America First populism, social volatility, and self-reported arrests

Significance

Using the 2016 American National Election Study, we develop comprehensive measures of the current populist moment in the United States. Our purpose is to develop a behavioral analysis of this current socially volatile moment. Using hierarchical modeling, we find that political narratives of America First populism are connected to reported lifetime criminal arrests, and this holds when taking into account political leanings or the economic precarity facing individuals. While we make no claims of causation, our findings provide important clues about the social volatility of the current moment. We find that political beliefs of America First express and reflect economic frustrations, and that the social boundaries these narratives draw against perceived outsiders and internationalism are associated with lifetime criminal arrests.

Abstract

Despite research on the causes of populism and on the narratives of populist leaders, there is little empirical work on the relationship between populist attitudes and behavior, notably including criminal behavior. Our overarching concern is the recurrent social volatility of metaphorical populist themes that are central to impactful political messaging. Drawing on a national United States survey conducted around the 2016 election, we use multilevel models to show that the politically charged exclusionary boundaries of “America First” populism are behaviorally connected to increased odds of having been arrested. We argue that the rapid redrawing of social boundaries that make up populist attitudes is closely connected with the effects of economic and political frustrations during times of rapid social change. In the process, we develop a behavioral analysis of the social volatility of the recurrent populist movement in America.

We live in an era in which populist leaders have gained political traction over much of Europe and the Americas. Although the paths to populist power have varied across countries, they are largely attributable to a rising backlash against globalization and market deregulation introduced through the politics of Margaret Thatcher and Ronald Reagan (1).

What populism means varies across parties and places (2). Yet there is an emerging social science consensus—among sociologists, psychologists, and political scientists—that three political threads underwrite populist attitudes. These consist of 1) a belief that “the people” have been excluded or deprived of their social status, 2) a rejection of domestic and international elites in favor of nationalist grassroots political movements, and 3) an antagonism toward globalization and “outsiders.” These interconnected threads of populism emphasize that decent working-class people have been victimized by elites, by globalization, and by foreigners (3–5). These political threads are based on a redrawing of social, political, as well as criminal boundaries (5, 6).

The ascendancy of the American President Donald Trump features precisely these elements of populist sentiment. Trump’s commitment to “drain the swamp,” a driving force in the 2016 Presidential campaign, promised to establish boundaries between a working population and elites. His antiimmigration pronouncements involved references to invading rapists and murderers, and the promise that “I will build a great, great wall on our southern border” (7). And resistance to globalization was a central trope in Trump’s inauguration address: “From this day forward,” Trump insisted, “it’s going to be only America first, America first” (8). As we see below, these rallying cries are recurring themes in American political life and echo underlying politics of division in weakly regulated market societies (9). Yet despite research on the causes of populism and on the narratives of populist leaders (5, 10), there is little empirical work measuring the relationship between populist attitudes and behavior (11).

In this paper, we demonstrate that criminological research uniquely provides analytical and evidentiary leverage for understanding the current American circumstance. The rapid redrawing of symbolic boundaries that make up populist attitudes is closely connected with sociological criminology’s longstanding analyses of the effects of economic and political frustration during times of rapid social change (12, 13). This research demonstrates that antisocial behavior such as crime stems from socioeconomic frustrations over not being able to achieve social goals, with these frustrations exacerbated in societies with weakly protective social institutions (14–17). This has been found to occur even during times of economic prosperity, since inequality erodes the political legitimacy of social institutions (18). In this paper we build on this research on social strain to assess whether political narratives of populism are also behaviorally connected to lifetime criminal justice contact. In so doing, we develop an analysis of the current socially volatile moment in America.

Populism as Boundary Creation

The election of Donald Trump in the United States, the Brexit vote in the United Kingdom, and the rise of new political voices across Europe are understood as heralding a new era of populist politics (19, 20). Yet the concept of populism can describe a wide array of political positions. In the United States, both Donald Trump and Bernie Sanders are identified as populist for their denunciation of elites and establishment figures in the name of “the American people” (21). Precisely what populism entails is often ambiguous and uncertain (10, 22, 23).

As a political term, populism relies heavily on an idea of “the people” as wrestling back political power from distant elites (4, 9, 24). The concept dates to the agrarian left-wing People’s Party of the 1890s: People who saw themselves as victims of urban elites, of internationalism, of finance, and of foreigners and outsiders generally (25). Building on this American history, populist movements often center on a political ideal of the people’s sovereignty that is at the basis of democratic theory generally, but with an antagonism toward elites for being out of touch, and a resentment toward others that is motivated by both economic and democratic grievances (4). The populist base of these movements identifies itself as ethnically homogeneous, virtuous, and unfairly divested of authority and influence (26–28).

While the terminology of populism is often used in ambiguous and disparaging ways, there is increasing consensus across social science disciplines that the analytical core of populism lies in the redrawing of social and symbolic boundaries (4, 9). In their analysis of Donald Trump’s political speeches, Lamont et al. (5) rely on the study of boundaries to demonstrate how Trump relies on rhetorical strategies that align him with a virtuous, White working class that is united against the negative effects of globalization. This is amplified through strong moral boundary-demarcation against perceived outsiders, including undocumented immigrants, refugees, and Muslims (5). Elites, in turn, are then seen as coopted by these outsiders, so that authority and power need to be restored to a virtuous ordinary people (29, 30).

This populist boundary-making extends beyond political rhetoric. Research from Europe demonstrates that populist political narratives are largely matched by everyday attitudes, including antielitism, an emphasis on local sovereignty, and the moral importance of promoting a virtuous and ethnically homogeneous people (31, 32). Recent research suggests that these populist tendencies can have effects on social trust in institutions more generally. A study of German populist opinion, for example, finds that holding antielite views predicts cynicism and lack of trust in the media (11).

In this paper, we develop this approach to analyze whether there may also be behavioral correlates that are linked to this skepticism over institutions, and to the political drawing of the boundaries of “us” and “them” that underwrite populist attitudes. We investigate whether the content of populist symbolic boundaries tap into underlying socially volatile contexts. Our approach parallels criminological findings on legal cynicism (33), which demonstrates how skepticism about law enforcement is linked to detachment from state institutions, and may reproduce and intensify long-standing inequalities (33–35).

America First Populism

Populist tropes can be mobilized across the political array from left to right, both within the United States and in countries worldwide (36). Within the current American context, the populist redrawing of symbolic boundaries is especially articulated through the concept of “America First” that is at the core of President Trump’s messaging. “I like the expression,” Trump told the New York Times during his election campaign: “I’m ‘America First’” (37). Since his election, President Trump has continued to frame his core domestic and international policies through this analytical lens, repeatedly declaring “America First”' to be “the major and overriding theme of my administration” (8).

The language of America First is a longstanding populist theme in the United States. The lynchpin of America First populism has been its antiinternational and antiforeigner emphasis, which are underpinned by antielite rhetoric and a sense that a virtuous people is being repressed. In the 1930s, this was expressed in the form of widespread resistance to intervention in World War II, and opposition to financial support for those fighting against Nazi Germany, and was called the “Emergency Committee to Defend America First” (38). In popular culture this is also chronicled by the novelist Philip Roth (39), whose book The Plot Against America provides a dark account of what might have happened if the America First Committee had gained power. Historian Timothy Snyder (40) forcefully argues that the current incarnation of America First romanticizes a 1930s view of the United States as a period of national virtue, and resentful of costly expenditures on internationalist policies. Today, the resurrection of the idea of America First is similarly antiinternationalist, with a narrative of ordinary Americans being the economic victims of an increasingly globalized economy. President Trump persistently invokes these American nationalist themes, proclaiming “America first, yes, we will not be ripped off anymore. We’re going to be friendly with everybody, but we’re not going to be taken advantage of by anybody” (37).

Both in the 1930s and in the current political conjuncture, domestic feelings of economic threat and social exclusion are blamed on internationalism. Recent research demonstrates that economic frustrations inform this resurgence of antiinternational resentment. In her ethnographic work with White Louisianans, Hochschild (41) finds that current populist narratives tap into a sense of economic precarity, with a loss of social stability largely blamed on globalization, on offshoring jobs, and on immigrants as a new and rising source of competition for the White working class. Hochschild (41) summarizes these collectively held sentiments as reflecting a pervasive “lost hope.” Lamont (42) similarly argues that globalization and unrestrained markets are creating fewer realistic options for achievement of the American dream, rendering it less effective as a collective myth, and leading to widespread hopelessness. In turn, the resulting pessimism and anger lead individuals toward right-wing populist views (42, 43). This sense of economic hopelessness is found to underwrite new exclusionary symbolic boundaries against foreigners and others who are not “us” (20, 42, 44). Taking these ideas together, Dodd et al. (20) thus conclude that “[i]nequality informs the Brexit/Trump phenomenon,” with exclusionary boundaries against minority group members mobilized in response to the inequalities produced by globalization (44).

Hochschild’s “lost hope” (41) or Lamont’s “hopelessness” (42) echo findings that have long been at the core of criminological research. Rosenfeld and Messner (45) argue that the relentless pursuit of material wealth that is at the core to the American dream, combined with the competitiveness that this pursuit produces among individuals, heightens rates of crime, and is acutely experienced among those less able to achieve economic success. This strain results in higher crime rates (46–48). Yet what we are currently witnessing is that this stress does not turn exclusively on a lack of wealth or the failure of economic growth. Rather than a lack of economic growth, it is inequality that is at the core of current frustrations, along with the visible concentration of wealth, resources, and lack of taxation enjoyed by the “one percent” (49–51). In criminology, this frustration is what LaFree (18) points to in signaling that inequality and a lack of social services, even during a time of economic growth, corrode institutional legitimacy in families, in the economy, and in civil society broadly, and that this is connected to crime. This inequality is the context in which the current manifestation of America First populism is being heard, and within which exclusionary boundaries are mobilized.

Social Volatility, Populism, and Political Scripts

We build on research findings that current cultural claims within America First populism are the outcome of economic frustration and perceived hopelessness. This social strain leads to strongly drawn exclusionary boundaries against perceived outsiders, with international engagement, immigrants, and refugees identified as causes of the “lost hope” that Hochschild (41) detects, and the “hopelessness” that Lamont (42) links to an increasingly elusive American dream. We connect this research with work in criminology on the effects of social strain, which finds that frustrations over economic inequality can be connected with a loss of institutional legitimacy and to crime.

Our overarching hypothesis is that the nationalistic political processes of America First populism harness and amplify social volatility as measured through nonnormative behaviors such as crime and delinquency. That is, we hypothesize that these underlying experiences are connected with political scripts and are amplified by an institutional context of inequality and the relative absence of social supports. Our thinking about social volatility is informed by sociological research on political conflict and change (52), and which we operationalize here through a measure of lifetime criminal justice contact.

Research from outside of the United States provides empirical support that social volatility is connected with resurgent economic and political scripts. For example, Hagan et al. (13, 53) conducted research with German youth following the collapse of the former German Democratic Republic. They found that the unrestrained economic aspirations of youth in the East, combined with cultural scripts emphasizing economic and social competition, are connected to youth violence (13, 53). Importantly, this social volatility was detected through involvement in everyday crime as well. In parallel work, Hagan et al. (13) found that beliefs, such as hostility toward foreigners and about the importance of restoring Germany’s “earlier greatness,” combined with anomic economic aspirations to predict school vandalism and violence. They concluded that “the economic challenges and the changing social and political order of a unified Germany raise concerns about the re-emergence of subterranean traditions of right-wing extremism,” and that these combined with a pattern of drift into everyday sorts of delinquency among youth (13). More recently, Hagan and Rymond-Richmond (54) identified analogous sources of participation in mass violence in Sudan’s Darfur region, where the combination of economic scarcity and political narratives were linked to the dehumanization of outgroups and predicted targeted interethnic violence and atrocities (54).

Each of these cases reflects situations of weakening formal economies and economic scarcity (55–57). Yet there is reason to believe that the same process may currently be ongoing in the United States. Speaking about changes in United States democracy, Snyder (58) explicitly connects current economic inequalities and political messaging with crime and violence. Snyder suggests that “violence is hugely important” to the current populist era, and that it builds on “fictional threats such as those posed by immigrants” that is augmented by economic concern, which “makes it hard for people to feel like they’re living in the same society,” so that material concerns “probably also contribute indirectly to violence” (58). Snyder’s hypothesized link points to economic inequality as a source of threat and hopelessness that is linked with perceptions of outsiders and internationalism, and with an indirect connection to crime. Some initial evidence supports this link, with empirical support for a “Trump effect” whereby reported hate crimes surged following President Trump’s election (59). We therefore get a sense, although empirically undeveloped, that the antiinternational and antiforeigner dimensions of America First populism might also be connected with an underlying social volatility, and that this can be measured through connection to crime and delinquency.

Methods and Analysis

Analysis.

The data for the present paper are drawn from the American National Election Studies (ANES) 2016 Time Series Study. Since 1948, the ANES have been conducting analyses of public opinion and voting behavior during United States elections. Here, we use the complete dataset, which was rereleased in 2018. The sample size is 4,270, including a combination of individuals interviewed both face-to-face (n = 1,180) and through the internet (n = 3,090). Both face-to-face and internet interviews are from independently drawn probability samples for the majority of the United States, with the sampling frame derived from residential addresses where mail is delivered. The sampling universe is thus United States eligible voters. Respondents had to reside at the sampled address, be a United States citizen, and 18 y or older.

Respondents were interviewed before the 2016 United States Presidential election, between September 7 and November 7, 2016. After the election on November 8, as many respondents as possible were interviewed again between November 9 2016, and January 8, 2017. Our analyses are based on the full sample, including both interview modes and questions from both before and after the election. To accomplish this, we used ANES weight V160102 (60). Further details on ANES 2016 can be found in the user’s guide (61).

As we discuss below, the ANES data are particularly well-suited for our analysis of populist attitudes. The dataset includes variables that capture the range of populist attitudes that have been identified in social science research, a wide array of socioeconomic data on respondents and their families and friends, and a measure of reported lifetime criminal arrest.

Because this is a national dataset and in order to account for within-state correlation error, we used multilevel models to assess the relationship between arrests and other variables of interest. These analyses were performed using the melogit command in Stata, v14.

Altogether, removing those with missing data on key variables, we retained an analytic sample of 2,422 individuals. Given inconsistent recommendations about how to impute data at multiple levels (62–64), we proceeded with listwise deletion, yet additional models run with multiple imputation present the same findings as those described below. We have 51 states in our analytic sample when including both internet and face-to-face modes. On average, states have a size of 83.75 individuals, ranging from 4 to 414 respondents. As a further robustness check, we dropped observations from states with fewer than 30 observations and our findings hold. Finally, to detect multicollinearity, we reviewed the variance inflation factor (VIF). This was 1.23 and so did not exceed the standard threshold of VIF = 2.00 (65).

Predicted Variable: Self-Report Arrests.

Given concerns over official data, particularly when it comes to lower-level offending and delinquency, criminological analyses often rely on survey self-reports of behavior (66). The importance of this methodological approach is magnified by the need to rely on richer measures of independent variables than official arrest data provide, so that criminological research in this field prioritizes survey data for both independent variables and dependent variables of nonnormative behavior (67, 68). In this paper we relied on a self-report measure of lifetime criminal justice contact. Respondents were asked, “Have you ever been arrested, or has that never happened to you?” This was coded into has been arrested = 1 and never arrested = 0; 20.46% of the sample report having been arrested. We have confidence in this retrospective measure given that an event such as an arrest is less subject to errors in recall that would be at issue with self-reported delinquent behavior more broadly (69, 70). This is further reinforced by research in social psychology, which indicates that rarer events are better remembered than frequent ones, and by the fact that this question does not ask respondents to recall a date or time period, which has been shown to be prone to recall error (71).

Predictor Variables: Controls.

We controlled for standard demographic variables, such as gender (female = 1, 52.9% of the sample), level of education (from 1 [less than first grade] to 16 [doctorate degree]), income,* and age (ranging from 18 to 90, mean = 49.6, SD = 17.58). We included race as three dummy variables—Black (9.4% of the sample), Latinx (10.6% of the sample), and other (8.3% of the sample)—with White (71.7% of the sample) as the reference category. Furthermore, we controlled for ideological identification based on self-placement on a 1 to 7 scale, from extremely liberal to extremely conservative (mean = 4.18, SD = 1.6).

While we are interested in the relationship between populist beliefs and crime, lifetime arrests could be attributable to a range of processes, whether they be person-level attributes or policing practices. To account for these, we included two further variables in our analyses. First, we included a measure of self-control, which is a combination of two variables: capturing how difficult it is for the respondent to control their temper (from 1 = easy to 5 = very difficult), and capturing whether, when provoked, they are likely to hit someone (from 1 = unlikely to 5 = very likely), such that the measure of control ranges 0–10, with higher scores denoting less self-control (mean = 2.56, SD = 1.66). These measures capture the dimension of “violent temper” that past research has identified as predicting both drug use and violent crime among youth; so, also, is this dimension likely to capture everyday forms of offending that result in arrest over the life-course (72–74). Second, we included a binary measure of police stops of the respondent or their family members during the past 12 mo (23% of the sample reported they were stopped or questioned), to address the possibility that our findings are a result of the concentration and frequency of police arrests, and thus mainly a measure of police behavior (75, 76).

Finally, in order to approximate a respondent’s sense of proximate economic threat, we included a binary measure of whether a respondent’s family or friends lost their job in the past year (43.4% of the sample answered yes). And to get at more contextual economic pressures, we included two state-level variables from the United States Census Bureau: The percent of the state’s population living in poverty (3-y average 2014 to 2016, ranging from 6.9 to 20.8, mean = 13.64, SD = 2.72), and the change in percent of people in poverty (3 y, 2014 to 2016, ranging from −4.9 to 2.1, mean = −1.65, SD = 1.2).

Predictor Variables: Populism.

We included three indices in our analyses, all of which pertain to specific dimensions of populism: 1) distrust of elites; 2) feeling excluded from the political system; and 3) a measure of what we call America First populism. The first two match broader analytical dimensions of populism, namely, that “‘elites’ harm the ‘people’.” The latter, meanwhile, is more specific to the current moment in American politics.

ANES 2016 includes a feeling thermometer for different groups or individuals, requiring respondents to rate 0 to 100 from very poor to very positive perceptions of the actor in question. For our antielite measure, we included reverse-coded perceptions of scientists, the US Supreme Court, feminists, and liberals, such that higher measures imply worse perceptions.

To test whether we could combine these measures into an index, we conducted a principal axis factor analysis. This supported the unidimensionality of this measure, since only one factor emerged with an eigenvalue above 1 and subsequent factors indicated a substantial decline in eigenvalues. We therefore combined these four variables into an index by averaging the four items. The full index ranges from 0 to 100, with a mean of 39.17 and an SD of 16.61 (α = 0.68).

The second element of populism is a sense of the people feeling disregarded by elites. To measure this construct, which we call “no say” populism, we used three variables. The first asked respondents how strongly they agree (1 to 5, from agree strongly to disagree strongly) with “[p]ublic officials don’t care much what people like me think.” The second is a measure of agreement (1 to 5) with “[p]eople like me don’t have any say about what the government does.” The third measured agreement (1 to 5) with “[m]ost politicians do not care about the people.”

We again conducted a principal axis factor analysis for these three items, which similarly indicated the unidimensionality of this measure. We then averaged the three variables and created an index of perceptions that the people have no say, reverse-coding each of the variables such that higher values mean stronger feelings of the people being marginalized from the political process. This index ranges from 1 to 5, with a mean of 3.37 and an SD of 0.91 (α = 0.70).

Our final measure is of America First populism. Following Lamont et al. (5), we postulate that the current moment of American populism has a number of distinctive features: Vilification of globalization combined with strong moral boundary exclusion of immigrants, refugees, and Muslims. We also observed a generalized sense of retrenchment from the world, in which survey respondents answered that “we’re not going to be taken advantage of by anybody.” These elements form our index of America First populism.

First in this index we included negative perceptions of immigrants. On these three variables, respondents report how strongly they agree (1 to 5, from agree strongly to disagree strongly) with the following statements: “[i]mmigrants increase crime rates in the United States,” “America’s culture is generally harmed by immigrants,” and “[i]mmigrants are generally good for America’s economy.” We reverse-coded the first two variables, so that higher values capture more negative perceptions of immigrants.

Second, we included a measure of negative views of refugees, and more precisely, whether respondents oppose Syrian refugees coming to the United States. Respondents were asked, “Do you favor, oppose, or neither favor nor oppose allowing Syrian refugees to come to the United States?” with responses ranging from 1 to 7, indicating favoring a great deal to opposing a great deal. In constructing our scale (i.e., with all variables in the same range), we recoded this measure by combining “favor a moderate amount” and “favor a little,” or conversely “oppose a little” and “oppose a moderate amount.”.

We also included perceptions of Muslims, namely a feeling thermometer (ranging 1 to 100), which we reverse-coded so that higher values indicate more negative perceptions of Muslims, and recoded into quintiles.

As a proxy for negative feelings about globalization, we included a respondent’s feelings on international trade. This is a three-category variable where respondents were asked, “Have increasing amounts of trade with other countries been good for the United States, bad for the United States, or neither good nor bad?” We recoded this so that trade = 1 if a respondent thought trade with other countries has been good for the United States, trade = 3 if trade is deemed neither good nor bad, and trade = 5 if trade is deemed bad.

Finally, we included two additional variables that get at both perceptions of globalization and the larger sense that “we’re not going to be taken advantage of by anybody.” First, we included a measure of whether respondents agreed or disagreed with “[t]his country would be better off if we just stayed home and did not concern ourselves with problems in other parts of the world” (agree = 5; disagree = 1). Second, we included a respondent’s level of agreement (1 to 5) with the statement: “[t]he world would be a better place if people from other countries were more like Americans,” which we reverse-coded.

Our index of America First populism thus includes eight variables measuring perceptions of immigrants, refugees, Muslims, trade, and a general sense of disengagement with the rest of the world. As with the other two indices, a principal axis factor analysis supports the unidimensionality of this measure. We thus averaged individuals’ scores on these eight variables, ranging 1 to 5 (mean = 2.73; SD = 0.89), with higher scores indicating stronger agreement with America First populism (α = 0.75).

Findings.

Our first model (model 1 in Table 1) includes demographic variables. As in much prior research, we find that women (odds ratio [OR] = 0.30; P < 0.001), those with more education (OR = 0.91; P < 0.001), or more income (OR = 0.96; P < 0.001), and those who are older (OR = 0.99; P < 0.001) all have lower odds of having been arrested. In this model, none of our race dummy variables (with White as the reference) have a statistically significant association with lifetime arrest.

Table 1.

Multilevel logit models of arrest, ANES 2016 (ORs shown with SEs)

	Model 1			Model 2			Model 3
Variable	Odds ratio	SD error	P > |z|	Odds ratio	SD error	P > |z|	Odds ratio	SD error	P > |z|
Female	0.30	0.03	***	0.29	0.03	***	0.30	0.03	***
Education level	0.91	0.02	***	0.91	0.02	***	0.92	0.02	**
Income	0.96	0.01	***	0.96	0.01	***	0.97	0.01	**
Age	0.99	0.00	***	0.99	0.00	**	1	0.00
Black	0.77	0.16		0.73	0.15		0.68	0.15
Latinx	1.14	0.20		1.14	0.19		1.23	0.22
Race other	0.93	0.18		0.90	0.18		0.91	0.18
Lost job (family and friends)				1.51	0.15	***	1.37	0.14	**
Alternative explanations of lifetime arrests
Self control							1.14	0.05	**
Police stop past 12 mo							2.02	0.23	***
Populism and political orientation measures
Conservative							0.87	0.04	**
Antielite populism							1.00	0.00
No say populism							0.99	0.06
America First populism							1.25	0.10	**
State level measures
Percent of state in poverty				1.03	0.02		1.03	0.02
State change in poverty				1.02	0.06		1.03	0.06
Constant	2.91	0.90	**	1.60	0.69		0.55	0.31

***P < 0.001, **P < 0.01, *P < 0.05.

In model 2, we added both individual and state measures of economic threat. At the individual level, we added a measure of whether a respondent’s family or friends lost their job in the past year. We find that respondents whose proximate individuals have experienced job loss have a 51% increase in the odds of having been arrested (P < 0.001). Neither of our state level variables, whether poverty levels or changes in poverty levels, have a statistically significant association with arrest. We note that in this model, we see the same patterns as in model 1. Gender, education, income, and age still have a statistically significant association with arrests, related to decreased odds of having been arrested.

In model 3, meanwhile, we added in the three populism indices we created, namely “antielite” populism, “no say” populism, and “America First” populism. As further controls in this model, given that populism is associated with both sides of the political spectrum (21), we included a measure of political orientation, operationalized as a self-reported conservativism scale, as well as the two alternative explanations of lifetime arrests, namely self-control, as well as whether the respondent and her family experienced police stops over the past 12 mo. With the exception of age, which is no longer significant, our previous findings are unchanged, suggesting that the relationship between the demographic variables and arrests is relatively stable, as is the association between proximate economic precarity and the odds of having been arrested. We therefore see further evidence that a sense of economic threat is indeed significantly associated with lifetime arrest.

When considering the new variables in the model, we see that the sense of threat implicated in America First populism is also a statistically significant predictor of lifetime arrest. Thus, a one-point increase on the America First scale is associated with a 25% increase in the odds of having been arrested (P < 0.01). This holds net of the demographic and economic threat variables in the model, and importantly net of political conservatism, which we find to be associated with a statistically significant drop in the odds of having been arrested (OR = 0.87; P < 0.01). This relationship is also net of alternative explanations of lifetime arrest, with both a respondent’s self-control (OR = 1.14; P < 0.01) and family experience of police stops (OR = 2.02; P < 0.001) statistically significant predictors of lifetime arrests. The other populism measures in our model, whether no say or antielite populism, are not significantly associated with arrest.

Taking these data together, we see that lifetime arrests are associated with both economic precarity and the perceived threats from immigrants, refugees, and global engagement that are articulated through America First populism. When plotting the relationship between these two variables and arrest (Fig. 1), we see that America First populism is related to a general increase in the odds of having been arrested, such that a higher score on the index is associated with higher odds of arrest. We also see that respondents with family and friends who have lost jobs (that is, the red line in the model in Fig. 1) will generally see a further increase in the odds of having been arrested. We thus find that economic threat and the symbolic boundaries implied by America First populism are both significantly associated with lifetime arrest.

Fig. 1.

Predicted odds of arrest, by America First populism and proximate job loss.

Conclusion

We used multilevel models and found that across the United States hostility toward foreigners and skepticism about international engagement are linked to lifetime self-reports of criminal arrests. We found, moreover, that holding America First populism and other variables constant, populist political claims, such as antielite sentiment, as well as a feeling of political disempowerment, are not positively connected with arrests. In other words, it is the recurrent beliefs of America First and perceptions of others as threatening—which are mobilized through exclusionary social boundaries, and which are persuasively depicted in writings ranging from the prominent historian Timothy Snyder (58) to the popular novelist Philip Roth (39)—that we find stubbornly associated with police arrests.

We suggest that this provides evidence that America First narratives, and the social boundaries they draw regarding immigrants, Muslims, and globalization, tap into and amplify strands of volatility. It is important to note that this finding, and what we describe as social volatility, are related to the holding of populist beliefs regarding social boundaries, rather than political orientation or voting patterns among respondents. Indeed in our model, conservative political views are instead associated with decreased odds of having been arrested.^† In addition, criminal arrests are exclusively associated with just one dimension of populism, namely the antiimmigrant, anti-Muslim, antitrade, and antiinternationalist beliefs that are closely associated with America First populism. We highlight that these are symbolic boundaries that are drawn against perceived outsiders. In contrast, a broader distrust of elites and a sense of political exclusion—although part of current populist political narratives—are not associated with lifetime arrests in our models.

In their recent work on populism, eminent political scientists Norris and Inglehart (36) highlight the move toward authoritarian populism in the United States and the United Kingdom. They end their book by asking “whether democratic cultures are sufficiently robust to resist the associated dangers,” including the damage that this political turn may do to civic culture, social tolerance, and through violent behavior (36). These recurrent symbolic boundaries and relationships are also found in America First populist views—which have involved expressing insularity and racist views since their early mobilization in the 1930s—and are central to the policy statements of the Trump administration. These views, in other words, tap into a boundary-setting process between “us” and “them” that is itself socially volatile. Similar to criminological research among rebellious youth during the aftermath of German unification (13), we find that in the current United States political circumstance, negative and hostile views of out-groups and internationalism are linked to crime and arrests. Given that we rely on a measure of lifetime arrests generally, we note that this is connected with a wide range of underlying nonnormative behavior, and extends beyond direct behavioral manifestations of hostility to outsiders such as hate crime.

We believe this provides unique insight into our current circumstance. Within the discourse of America First populism, claims over these symbolic boundaries are at the core of defining the authentic nation, making up what Bonikowski (77) refers to as an “ethno‐nationalist populism” that is fueled by resentment. This dimension of American populist discourse explicitly identifies the economic strains and hopelessness that people are experiencing with global trade, and ongoing social change with the presence of minorities and immigration to the United States (77, 78). In contrast to other populist claims of economic and political inclusion, this is the element of America First populism that taps into social volatility, and which also mobilizes and legitimates exclusion (77). This further resonates with research in social psychology, which identifies in- and out-group characterizations as responses to status threat and competition (79), and with research on when democracies turn to violence (80). Importantly, our point is not that these narratives themselves cause social volatility, but that these social views tap into a resentment of others that is empirically connected to criminal justce contact and is further connected with perceived threats to status at a time of intense concern over economic inequality.

Of course, despite controlling for other police arrest practices it may still be that we are tapping into police behavior as much as individual crime. Here too we identify arrests as a proxy for broad social volatility. Recent research finds that police stops can increase crime and decrease psychological well-being (81). We find further support for this thinking of the link between arrests and volatility from the suggested relationship between arrest and the continued reproduction of low self-control, as well as the weaker labor market outcomes that result from arrest (73, 82, 83). While work to date has sought to connect populist view to crime through a theoretical and causal model of incitement and hate crime, lifetime self-reported arrests, in our model, speak to a dynamic and volatile process that in the current period is empirically connected to the denigration of foreigners and of internationalism, under the populist rubric of America First. We believe this gives us a lens into the relationship between behavioral correlates and symbolic boundaries that are tapped into during a socially volatile era.

We emphasize that we make no claims of causation in our analysis. As we indicate above, the social volatility we identify and measure through lifetime arrests may work in several ways. Our basic finding is that symbolic boundaries that favor exclusion of immigrants and refugees, and which also reject globalization, are predictive of criminal conflicts with the law, measured through self-reports of arrests across respondents’ lifetimes. Methodological approaches to establishing causal order are a persistent debate in establishing individual-level engagement in crime, with an emphasis on the degree to which criminal careers are static or dynamic over time (73, 84, 85). Further research could usefully identify the proximate and ultimate sequences that are involved in the relationships we observe and establish the recurrent role of politically charged boundaries in different historical moments (86). In addition, while the dichotomous measure we rely on allows us to capture a broad relationship between the holding of these America First views and the prevalence of arrest across our sample (87), it leaves unanswered the question of whether the incidence of arrests is related to holding these values more intensely, and whether the holding of these values waxes and wanes with the timing of criminal justice contact. The first important step that we have taken here, however, is to observe that such symbolically important relationships exist.

Our findings offer some important clues about the instabilities of the current era. We find that political narratives express and reflect economic frustrations, and that the social boundaries that are drawn politically are also behaviorally associated with lifetime criminal arrests. If we have not captured the heart of a causal process, but simply a conjuncture of politics, symbolic boundaries, and social volatility, we nonetheless contend that this connection cries out for explanation, and that this is of vital importance.

Data Availability.

ANES data can be obtained at https://electionstudies.org/data-center/2016-time-series-study/.