intended by the candidate to create an emotional reaction in potential voters. Therefore, press releases, which are usually straightforward policy or logistical announcements, and interviews, which reflect less of a strategy than prepared campaign speeches because they are extemporaneous, are less likely to indicate a consistent pattern of language usage. A consistent pattern of usage is important because it reflects meaningful messaging from the campaign.
The speeches of the two most recent Republican nominees were available on the APP website, as were the speeches Richard Nixon gave as Republican nominee in 1968. For the reasons outlined above, all the campaign speeches available for these candidates were included in the analysis. I considered that a comparison of Trump’s 2016 Republican primary opponents might be valuable. A sample of the speech transcripts of Trump’s final two Republican opponents, candidates Senator Ted Cruz and Governor John Kasich, was obtained for preliminary analysis. After reviewing the speeches by Sen. Cruz and Gov. Kasich, it was determined that including them posed some methodological issues. Even in the preliminary review, it was clear to coders that the language used by Cruz and Kasich did not include any uses of the dog whistle terms as understood in common parlance and eventually codified during the study’s procedure. The recent elongation of the primary period and the potential for introducing more variation related to primary competition also indicated that including these speeches could create additional, internal consistency issues.
Based on research on the necessity of coded language in racial appeals, a coded racial appeal is more important to deploy in the general election than in the primaries: the appeal to White solidarity (rather than ideologically divisive policy issues) could be an effective strategy to attract undecided and non-Republican voters. Therefore, speeches given during the primaries were removed so that the comparison between presidential candidates would be as even as could be reasonably expected. The final set of speeches included for each Republican presidential candidate begins with the acceptance speech given at the Republican National Convention that election year and concludes with the final speech given before the date of the election. Every available speech given by the candidate during said period was subject to analysis.
6.2. Procedure
Based on prior research, a set of terms, hereafter referred to as dog whistles, were chosen to be included in the analysis. The process of creating a list and, thus, determining relevant constructs before analysis is in line with a priori (Neuendorf, 2002) or deductive (Elo & Kyngas, 2007) content analytic method, in which a predetermined structure based on prior research is used to develop themes and, eventually, codes to guide the textual analysis. While this is a qualitative study, one of its goals was to be able to quantify dog whistles to indicate consistency across speeches and determine which areas were mentioned most. Though the speeches analyzed come from four different years, for the sake of consistency the terms were operationalized the same way across speeches. In order to account for time differences in terminology use, thereby addressing potential inequalities in dog whistle totals between the 1968 set and the modern speeches, I included several 1960s-era dog whistles in the analysis (see Appendix 1).
To be included in the study as a dog whistle, a term needed to have a documented history of usage as code for activating White solidarity against a minority group. A history for these terms is needed because such terminology, once developed, is only useful because it is linked over time to a group to the point that the group need not be mentioned (Hanley López, 2014). Though dog whistles for non-racial issues such as religion (e.g., Albertson, 2006) exist, racial dog whistles are pursued here due to the nature of the Trump campaign events discussed in the introduction to this paper. The dog whistle terms chosen for this study were rooted in the Hanley López (2014) book on race-baiting, political dog whistles. As noted in the Dog Whistle Terms section, though, many of the terms cited by Hanley López appear in numerous other works and have come to be accepted as coded terminology among historians and political scientists. The dog whistles, italicized in this section, were law and order,illegal,Islam,school choice,welfare,and voter ID.
In the initial listing of dog whistles, three additional terms were considered: school busing,food stamps,andstates’rights.Though these terms are not often used today as racial code, they were more prevalent during the late 1960s and into the 1970s. I included these terms to allow for a fairer comparison specifically between Trump’s and Nixon’s dog whistle totals. However, Nixon did not use these terms in his 1968 campaign speeches and thus the terms were dropped from the list. Though the attempt at a more even comparison between Trump and Nixon failed in that respect, the information learned through content analysis of Nixon’s speeches was helpful toward the second hypothesis, as discussed in the results. Additional details on the coding process are included in Appendix 1.
A codebook for the dog whistles (see Appendix 2) was created using definitions based on research cited in the “Dog Whistle Terms” section. Based on recommendations drawn from similar, qualitative studies of content (e.g. Braker-Walters, 2014; Freeman, 2017) the analysis utilized “hand coding.” Hand coding requires human coders to read and assess the text during the coding process. Hand coding also allows coders to consider the importance of context, as noted above in discussion of the code definitions, and meaning before completing analysis (Cohen, Manion, & Morrison, 2018).
The analysis was conducted by the author and four trained analysts. Once the codebook was created, an analyst used it to guide scanning of each assigned speech, mark dog whistles found in the text (e.g., LO for law and order), review for accuracy based on the codebook, and then conduct a dog whistle count for that speech. Each reviewer examined a set of speeches, which included a subset of speeches examined by other coders as a check on the accuracy of the codebook. As assessed by Krippendorff’s alpha3, interrater reliability on coding was good (α = 0.824). Final coding resulted in 94.77% interrater agreement.
7. Results
The results below are presented in terms of their relevance to the hypotheses generated prior to data analysis. Therefore, only major themes, i.e., frequency of dog whistle codes within candidate data sets, are reported. Topics consistently mentioned by the candidates that did not meet the criteria for coding or did not align with the racism theme underlying the dog whistle analysis are not presented here.
7.1. Hypothesis One
It was expected that Donald Trump used a variety of dog whistles during the 2016 presidential campaign. Trump did indeed use all the following terms in his campaign speeches: law and order,illegal,Islam,school choice,welfare,voter ID. Descriptive data for each of the dog whistles Trump used during the campaign are presented in Table 1. Trump did not consistently use the terms voter ID orwelfare.
Trump used the immigration dog whistles the most, which is not a surprise given that immigration was the dog whistle topic emphasized in his candidacy announcement. He mentioned it in all but 14 of the 62 speeches examined in this study, finding time even in a policy speech about the military on November 3 to mention it six times. Often, Trump used a fear-based approach to speak about immigration. In his August 23 speech in Austin, he said, “Today I met with the moms of American children killed by illegal immigrants as a result of the policies Hillary Clinton supports”. Trump repeatedly, graphically used the example of Kathryn Steinle’s murder to stoke fear of immigrants, such as in his November 2 speech in Miami: “(Hillary) strongly supports sanctuary cities like in San Francisco where…Kate Steinle was killed by a five-time deported illegal immigrant [audience boos]”. Trump also spoke negatively about Islam in 39 of his speeches and mentioned it third most of the dog whistles examined. When
![]()
Table 1. Dog whistles in Donald Trump’s 2016 Campaign Speeches.
Note: Numbers in bold indicate consistent usage. *The highest count occurred in two speeches.
Trump spoke about Islam, it was almost always in the context of terrorism. On November 2 in Miami, he said, “When I’m president, we will suspend the Syrian refugee program (applause) and we will keep Radical Islamic Terrorists the hell out of your community”.
Trump used law and order most in his address to the Republican Convention (n = 9) and his August 31 speech on immigration (n = 11). In that acceptance address, he stated, “I have a message to every last person threatening the peace on our streets and the safety of our police: when I take the oath of office next year, I will restore law and order to our country”. Based on the results shown in Table 1, Donald Trump consistently race baited during the 2016 campaign.
7.2. Hypothesis Two
To determine whether Trump’s use of these terms is out of line with prior Republican campaigns, Trump’s rate and consistency of dog whistle usage were compared to that of Mitt Romney, John McCain and Richard Nixon. The dog whistle usage rates of the four candidates, including the percentage of speeches each dog whistle appeared in, means, and standard deviations, are displayed in Table 2.
Trump used the most dog whistles and used them most consistently, i.e. they appeared in the highest percentage of his speeches, in five of the six cases in which there is a meaningful difference between candidates.
The dog whistle usage of all three candidates was compared using an ANOVA test. The test showed significant differences among the groups (p < 0.01) on each of the dog whistles except for voter ID, which no other candidate used. Games-Howell4 post-hoc tests were performed due to unequal variance and unequal sample sizes. Figure 2 is a bar graph of average dog whistle usage per candidate, along with whether the usage rate was significantly different from Trump’s
![]()
Table 2. Dog whistle usage consistency comparison.
Note: Highest appearance percentages for each dog whistle noted in bold type. Speeches analyzed for Trump: 62, Romney: 52, McCain: 35, Nixon: 21.
![]()
Figure 2. Candidate comparison on average usage per speech for five dog whistles. *difference between Trump and other candidate significant at p < 0.05; **difference between Trump and other candidate significant at p < 0.001.
usage, based on the post-hoc t tests comparing candidate usage rates.
Trump’s usage of dog whistles was found to be significantly greater (p < 0.001) in almost every case. The comparisons in Table 2 show a meaningful difference as well, as in multiple cases Trump used a dog whistle that the comparison candidate used inconsistently or not at all. Based on the criteria for consistent usage, no candidate except Trump consistently used a dog whistle, though law and order was used by Nixon an average of about once per speech. Hypothesis Two was supported with one exception: Nixon’s use of law and order did not exceed Trump’s usage in a meaningful or statistically significant way.
8. Discussion
The results for both hypotheses support the claim that Donald Trump consistently race baited during the 2016 US presidential campaign and did so at a rate significantly and meaningfully higher than comparable Republican presidential candidates. These results fit into the integrative framework for linking messenger, receiver, and content, as presented in the literature review. The racist stereotypes inherent in dog whistles, including the fear they inspire, align with studies demonstrating how racial resentment and status threat motivated White Americans to vote for Trump. Donald Trump’s swaggering persona, blunt language, and emotional appeal align with studies showing the susceptibility of Trump supporters to such an approach. The anger inherent in much of his message (Smith & Hanley, 2018; Stevenson, 2016), and echoed by rowdy rally crowds (Mathis-Lilley, 2016), has the power to enhance the influence of ideas about race on political behavior.
As displayed in Figure 1, the integrative model I use to evaluate the evidence for whether Trump dog-whistled is built on the alignment of three elements. Trump’s persona, based on evidence presented in the paper, is anti-establishment (Hall et al., 2016) and appealing to low-information (Zhong, 2016) and low-education audiences. Trump voters were found likely to be distrustful of expertise (Oliver & Rahn, 2016), racially isolated (Rothwell & Diego-Rosell, 2016), holding negative racial animus (Schaffner et al., 2017; Wood, 2017), and threatened by demographic change (Fowler et al., 2017; Mutz, 2018). Racial appeals have been found to work well with low-information (Silver, 2016), angry (Banks, 2014), and anxious (Oliver & Rahn, 2016) audiences. The final piece to this integrative framework is the content of the messages themselves and, regarding that, the results are clear.
Donald Trump consistently used race-baiting dog whistles during the 2016 Presidential campaign. On average, speeches contained between 11 and 12 dog whistles. Only one of Trump’s speeches, his October 3 speech on cybersecurity, contained zero dog whistles. Trump spoke most about immigration: he maligned undocumented immigrants at an average rate of over five times per speech. Trump warned a crowd in Colorado Springs that, “criminal aliens” are coming for “your job”. Lest anyone believe his words on immigration were empty rhetoric, statistics show that, under Trump, arrests of immigrants increased significantly—41%—while arrests of non-criminal immigrants have risen 171% (Kopan, 2018). The Trump administration’s treatment of immigrants and asylum-seekers, most notably the treatment of immigrant children including separating them from their parents and later attempting to deport them sans parents, matches the dehumanization described in the prior research on immigrant rhetoric (Brown, 2016a; Chavez, 2001; Chavez, Whiteford, & Hoewe, 2010; Fernández & Pedroza, 1982).
The term “law and order” had been exposed as race baiting even before Lee Atwater’s smoking gun on the Southern Strategy, yet Trump used it 76 times in 62 campaign speeches. His usage rate and consistency were not significantly different than Richard Nixon’s in 1968. One potential excuse for talking about law and order is that there was unrest in the country (much like in 1968), particularly the July 7, 2016 shooting of Dallas police officers by Micah Johnson less than two years following the Michael Brown shooting and subsequent Ferguson unrest that contributed greatly to the growth of the Black Lives Matter movement (BBC News, 2016). But there were many lawful, peaceful demonstrations against police brutality (Quintana, 2016; Yan & Park, 2016) and declarations of sanctuary cities (Luhby, 2016); warnings and complaints about these often served as context to law and order and illegal in Trump’s speeches. Some could argue, as Richard Nixon once did, that more law and order would help all citizens. I would argue, considering the response Trump provoked from his supporters and research showing how dog whistle approaches work (Banks & Bell, 2013; Brown, 2016b; Domke, 2001), that it is not accidental that the bulk of the people who would be presumably locked up or deported once “law and order” are restored—the people invoked in Trump’s warnings—are not White: supporters and participants in a Black-led protest movement and Mexican and Central American immigrants.
Trump spoke frequently about school choice—fifty of his sixty-two speeches use the words. Support for vouchers has been a Republican policy point for at least twenty years; however, if the school choice dog whistle is merely Republican terminology, it does not explain why his usage difference with each of the two preceding Republican presidential nominees is so large, despite it being the dog whistle Romney used most overall. John McCain barely mentioned school choice: he used the term four times, and usage stops after September 13, 2008. The differences found here between Trump’s words and those of his predecessors is supported by recent research on the type of voter each candidate attracted: Schaffner and colleagues (2018) found that there was no relationship between racism and voting for the previous two Republican presidential nominees.
Trump’s comments on Islam were universally negative. Though the coding instructions mandated that only negative uses of Islam or Muslim would be counted, during the process it became apparent that other usages were not in his speeches. Trump’s crowds were recorded as booing the mention of Syrian refugees on several occasions. Trump followed “Islam” or “Islamic” with “terror” over 90% of the time. On October 22 he warned, “Radical, Islamic terror is right around the corner” and proposed, “extreme vetting” of refugees. In context, Trump’s message uses an us-versus-them dynamic with religious and ethnic minorities portrayed as “them” and potentially sets those minorities up as targets for aggression. It is not surprising that these statements coincided with clashes between protesters, sometimes with Trump encouraging his crowd (Finnegan & Bierman, 2016; White, 2016), and an increase in anti-Muslim groups (Struyk, 2017). Less than a week after the 9/11 attacks, George W. Bush famously said, “The face of terror is not the true faith of Islam…Islam is peace”. When Trump called Mexican immigrants rapists and drug-runners during his candidacy announcement, he added, “Some, I assume, are good people”. Donald Trump did not make that barest of concessions when speaking about Muslims.
9. Conclusion
Words matter, especially when they come from a candidate for the highest function in a state. Donald Trump’s words mattered not only because of the position he was in but also because of the strong reactions that came along with them—reactions, according to the research (Banks & Hicks, 2016; Brown, 2016a; Hurwitz & Peffley, 2005; Mendelberg, 2008; Slocum, 2001; Wells & Roda, 2016), he should have expected. Trump’s specific, race-baiting words also mattered because they were tied to campaign promises like the US-Mexico wall (illegal; McCaskill, 2016), “extreme vetting” of refugees and a travel ban from Muslim-majority nations (Islam; Diamond, 2015), and cracking down on antiracist protests (law and order; Levitz, 2016)—while boasting an endorsement from racial profiling ex-Sheriff Joe Arpaio (Siegler, 2016). Though the idiosyncrasy of the Trump campaign is well known, there is a difference between an unscripted remark or tweet and words repeated in speech after speech from July through November 2016. The results show how consistent he was in using the four dog whistles listed above.
Donald Trump mixed the dog whistle approach with more open race-baiting rhetoric as president. Trump famously created an equivalency between the actions of “both sides”, White supremacists and antiracist protesters, following the violent clashes in Charlottesville, Virginia in August 2017. Post-Charlottesville, Trump held a rally in Arizona and complained about groups requesting the removal of Confederate monuments: “They’re trying to take away our culture, they’re trying to take away our history”, to which CNN’s Chris Cilizza (2017) replied, “[dog whistle]”.
The next month, Trump called for the firing of National Football League players demonstrating during the pre-game national anthem. He continued to complain about the anthem protests, though most of them had stopped, into the 2018 NFL season. The NFL is almost 70% Black men (Gertz, 2017) and, until Trump’s comments, only one White NFL player had demonstrated during the anthem. Within the NFL discussion, Trump supporters can argue that the statistic is cherry-picking. In the larger context of Trump’s words, the statistic fits a pattern of Trump agitating White people against non-White people.
In January 2018, in discussion with lawmakers working on an immigration deal, Trump referred to El Salvador, Haiti, and a group of African nations as “shithole countries” (Dawsey, 2018). During the 2020 presidential campaign, Trump stumbled when asked to disavow the support of White supremacist groups like the Proud Boys (Collins & Zadrozny, 2020; Fabian, 2020). He agitated for support of Kyle Rittenhouse (Brewster, 2020; Wise, 2020), who killed two men at an antiracist protest over the police killing of Jacob Blake. Trump went back to the dog-whistling about law and order in warning “suburban women” that Democratic candidate Joe Biden was going to “destroy suburbia” with an “invasion” of low-income housing (Karni, Haberman, & Ember, 2020; King & Barrón-López, 2020). When Trump was confronted with accusations of race-baiting, he cited statistics on minority populations in the suburbs.
Whether Donald Trump personally holds racist beliefs cannot be concluded here. What can be concluded—what cannot be avoided—based on the words Trump used repeatedly, the history of those specific words in political communication, the response those words provoked among his supporters in real time, and the overwhelming support he received among certain White people, specifically those found to be most likely to hold implicit racist beliefs and support a symbolically racist agenda, is that a cohesive theme of Trump’s campaign speeches is language appealing to the racism of American voters.
As Tali Mendelberg (2008: p. 113) noted in her review of studies on coded racial messages, “The power of elites to promote or deflate racial politics is strong and consistent”. The preceding findings add to research underlining the value and influence of candidate language, especially in the context of the racial and ethnic angst during the 2016 US presidential campaign and since. Finally, the results contribute to the growing literature demonstrating that race played a role in the 2016 election by providing proof that Trump spoke in a way designed to attract the very voters researchers found to be crucial to his victory: those threatened by the rise of minorities and those exhibiting some form of racism.
Whether Trump benefitted from race-baiting, and the results indicate he did, is different from being able to conclude he did so strategically and intentionally, let alone predict during the campaign whether he would win by preying on the implicit racism of the American public. We cannot, based on the results, conclude that racism is solely responsible for his election. However, the results illustrate that political analysists are on solid, empirical ground in claiming that Donald Trump dog-whistled during the campaign. The findings of this study also indicate that racism, especially the opportunistic racism of consistently arousing and benefitting from racial polarization, belongs in honest discussions of why Donald Trump was elected President of the United States.
Appendix 1: Details on Coding Process
There was a pilot coding to test the first version of the codebook using a subset (n = 15) of Trump’s campaign speeches. During the pilot coders noted that guidance would be needed before official coding began; namely, coders needed to know whether the terms must appear as they are listed or if alternate forms would count. For example, some terms, like Islam, were straightforward: coding required use of any form of the word Islam or Muslim (e.g., Islamic,Islamist,Muslims) in a negative connotation for a usage to count. Law and order coding required those three words used in that exact order. Coding of welfare required usage to be in reference to the social program, i.e. not regarding general well-being.
The term illegal, however, involved a more complicated code development. The dehumanization and “othering” of undocumented immigrants, specifically those from Mexico and Central America, as discussed in the literature review, was found to include alternate terms such as “criminal”and“alien”. Therefore, the illegal code was expanded to include forms of criminal and alien. For a usage of illegal to count as a dog whistle the context must be undocumented immigrants—often referred to by Trump as “criminal illegal aliens”. Similarly, voter ID coding included “voter ID” or “voter identification” as well as warnings about “voter fraud” or “voting fraud”. The latter terms were included because the mechanism is the same: the words are referring to an unspoken group of people who supposedly are gaming the voting system, resulting in an appeal for voter ID requirements.
School choice was another case in which the words were coded as a dog whistle even when they were not adjacent, as in this example from Trump’s October 13 speech in Columbus, Ohio (dog whistles in italics): “Under a Trump Administration, disadvantaged children will be able to attend the public, private, charter or magnet school of their choice”. The term also included alternate forms such as choose or schools. The rationale in including these other versions was to capture phrasing conveying a support for vouchers—sometimes appearing as criticism of public schools—while excluding unrelated discussions of choice or schools.
Coders were trained to mark usage of the dog whistle terms first, then look at the context to be sure the meaning was what was indicated in the codebook before counting the usage. The coders were not instructed to look for specific racial groups in the context of the dog whistles; initially, this may seem counterintuitive. However, the development of race-relevant issue frames and historical usage of dog whistles, as noted in the preceding review, indicates that the value of using coded language is in the concomitant plausible deniability regarding race: because a specific racial group is not mentioned alongside a dog whistle, the speaker can deny race-related interpretations of those words.
Appendix 2: Codebook for Dog Whistle Text Analysis
1) Search for the following terms and mark the speech with the code (code in brackets).
a) Law [LO]
b) Schools [SC]
c) Illegal/criminal/alien [IM]
d) Islam/Muslim(s) [IS]
e) Welfare [W]
f) Voter [V]
2) For each term, consider the context.
a) Law: term must be “law and order”, e.g. “law and order”
i) Context: should be about raising safety concerns or discussing violence (not in immigration context)
ii) e.g. “we will return to law and order”
b) Schools: term must be words “school” and “choice”/“choose” together
i) Context: painting public schools in a negative light (“failing public schools”)
ii) e.g., “school of your choice”
c) Illegal: term must refer to “illegal immigrant,” i.e. not related to opponent’s actions
i) Includes: Criminal: “criminal immigrants” or referring to immigration as a crime or immigrants as criminals; Alien: as in “illegal alien”
d) Islam: term must be negative reference to Islam/ic/ist or Muslim/s
i) e.g. “radical Islam” or “Islamic terrorism”
e) Welfare: term must be welfare
i) Context: negative portrayal of welfare system or people who receive welfare
ii) e.g. “welfare mentality”, “welfare state”
f) Voter: term should be voting or voter in context of the need for “voter ID laws”, concern about “voter fraud”, or proposing voter ID as a way to fix the “rigged system”
NOTES
1There is little guidance in the qualitative literature for quantitative benchmarks that should be applied to content analysis from one source or messenger (Mayring, 2014; Stemler, 2000). This is because frequency data in content analysis are more commonly used for comparisons (e.g., Manganello & Blake, 2010; Nimegeer, Patterson, & Hilton, 2019; Vokey, Tefft, & Tysiaczny, 2013)—a method employed in addressing Hypothesis 2. The combination of quantitative techniques with qualitative content analysis is undertaken in addressing Hypothesis 1 as recommended by Neuendorf (2002) to strengthen the claims of validity when the analyses are aligned.
2Nixon’s usage is unique because the divisiveness of some issues, and therefore the dog whistles that reference them, did not develop until after his candidacy. As discussed in Appendix 1, I attempted to correct for this by including in the analysis dog whistles used in the late 1960s. Nixon’s inclusion is based foremost on the meaningfulness of comparison on the “law and order” dog whistle.
3Interrater agreement is the percentage of codes that are the same among the coders when the coding is finished. It is one way to measure the accuracy and reliability of the codebook used for that process. Krippendorff’s alpha is a statistic used to determine inter-rater reliability. Much like the statistic computed to estimate internal consistency of survey responses, Krippendorff’s alpha is a measurement of the extent of agreement among raters. Because it takes into account interrater disagreement (rather than just agreement), it is considered more stringent than inter-rater agreement percentages and a more robust estimate of reliability. It is expected that this statistic be above 80 in order to be able to claim good inter-rater reliability (Krippendorff, 2011).
4An ANOVA test is used, typically, to determine whether there is a significant difference among three or more groups on a specified criterion without the accumulation of Type 1 error that would come with iterative testing of groups in sets of two. If the F statistic in the ANOVA is significant, post-hoc tests can determine which groups are significantly different. Games-Howell post-hoc tests do not assume equal variances or equal sample sizes.