Faith in Trump and the willingness to punish white-collar crime: Chinese Americans as an out-group.

Objectives: The first goal of the study was to investigate the willingness of former President Trump's supporters to punish a particular form of white-collar crime (i.e., bank fraud). The second objective was to test whether the race of the person who committed the bank fraud influenced Trump supporters' willingness to punish.
Methods: This study used data from factorial vignettes that were administered to a national sample of adults in 2021 (N = 1509). A 2 (race of the individual who committed bank fraud) × 2 (prior criminal record) × 2 (COVID-19 related fraud) between-subject experimental design was used. Multivariate techniques were used to regress the dependent variables (e.g., length of prison sentence) onto the faith in Trump scale, the experimental conditions, and other variables.
Results: Participants who expressed a strong faith in Trump were less likely to support sending an adult male who committed bank fraud to prison, but they were more supportive of deporting the individual to another country. The effect of faith in Trump changed when the race of the person who committed bank fraud was manipulated. Specifically, participants who expressed greater faith in Trump were more likely to view bank fraud as harmful and wrong, more likely to support the use of prison and recommend longer prison sentences, and expressed greater support for deporting the individual when he was depicted as Chinese American. Conclusions: Allegiance to the former president likely increased the targeting of Chinese Americans as out-group members in need of greater social control. Supplementary Information: The online version contains supplementary material available at 10.1007/s11292-022-09528-8.

Research suggests that US presidents are capable of shaping public opinion. They represent the country and are the lead policy maker; their words matter (Cohen, 1995, 1997; Hawdon, 2001; Shi et al., 2020). This thesis is consistent with the social constructionist view that positions espoused by political elites shape rather than are a response to the will of the public (Beckett, 1997; Pickett, 2019). Donald Trump honed his ability to shape public opinion into an art form. He claimed that he alone could “Make America Great Again” (MAGA). His campaign rallies were attended by thousands of supporters. Trump flooded the airways with Tweets—more than 23,000 during his presidency, reaching close to 80 million followers before his account was shut down (Madaminov, 2020; Twitter Followers, 2021). He was a frequent call-in guest on Fox News shows, a network with 3.5 million viewers (Katz, 2020). As his presidency unfolded—and beyond—Trump demanded fidelity from Republican office holders, seeking to admonish those who crossed him. Despite being twice impeached, he eventually achieved hegemonic control over his party.

A key issue is why a healthy slice of the American electorate not only voted for but also have a deep faith in Trump. In his Moral Foundation Theory, Haidt (2012) argues that moral intuitions fall into two distinct orientations: (1) “individualizing,” emphasizing caring and fairness, and (2) “binding,” emphasizing hierarchical authority, group loyalty, and sanctity (e.g., religion; see Graham et al., 2009). Republican politicians have long connected with their voters by trumpeting binding moral intuitions. Ronald Reagan campaigned on the slogan that “We can make America great again” (Ronald Reagan, 2021) and John McCain on the slogan of “Our country first” (Grollman, 2020). Underlying these slogans was often an implicit appeal to racial resentment, part of a so-called “southern strategy” that stoked White dislike of welfare distributed to undeserving Blacks (e.g., “welfare queens”; see Levin, 2019). Trump not only followed this playbook, he amplified it (Newman et al., 2020).

Trump also tapped into the sense among some Whites that they were now an in-group whose cultural and demographic status was being threatened by people of color, whether they were from the urban core, Muslim nations, or south of the border. Attacks on legal immigration were prominent in Trump’s messaging and policies (Ibe, 2020). This distinct out-group threat was captured by Hochschild (2016) in the title of her book, Strangers in Their Own Land. As Fording and Schram (2020, p. 40) note, the MAGA slogan was decoded by some as “Make America White Again.” Evidencing as well a strong “hegemonic masculinity,” Trump also promised to protect “good Americans” from the attacks by liberals and feminists—promising to lock up Hillary Clinton who devalued the common people as “deplorables.” He would protect the White Christian order (Whitehead & Perry, 2020). According to Trump, statues commemorating Confederate heroes of the Civil War should stay put, and transgendered people had no place in the military. He claimed to have his supporters’ backs.

Willing to insult and bully just about anyone, Trump tapped into the deep moral intuitions of his followers and thereby served their “emotional interests” (Hochschild, 2016). Although his economic policies favored the rich, his cultural campaign and populist messaging led many Americans to believe that Trump “understood them” and “made them feel safe.” The Republican 2020 platform was reduced to “anything Trump said.” The average voter was happy to follow his lead. Faith in Trump mattered.

Substantial evidence exists of Donald Trump’s capacity to influence the views of his followers. Polls show that once Trump voiced a strong position, Republican voters would conform their opinion to his. This conformity has been manifested in their sudden opposition to mail-in voting during the 2020 presidential election, favorable opinion of Vladimir Putin and Russia, resistance during the COVID-19 pandemic to wearing masks, embrace of the idea that Trump, not Biden, won the 2020 election, and the claim that the January 6th protestors were exercising their free speech (Bump, 2022; Cillizza, 2021; Clinton et al., 2020a; Clinton et al., 2020b; Easley, 2021; Itkowitz, 2021; Morris, 2021; Oliphant & Kahn, 2021).

Two relatively recent experimental studies showed that a connection to Donald Trump can influence a range of political policy preferences. First, using survey data collected immediately after Trump’s 2017 presidential inauguration, Barber and Pope (2019) examined respondents’ views toward 10 policy issues (e.g., raising the minimum wage, increasing taxes on the wealthy, background checks on gun purchases) that Trump had taken both conservative and liberal positions on. This fact allowed respondents to be randomly assigned to a conservative or a liberal Trump treatment. Depending on which view they were told Trump held (the “Trump cue”), the policy attitudes of “strong Republicans” swung in that direction, suggesting that party allegiance “trumped” political ideology. Note that those who were most approving of Trump were more likely to conform their policy preference to the Trump cue.

In the second study, Kunst et al. (2019, p. 1184) investigated how White Republicans’ allegiance to Donald Trump affected their “willingness to violently challenge elections and to persecute—with legal or authoritarian justifications—religious, immigrant and political out-groups” (e.g., Muslims, Iranians, immigrants at the southern border). Drawing from self-expansion theory, they examined the respondents’ “identity fusion” with Donald Trump. This construct emphasizes people having a “visceral feeling of oneness (that is, psychological fusion) with a political leader” (p. 1180). This fusion can have a “self-enhancing function,” increasing a sense of self-confidence as a means of coping with economic insecurity and a feeling that “they and their group are losing ground” (p. 1180). A sample item from the scale is: “I am one with Donald Trump” (p. 1185). Notably, identity fusion with Trump was associated with out-group animus and the willingness to consider political violence.

Of special relevance to the current project, a research team led by Graham and Cullen conducted a national survey in the beginning stages of the COVID-19 pandemic. One study by Graham et al. (2020, p. 8) measured intentions to defy social distancing norms. The second study assessed social distancing techniques of neutralization (Cullen et al., 2021, p. 9). Controlling for a range of political, COVID-related, and socio-demographic variables, faith in Trump was strongly associated with neutralizations and defiance intentions. Both studies contained a multiple-item measure on how much faith that the respondents had in statements that Trump made about COVID-19. More important, Graham et al. (2020, p. 9) developed a psychometrically-sound five-item “faith in Trump” measure. These items drew from statements made by Trump about his unique views and abilities, such as he can protect America from “threats around the world.”

Building on this work, the current study investigated how “faith in Trump” affects the willingness to punish under certain circumstances. The analyses focus not on street crime but on a form of white-collar crime—bank fraud (i.e., illegally obtaining loans) to support struggling businesses due to either COVID-19 or other reasons. Given Trump’s critique of government efforts to curtail COVID-19, his own alleged business corruption, and his willingness to pardon upperworld lawbreakers, one possibility is that loyalty to Trump may insulate against a willingness to punish white-collar offenders (Cohen, 2019). The analyses also focus on how willingness to punish is affected by faith in Trump when the perpetrator of the offense is a member of a racial minority group. Much attention is focused on Trump’s animus toward various racial/ethnic groups. However, during his 2016 election campaign, Trump criticized China for unfair trade practices. His anti-China rhetoric accelerated with the COVID-19 pandemic. At issue is whether those with faith in Trump harbor animus that would make them more likely to punish a Chinese American than a White American. Such a finding would suggest that allegiance to Donald Trump has the effect of eliciting negative feelings. These issues are examined using an experiment embedded in a survey that was administered to a national sample of adults. Two additional issues are discussed below: white-collar crime during the pandemic and former President Trump’s growing animus toward the Chinese.

Extent and opportunities for COVID-19 fraud

During the COVID-19 pandemic, available data suggest unprecedented levels of white-collar crime—which includes fraud by small business owners against the federal government and financial institutions (Bruggeman, 2021). Many circumstances during the pandemic created new opportunities for fraud. Working remotely became the norm for businesses and for officials responsible for investigating and prosecuting white-collar crime, resulting in less oversight and enforcement (McDermott, 2021). In spring 2020, then-President Trump signed the largest economic stimulus package (2.2 trillion USD) in US history into law (Coronavirus Aid, Relief, and Economic Security Act, 2020). The CARES Act included a number of programs intended to provide relief to struggling small businesses and employees, such as the Paycheck Protection Program (PPP) and Economic Injury Disaster Loans (EIDL). Although many government programs are prone to fraud, the pressure to promptly provide funding while simultaneously conducting a thorough review of applications makes emergency loan programs particularly vulnerable (Bailey et al., 2022).

Much like other crises requiring governmental intervention (e.g., FEMA’s response to Hurricane Katrina), the COVID-19 pandemic provided criminally motivated individuals with many opportunities (Dilanian & Strickler, 2022; Federal Bureau of Investigation, 2008; Kratcoski, 2018). Between March 2020 and April 2021, the Small Business Authority (SBA) Office of the Inspector General—who helps investigate applicants and participants in governmental programs like the PPP and EIDL—received over 150,000 complaints of loan fraud, up from around 1000 in 2019 (Schwellenbach & Summers, 2021). While elements of these fraud cases varied, many involved creating fictitious businesses to apply for the emergency loans, or otherwise misrepresenting application information (e.g., by using details from a legitimate business or not reporting prior criminal convictions) to obtain funding. To secure bigger loans, tax information and payroll numbers are frequently falsified (United States Department of Justice, 2021). Other criminally motivated individuals applied for smaller loans from multiple lenders to avoid generating suspicion (Bailey et al., 2022). In May of 2021, the COVID-19 fraud task force was established to work with other government agencies to prosecute the most egregious cases of fraud (Ngo, 2022). Underscoring the seriousness of COVID-19 fraud, a special prosecutor was also appointed, and nearly every US Attorney’s Office now has a PPP Loan Fraud Coordinator position tasked with identifying and prosecuting fraud (McLain et al., 2021; Morse, 2021).

The costs of SBA-related fraud

Estimates of taxpayer funds lost due to fraud in the PPP, EIDL, and other SBA-affiliated loan programs (e.g., unemployment insurance) are staggering: potentially 80 billion USD to the PPP alone, and between 78 and 100 billion USD in the EIDL program (Dilanian & Strickler, 2022). Examples of excessive greed are prevalent, with individuals spending fraudulently received COVID-19 relief funds on sports cars, travel, jewelry, and even a rare and valuable Pokémon card (Dilanian & Strickler, 2022; Ngo, 2022). Victims include the federal government, lenders, the American taxpayers, and legitimate businesses who needed funding to stay afloat. Many of the known perpetrators fit the profile of a “typical” white-collar offender in that they held high-level positions and are considered respectable members of their community (Hollis, 2022). Like other instances of fraud, the emotional toll on victims and other consequences—such as decreased confidence in governmental institutions—are difficult to measure.

Extent of prosecution of the fraud

To date, the United States Justice Department has charged over 1000 defendants with fraud against the EIDL program, and seized fraudulently obtained assets estimated to be worth 1 billion USD (Ngo, 2022). Approximately 200 individuals have been prosecuted for PPP fraud (Solow, 2022). There are also over 240 civil investigations into 1800 individuals and organized crime groups for other forms of misconduct related to COVID-19 relief loans amounting to more than 6 billion USD (Morse, 2021). The punishments meted out so far suggest that federal government officials have followed through on their promises to hold white-collar criminals accountable. For example, three men who applied for 2.7 million USD in fraudulent PPP and EIDL loans received sentences ranging from 60 to 72 months. Each of the men was ordered to pay about 500,000 USD in restitution (United States Department of Justice, 2021).

Trump’s animus toward China

Although voluminous research exists explaining Trump’s 2016 surprise election, a common theme is that he drew much support by emphasizing white identity and outgroup hostility toward Blacks, immigrants, and Muslims (Fording & Schram, 2020; Graham et al., 2021; Jardina, 2019). Direct anti-Asian rhetoric from Trump was relatively sparse. When asked how his administration would combat hate crimes against Asian Americans, Trump responded: “Well, I don’t know. All I know is this: Asian Americans in our country are doing fantastically well” (Kai, 2021). Trump’s animus toward China, however, is another story.

During the 2016 presidential campaign, Trump argued that China was engaging in unfair trade practices, leading to the loss of manufacturing jobs, a massive trade imbalance, and theft of intellectual property. He called these practices “the greatest theft in the history of the world” and argued that “We can’t continue to allow China to rape our country” (Hass & Denmark, 2020). Upon taking office, Trump launched a trade war and kept up his rhetoric. Analyses have shown that this policy cost nearly 300,000 jobs in the USA, shrunk the GDP, and cost companies nearly 2 trillion USD (Hass & Denmark, 2020). A trade deal was consummated in January of 2020.

Around this time, the pandemic began to unfold. Despite being aware of the health risks, Trump publicly downplayed the dangers of COVID-19. As infections and deaths mounted, his denials rang hollow (Bump, 2020; Cullen et al., 2021). Part of his response was to blame China for the origins of the disease. In a March 16, 2020 tweet, Trump substituted the term “China Virus” for the “COVID-19 Virus” (Hswen et al., 2021). He later used the terms “Kung Flu” and “Wuhan Flu” because they were popular in conservative circles. Trump and his allies claimed that these labels were not racist, but merely a way COVID-19 originated (Mangan, 2020; Rogers et al., 2021).

An analysis of 1.2 million hashtags showed that anti-Asian sentiments accompanied one-fifth of the #covid-19 hashtags and one-half of #chinesevirus hashtags (Hswen et al., 2021). Commentators now link the rise in hate crimes against Asian Americans to the toxic environment created by Trump (Kai, 2021; Tavernise & Oppel, 2020). A 2021 study by the Pew Research Center reported that one in five Asian Americans blamed Donald Trump for increases in violence against their group (Ruiz et al., 2021). In the same survey, Asian Americans reported that they feared a physical attack (32%), perceived that others acted uncomfortable around them (27%), were subject to racial slurs (27%), were told to go back to their country (16%), and were blamed for the coronavirus outbreak (14%). Another survey conducted during the early stages of the COVID-19 pandemic found that anti-Asian attitudes were associated with increased support for racially motivated policies, including reducing the number of resident alien visas from Asian countries (Reny & Barreto, 2022).

These considerations suggest that those with faith in Trump not only are hostile to Blacks, Muslims, and immigrants from south of the border, but also have been directed by this rhetoric to be biased against anyone of Chinese origin—regardless of whether they were born in China or in the USA.1 If so, then those with an allegiance to the former president will likely impose harsher punishments when a perpetrator is identified as a Chinese American.

Current focus

Using factorial vignette methodology and data from a survey that was administered to a national sample of adults, the objective of this study was twofold. This first aim was to test the direct effect of faith in Trump on a series of punishment-oriented outcomes for a specific type of white-collar crime—bank fraud—that was widespread during the COVID-19 pandemic. Although factors that are associated with faith in Trump, such as political ideology and party affiliation, have been shown to predict punitive attitudes when it comes to street crimes, such relationships are less pronounced when it comes to white-collar offenses (Simpson et al., 2022; Unnever et al., 2008). Accordingly, this portion of the study was largely exploratory. The second goal of the study was to test whether the race of the person—Chinese American versus White—who committed the bank fraud influenced the willingness to punish among those with high levels of faith in Trump. Because the race manipulation was expected to trigger discriminatory beliefs about outgroup members among Trump loyalists, this portion of the study focused on interaction effects. To ensure robust parameter estimates, a number of statistical control variables were included that have been shown previously to explain punitive attitudes, including political ideology, party affiliation, national identity, and demographic characteristics.

Methods

Participants

The data for this study come from a nationally matched (i.e., age, race, and sex) opt-in sample of Americans aged 18 years and older. The survey was administered from July 19 to July 22, 2021, by Prolific Academic Ltd. (Oxford, UK) to a sample of 1509 participants. The age of participants ranged from 18 to 92 years (median = 43, mean = 44.4, SD = 16.08). The sample was evenly split between sexes (50.6% female and 49.4% male). In terms of race, the majority of the sample was White (75.4%), 13.4% were Black or African-American, 6.4% were Asian, 0.4% were American Indian or Alaskan Native, 2.7% identified as multiracial, and 1.8% were “other.” Nearly 6% of respondents identified as Hispanic, Latino, or Spanish origin. About half of the sample was married (49.4%) at the time of the survey, 25.3% had never married, 12.1% were not married but in long-term relationships, 10.6% were divorced, and 2.6% were widowed. A majority of participants (56.1%) owned their home. The median amount of time individuals had lived in their current home was 6 years (mean = 10.5, SD = 10.70). The sample was heterogeneous in terms of political ideology: 14% considered themselves extremely liberal, 41.2% were either liberal or slightly liberal, 14% were centrists, 25.4% were either slightly conservative or conservative, and 5.2% said they were extremely conservative.

To determine how well the sample represented the population of adults in the USA, demographic information was compared to 2021 estimates from the U.S. Census Bureau (see Supplemental Appendix Table S1). The sample appeared representative in terms of sex and race. The sample also reflected the adult population when it came to civilian labor force participation. The sample underrepresented Latinos, individuals who were born outside the USA, and homeowners. Finally, the sample overrepresented the percentage of adults who had resided in their home for at least 1 year and those who had earned a Bachelor’s degree or higher.

Procedure

Prior to beginning the survey, participants were informed that their participation was voluntary and that their responses were anonymous. Participants were paid the equivalent of $9.32 per hour. The protocols used in this study were approved by the university’s institutional review board prior to data collection. The survey contained a variety of open- and closed-ended items that queried respondents about a variety of criminal justice issues, including police fairness, self-reported compliance with laws, and the morality of laws. Survey items for three composite scales were also included in the front portion of the survey. These items captured attitudes toward political matters and leaders that were of interest in this study. Following these items, a factorial vignette was embedded in the survey instrument. The vignette included experimental conditions that were central to this study. The hypothetical scenario was followed by a series of closed-ended items; some were used as outcome measures and others used to evaluate the quality of the vignette. Finally, the survey concluded by asking a series of questions about respondents’ personal characteristics. After using listwise deletion to remove cases with missing information, the analytic sample was reduced by approximately 6%.2

The vignette that was administered to participants described an incident involving a specific type of white-collar crime (i.e., bank fraud; see Table S2) where a 41-year old man (the average age for a white-collar offender, see Piquero & Piquero, 2016) received more than 1 million USD after applying for federally insured loans to save his nonexistent struggling business. Participants were also told what the maximum penalty for this level of bank fraud was. The vignette used a 2 (race of the individual who committed bank fraud) × 2 (prior criminal record) × 2 (COVID-19 related fraud) between-subject experimental design. Three conditions were randomly administered to participants. The first condition concerned the name and race of the fraudster: “Jeffrey Miller, a 41-year old white man” (control condition) and “Jian Zhang, a 41-year old Chinese American man” (experimental condition). A second condition concerned whether the criminally involved man in the scenario has a criminal history: “no prior criminal record” (control condition) and “a prior criminal record” (experimental condition). The third condition captured the type of bank fraud that was perpetrated. One form of fraud involved obtaining loans meant for businesses struggling with “bad credit” (control condition) and the other type involved loans to help businesses hurt by “COVID-19” (experimental condition).3 To test the effect of the experimental conditions in a multivariate context, a binary coding scheme was used for each of the three experimental conditions—Chinese American, criminal history, and COVID-19 fraud.

Measures

Three composite indices were included as independent variables. Faith in Trump was a summated scale that captured respondents’ allegiance to former-President Donald Trump (Graham et al., 2020). This scale was constructed using five survey items (e.g., “Former President Trump is the only politician who really cares about the common man” and “I love former President Trump’s style because he is strong and tells it like it is”). The second additive index, American identity, captured a sense of belonging and connection to the USA (Schwartz et al., 2012; Wolfe & McLean, 2021). This scale was composed of six closed-ended items (e.g., “I am happy that I am an American” and “I have a lot of pride in the United States”). The third summated scale, immigrant threat, was designed to measure variation in the perceived threat of immigrants to citizens’ economic and material interests (Paxton & Mughan, 2006). This scale was constructed using five survey items (e.g., “Immigrants take jobs from Americans” and “Immigrants increase crime”). Each of the 16 survey items that represented the three scales featured a closed-ended response set ranging from strongly disagree (coded 1) to strongly agree (coded 4). The internal consistency associated with each of the three scales was high (Cronbach’s alpha > 0.90), and the results from the promax-rotated principal-axis factor model demonstrated that the survey items load on the hypothesized latent constructs (see Table 1).

Table 1

Principal axis factor model with promax rotation

		Factor 1	Factor 2	Factor 3
		Loadings	Loadings	Loadings
Faith in Trump (α = .962)
1	I believe that former President Trump will make America great again	.001	.947	-.016
2	Former President Trump is 100% correct that we need a wall to make sure that gang members, criminals, and rapists do not come into the U.S.	.057	.773	.126
3	Former President Trump is the only politician who really cares about the common man	− .051	.883	.047
4	Former President Trump knows how to protect America against threats from around the world	.036	.930	-.015
5	I love former President Trump’s style because he is strong and tells it like it is	− .016	.942	-.010
American identity (α = .942)
1	I have a clear sense of the United States and what being American means for me	.837	− .010	-.056
2	I am happy that I am an American	.888	− .019	.018
3	I have a strong sense of belonging to the United States	.895	− .031	.005
4	I understand pretty well what being American means to me	.797	.011	-.062
5	I have a lot of pride in the United States	.818	.056	.058
6	I feel good about being an American	.895	.008	.033
Immigrant threat (α = .928)
1	Immigrants take jobs from Americans	− .016	.028	.822
2	Immigrants lower the quality of education in schools	− .052	− .017	.869
3	Immigrants bring disease into the country	− .009	.025	.833
4	Immigrants raise taxes for Americans	.024	.059	.802
5	Immigrants increase crime	.038	− .004	.855
Eigenvalue =		8.507	2.557	1.167
% Variance =		53.171	15.982	7.296

Pattern loading estimates greater than |.400| are shown in boldface

Five criterion variables were used in this study. Perceived wrongfulness was a single item measure (“Please indicate ‘how wrong’ you believe the crime committed by the person in the scenario was”). Perceived harmfulness was also a single item variable (“Please indicate ‘how harmful to society’ you think the crime described in the scenario was”). Each of these items featured a four-point response set that was coded so that higher scores indicated more negative assessments of the hypothetical bank fraud. A third variable, prison sentence, reflected whether participants believed that the “person described in the scenario should go to prison” (1 = yes, 0 = no). A follow-up item, termed length of prison sentence, was administered only to those who felt prison was necessary. These participants were asked what the maximum number of years the person in the scenario should spend in prison. Consistent with federal sentencing guidelines, options ranged from 1 to 30 years. The original distribution of scores was widely dispersed (mean = 13.06, SD = 8.80, variance = 77.41). Transforming the distribution by taking the natural log reduced the dispersion of scores (mean = 2.30, SD = 0.79, variance = 0.63). Finally, one extra-constitutional sanction was included. Support for deportation was a single item variable (“The person described in the scenario should lose their citizenship and be permanently removed from the United States”). This item was coupled with a four-point response set that ranged from strongly disagree (coded 1) to strongly agree (coded 4).

To better discern the unique effects of the key independent variables and the guard against potential spuriousness, demographic and politically oriented measures were included as statistical controls. The demographic measures included age (in years), male (1 = yes, 0 = otherwise), White (1 = yes, 0 = otherwise), married (1 = yes, 0 = otherwise), and socioeconomic status.4 The two political variables were conservative ideology (1 = very liberal to 7 = very conservative) and Republican identification (1 = yes, 0 = otherwise). Summary statistics for the variables used in the study are provided in Table 2.

Table 2

Descriptive statistics

	Mean	SD	Minimum	Maximum	N
Outcome Measures
Perceived Wrongfulness	3.74	.52	1	4	1507
Perceived Harmfulness	3.35	.73	1	4	1503
Prison Sentence	.90	–-	0	1	1507
Length of Prison Sentence	2.30	.79	0	3.40	1346
Support for Deportation	2.13	.89	1	4	1508
Experimental Conditions
Chinese American	.50	–-	0	1	1509
COVID-19 Fraud	.50	–-	0	1	1509
Criminal History	.50	–-	0	1	1509
Independent Variables
Faith in Trump	9.23	4.93	5	20	1494
American Identity	18.78	4.37	6	24	1496
Immigrant Threat	9.85	3.92	5	20	1484
Control Variables
Age	44.40	16.08	18	92	1509
Male	.49	–-	0	1	1495
White	.75	–-	0	1	1509
Socioeconomic Status	.00	1.00	-2.11	1.45	1496
Married	.49	–-	0	1	1507
Conservative Ideology	3.46	1.86	1	7	1505
Republican Identification	.20	–-	0	1	1506

Descriptive statistics for Length of Prison Sentence include only cases endorsing a prison sentence for the criminally involved individual described in the hypothetical vignette. The distribution for Length of Prison Sentence was transformed using the natural log

Analytic strategy

Two multivariate data-analytic techniques were used to accomplish the research objectives. The main effects for three dependent variables—perceived wrongfulness, perceived harmfulness, and support for deportation—were assessed using ordinary least-squares (OLS) regression. A two-stage Heckman model was used to estimate the main effects of key independent variables on both the prison sentence and length of prison sentence. The Heckman model was a reasonable choice because the modeling process involved drawing a nonrandom subsample (i.e., participants who recommended that the person committing fraud in the scenario serve a prison sentence) from the full sample. As such, sample selection bias was a concern. The Heckman model can correct for potential bias by calculating an “inverse Mills ratio” (λ) at the first stage, and then including the variable at stage two (see Bushway et al., 2007). To do so, at least two exclusionary restrictions must be included in the stage one model. The first exclusionary restriction, criminal involvement, was a frequency scale that captured both minor and serious forms of crime (Sweeten, 2012). The second such variable, bounded authority, reflected participants’ judgements of how frequently the police exceed their authority in practice (Trinkner et al., 2018).5 Importantly, both of these variables were significantly correlated with the selection variable, but neither had an observable association with the stage-two outcome. Finally, interaction terms were constructed to test whether the race of the man committing fraud in the scenario differentially influenced the outcomes among participants scoring higher on the faith in Trump scale. The latter measure was mean centered prior to creating the interaction term to reduce levels of harmful collinearity (Aiken & West, 1991).6 To ensure that the findings were robust, the regression models were re-estimated using alternative multivariate models.

Results

The first series of multivariate regression models assessed the effects of the experimental stimuli and other independent variables on the perceived seriousness of the white-collar offense depicted in the vignette (see Table 3). Importantly, the F statistic in all of the models was statistically significant, which indicated that the amount of variance explained was not due to chance alone. On the left side of Table 3, perceived wrongfulness was regressed onto the independent variables. The findings from model 1, which explained 9.2% of the variance, indicated that older participants, females, and minorities perceived bank fraud to be more wrong. The test statistics for two of the additive scales were significant at the 0.05 level. More formally, a one-unit increase in the American identity scale corresponded to a 0.018 increase in perceived wrongfulness. In contrast, immigrant threat was negatively correlated with wrongfulness judgments. Variation in the faith in Trump scale was not associated with the dependent variable. One of the experimental stimuli exerted a significant effect. Specifically, individuals who were administered the vignette depicting a Chinese-American man perceived bank fraud as being significantly less wrong. The interaction term was included in model 2 (R2 = 0.095). While the faith in Trump scale was significant in model 2, variables used to construct interaction terms no longer represent main effects when interactions are also included in the model (Aiken & West, 1991, p. 32). Most importantly, the positive interaction effect indicated that the perceived wrongfulness of bank fraud increased as individual faith in Trump also increased when the person committing fraud in the scenario was depicted as Chinese American.

Table 3

Ordinary least-squares regression models for Perceived Wrongfulness and Perceived Harmfulness

	Perceived Wrongfulness				Perceived Harmfulness
	Model 1		Model 2		Model 1		Model 2
	b	t ratio	b	t ratio	b	t ratio	b	t ratio
	(s.e.)		(s.e.)		(s.e.)		(s.e.)
Chinese American	− .060	− 2.343*	− .060	− 2.338**	− .032	− .840	− .032	− .830
	(.026)		(.026)		(.038)		(.038)
COVID-19 fraud	− .013	− .493	− .012	− .472	.026	.679	.027	.708
	(.026)		(.026)		(.038)		(.038)
Criminal history	.034	1.319	.034	1.312	.096	2.516*	.095	2.509*
	(.026)		(.026)		(.038)		(.038)
Faith in Trump	− .006	− 1.404	− .011	− 2.278*	− .005	− .809	− .014	− 1.956
	(.004)		(.005)		(.007)		(.007)
American identity	.018	5.032**	.018	4.962**	.028	5.223**	.027	5.143**
	(.004)		(.004)		(.005)		(.005)
Immigrant threat	− .011	− 2.378*	− .011	− 2.345*	.002	.273	.002	.318
	(.005)		(.005)		(.007)		(.007)
Faith in Trump × Chinese American	–-	–-	.011	2.185*	–-	–-	.020	2.603**
	–-		(.005)		–-		(.008)
Age	.007	8.132**	.007	8.166**	.005	4.126**	.005	4.164**
	(.001)		(.001)		(.001)		(.001)
Male	− .053	− 1.999*	− .054	− 2.031*	.006	.149	.004	.113
	(.027)		(.027)		(.039)		(.039)
White	− .076	− 2.439*	− .075	− 2.421*	− .089	− 1.935	− .088	− 1.913
	(.031)		(.031)		(.046)		(.046)
Socioeconomic status	− .018	− 1.237	− .018	− 1.210	.035	1.622	.036	1.661
	(.015)		(.015)		(.022)		(.022)
Married	.011	.385	.011	.361	.052	1.186	.051	1.156
	(.030)		(.030)		(.044)		(.044)
Conservative ideology	.004	.402	.004	.380	− .018	− 1.146	− .018	− 1.177
	(.010)		(.010)		(.015)		(.015)
Republican identification	.070	1.628	.068	1.588	− .023	− .360	− .026	− .409
	(.043)		(.043)		(.064)		(.064)
Constant	3.325	46.326**	3.377	44.718**	2.676	25.191**	2.767	24.776**
	(.072)		(.076)		(.106)		(.112)
R2	.092		.095		.055		.060
F	10.992**		10.575**		6.367**		6.420**
N	1428		1428		1423		1423

Entries are unstandardized partial regression coefficients (b), standard errors (s.e.), and test statistics (t ratio). *p < .05; **p < .01 (two-tailed test)

An identical modeling strategy was followed for perceived harmfulness (see right side of Table 3). Once again, age and American identity were positively correlated with the dependent variable in model 1 (R2 = 0.055). With regard to experimental stimuli, individuals who received the scenario involving a fraud offender with a criminal history judged the situation to be more harmful. Turning to model 2 (R2 = 0.060), the interaction for faith in Trump and Chinese American was statistically significant. More specifically, individuals who expressed greater faith in former-President Trump who received the scenario depicting a Chinese-American man committing bank fraud reported that the criminal violation was more harmful to society when compared to Trump loyalists who judged the harm of bank fraud committed by a White man.

The regression models in Table 4 focused on prison-related outcomes. Participants were asked whether the individual who committed bank fraud in the vignette should be sent to prison. And, if so, for how many years? Heckman two-stage models with sample selection were estimated. Model 1 provided the main effects. As expected, both of the exclusionary restrictions in the selection model were significant: participants with more extensive criminal involvement and those who were more critical of police authority were both less likely to recommend a prison sentence. The inverse Mills ratio (not shown in table) was not statistically significant (λ = 0.416, z = 1.71, p = 0.087), which indicated the absence of harmful sample selection bias. Nevertheless, this two-step modeling strategy was used because it is more efficient when compared to estimating separate equations (Bushway et al., 2007). Older participants were more likely to believe a prison sentence was appropriate as were individuals who were administered the criminal history experimental condition. In contrast, those who received the Chinese-American stimulus and participants who scored higher on the faith in Trump scale were significantly less likely to recommend prison. This model specification was replicated in model 2, but with the inclusion of the interaction term. Importantly, the parameter estimate and corresponding significant test statistic for the interaction term indicated that participants who scored higher on the faith in Trump scale were more likely to recommend a prison sentence when the person committing bank fraud in the scenario was Chinese American.

Table 4

Heckman two-stage regression models with sample selection for Prison Sentence and Length of Prison Sentence

	Model 1				Model 2
	Prison Sentence		Length of Prison Sentence		Prison Sentence		Length of Prison Sentence
	b	z test	b	z test	b	z test	b	z test
	(s.e.)		(s.e.)		(s.e.)		(s.e.)
Chinese American	− .294	− 2.51*	− .071	− 1.50	− .261	− 2.60**	− .070	− 1.48
	(.099)		(.047)		(.100)		(.047)
COVID-19 fraud	− .176	− 1.78	− 041	− .91	− .172	− 1.74	− .040	− .88
	(.099)		(.046)		(.099)		(.045)
Criminal history	.218	2.19*	.057	1.23	.212	2.13*	.059	1.28
	(.099)		(.046)		(.100)		(.046)
Faith in Trump	− .034	− 2.00*	.000	.05	− .055	− 2.80**	− .016	− 1.68
	(.017)		(.008)		(.020)		(.009)
American identity	.006	.45	.030	4.55**	.005	.35	.029	4.49**
	(.014)		(.007)		(.014)		(.006)
Immigrant threat	.019	1.05	.014	1.82	.020	1.09	.015	1.89
	(.018)		(.008)		(.018)		(.008)
Faith in Trump × Chinese American	–-	–-	–-	–-	.042	2.10*	.036	3.86**
	–-		–-		(.020)		(.009)
Age	.011	3.01**	− .004	− 2.22*	.011	3.07**	− .004	− 2.14*
	(.004)		(.002)		(.004)		(.002)
Male	− .114	− 1.10	− .024	− .52	− .117	− 1.12	− .027	− .58
	(.104)		(.047)		(.104)		(.046)
White	− .036	− .32	− .042	− .78	− .040	− .34	− .040	− .73
	(.115)		(.054)		(.115)		(.054)
Socioeconomic status	.003	.04	− .053	− 2.11*	.004	.06	− .051	− 2.04*
	(.058)		(.025)		(.059)		(.025)
Married	− .220	− 1.82	.134	2.57*	− .231	− 1.91	.132	2.53*
	(.121)		(.052)		(.121)		(.052)
Conservative ideology	.026	.70	− .026	− 1.36	.024	.65	− .026	− 1.40
	(.037)		(.019)		(.037)		(.019)
Republican identification	.048	.28	.054	.72	.050	.29	.044	.59
	(.170)		(.076)		(.170)		(.076)
Criminal involvement	− .048	− 2.52*	–-	–-	− .047	− 2.46*	–-	–-
	(.019)		–-		(.019)		–-
Bounded authority	− .050	− 4.90**	–-	–-	− .051	− 4.97**	–-	–-
	(.010)		–-		(.010)		–-
Constant	2.49	6.18**	1.78	11.37**	2.724	6.47**	1.930	12.48**
	(.402)		(.157)		(.421)		(.155)
Wald χ2	56.00**				71.12**
N	1404		1265		1404		1265

Entries are unstandardized partial regression coefficients (b), standard errors (s.e.), and test statistics (z test). *p < .05; **p < .01 (two-tailed test)

Turning attention to the outcome equations (i.e., length of prison sentence) in Table 4, the main effects model (located second from the left) showed that younger participants, lower SES individuals, and those who were married recommended longer prison sentences. More notably, the z test for the interaction term was statistically significant in the remaining outcome equation (located on the far right). Simply put, participants who scored higher on the faith in Trump scale said that the person committing bank fraud in the scenario should receive a longer prison sentence when they were depicted as Chinese American.

The support for deportation models are featured in Table 5. Model 1 explained approximately 21% of the variance in the dependent variable. The test statistics in model 1 indicated that participants who received the Chinese-American and criminal history experimental conditions reported significantly higher levels of support for this extra-constitutional sanction. Higher levels of support for deportation were also observed among participants with higher levels of faith in Trump, individuals who reported more negative feelings toward immigrants, and nonwhite participants. Most interestingly, the interaction term in model 2 showed that participants with higher faith in Trump scores expressed greater levels of support for deporting the bank fraud perpetrator in the hypothetical scenario if they received the Chinese-American experimental condition.7 It is worth noting at this point that the effect of the interaction term (i.e., faith in Trump × Chinese American) was largely consistent across five dependent variables.8 The takeaway was straightforward: for Trump supporters, the extra-legal factor of Chinese ancestry mattered when judging the seriousness and the appropriate sanction for bank fraud.

Table 5

Ordinary least-squares regression models for Support for Deportation

	Support for Deportation
	Model 1		Model 2
	b	t ratio	b	t ratio
	(s.e.)		(s.e.)
Chinese American	.605	14.445**	.605	14.540**
	(.042)		(.042)
COVID-19 fraud	.032	.766	.034	.811
	(.042)		(.042)
Criminal history	.127	3.037**	.126	3.036**
	(.042)		(.042)
Faith in Trump	.014	2.009*	− .001	− .169
	(.007)		(.008)
American identity	.009	1.527	.008	1.396
	(.006)		(.006)
Immigrant threat	.053	7.204**	.054	7.310**
	(.007)		(.007)
Faith in Trump × Chinese American	–-	–-	.034	4.083**
	–-		(.008)
Age	− .001	− .847	− .001	− .806
	(.001)		(.001)
Male	− .072	− 1.670	− .075	− 1.735
	(.043)		(.043)
White	− .117	− 2.318*	− .115	− 2.291*
	(.051)		(.050)
Socioeconomic status	.033	1.371	.034	1.432
	(.024)		(.024)
Married	.031	.651	.029	.609
	(.048)		(.048)
Conservative ideology	.007	.389	.006	.349
	(.017)		(.017)
Republican identification	− .112	− 1.598	− .118	− 1.686
	(.070)		(.070)
Constant	1.085	9.287**	1.243	10.149**
	(.117)		(.122)
R2	.210		.219
F	28.908**		28.331**
N	1428		1428

Entries are unstandardized partial regression coefficients (b), standard errors (s.e.), and test statistics (t ratio). *p < .05; **p < .01 (two-tailed test)

Visual depictions of the interaction effects are provided in Fig. 1. The scatter plots featured in the five panels illustrate the differential influence of faith in Trump on the punishment-oriented outcomes for those who received the experimental condition (Chinese American) and for individuals who were administered the control condition (White). The plots show that the relationship between faith in Trump and the dependent variables was positive among those individuals who received the Chinese-American condition. With but one exception, the relationship between faith in Trump and the punishment-oriented outcomes was either negative or null among participants who received the control condition. The sole exception to this pattern concerned the support for deportation, where the effects for faith in Trump were positive for both conditions. However, the effect among those receiving the Chinese-American condition was much stronger—as indicated by the steeper slope—relative to the White condition.9 Regardless of this exception, the results were clear: individuals who expressed greater faith in Trump were more critical of the bank fraud—perceived it as more wrong and harmful—and were more willing to punish the person—send them to prison, sentence them to longer prison terms, and even deport them from the USA—when he was depicted as Chinese American.

Fig. 1

Interactions between faith in Trump and race manipulation on outcome measures. A Perceived Wrongfulness. B Perceived Harmfulness. C Prison Sentence. D Length of Prison Sentence. E Support for Deportation

Discussion

This study explored public views about the punishment of white-collar crime, with a focus on whether bank fraud exploiting COVID-19 government support was judged differently than a standard fraud offense. The experiment introduced a control for prior criminal history and added the innovation of varying the race of the fraud perpetrator—White versus Chinese American. The focus on a Chinese American was relevant because of accusations that COVID-19 originated in China, and was also important given that much of the prior research on public perceptions suggests a longstanding association with white-collar crime and White perpetrators. The analysis also included a measure of faith in Trump to assess whether endorsement of the former president shaped preferences. The analysis examined five outcomes to assess the generality of effects across views associated with punishment. Three key findings emerged.

First, a COVID-19 fraud did not exert a significant effect on any outcome. It appears that the perceived wrongfulness and harmfulness of the act, as well as preferred sanctions, were not affected by whether the opportunity for misconduct was linked to the onset of COVID-19. Rather, individuals who were told that the perpetrator had a criminal history viewed the scenario as more harmful. Notably, the public displayed a willingness to punish the commission of fraud. For example, 90% of the sample favored sentencing the perpetrator to a prison sentence, with a mean term of 13.06 years. These findings belie the claim, once popular in the literature, that the public is unconcerned about white-collar crime and does not support stringent sanctions (Cullen et al., 2009; Holtfreter et al., 2008). The results add to the growing body of research that shows public perceptions about the seriousness of white-collar crime are linked to case characteristics and to individual differences (Simpson et al., 2022). These views also align with increased governmental enforcement and prosecution of white-collar crime.10

Second, somewhat surprisingly, the Chinese-American stimulus was negatively related to perceived wrongfulness and to both measures of support for imprisonment. This status had no significant effects on the perceived harmfulness of the act. It is possible that being a “minority” was seen as a disadvantage that lessened perceptions of culpability and deserved sanctions. Similarly, pre-existing stereotypes of Asian Americans as “model minorities” who have achieved significant educational and economic success may have influenced perceptions (Kim, 2008; Ng et al., 2007). Evaluations of harm, however, were distinct because the substantive effects were an empirical fact. A notable exception was with support for deportation—an unconstitutional penalty that no perpetrator could actually receive in these situations due to their American citizenship. Here, being Chinese American greatly increased support for deportation, and a “Trump effect” was apparent, which will be addressed shortly. Also relevant, belief that immigrants posed a threat significantly increased the willingness to deport, a finding that is consistent with the broader literature on out-group threat and punitive attitudes toward crime (Trahan & Pierce, 2022).

Third, a central aim of this study was to assess the extent to which faith in Trump affected the punishment attitudes of the sample. Such a finding would suggest that belief in certain political leaders may shape sanctioning policy preferences. The implication for future research would be salient because views of leaders are typically not included in surveys. Of course, the effects of Donald Trump could be unique (Barak, 2022). Still, the ways in which political and other elites shape policy preferences is an empirical question (Beckett, 1997). The role of the media in perpetuating stereotypes and influencing views on crime and public policy also warrants research attention (Ramasubramanian, 2011; Weitzer & Kubrin, 2004).

In the current context, it is instructive that two divergent findings regarding the faith in Trump variable emerged in the analysis. In Table 4 (see model 1), note that faith in Trump is negatively related to support for a prison sentence. As suggested, one reason for this finding might be that Trump is a corporate icon who has been accused of business fraud, which he rejects as unfair and a witch hunt (see, e.g., Cohen, 2019). An allegiance to him might diminish the willingness to punish white-collar crime. Whether faith in Trump exerts a similar influence on punishment perceptions when it comes to other crime contexts (e.g., violence or property offenses) remains an open empirical question that should be addressed in future public opinion studies. Along those lines, a recent exploratory study reported that both targets and victims of consumer fraud scored higher on their faith in Trump (Kennedy et al., 2021). This raises the question of whether it is possible for faith in a political leader to be truly “blind” to the point of obscuring judgment about the potential risks of victimization, or in other situations. Indeed, as noted at the outset, there is considerable evidence of Trump’s capacity to influence his supporters; many still believe that he won the 2020 election despite authority figures and established institutions declaring otherwise. As further documentation, numerous individuals charged in the January 6 capitol insurgency cited Trump’s promises that their actions would be pardoned as justification for engaging in violence (Pengally, 2022).

The other finding was the significant and positive effect that the interaction term (i.e., faith in Trump × Chinese American) had on the dependent variables. In these instances, those endorsing Trump apparently were activated to express more punitive evaluations of the Chinese-American man committing bank fraud when compared to the White man. The act of this individual was seen as more wrong and harmful. The act was also viewed as more deserving of imprisonment and for longer durations, and as more meritorious of deportation—all of which is contrary to the democratic value of equal treatment under the law. This finding is consistent with the connection between Whites’ racial attitudes and support for Trump (Fording & Schram, 2020; Graham et al., 2021). Trump’s “zero-tolerance” immigration policy, his repeated linking of immigration and crime, and public calls for undocumented immigrants to be “sent back from where they came from” may have similarly influenced the inaccurate perception that deportation is a valid legal sanction for US citizens (Frazee, 2018). Racial animus toward Chinese Americans may stem from a variety of sources apart from Trump, including existing prejudices about Asian Americans more generally, such as the negative stereotype that they are threatening outsiders who will never be viewed as “true Americans” (Matsuoka & Junn, 2013). The broader point is that allegiance to the former president likely increases support for policies that target out-groups for greater social control.

An additional finding that merits discussion concerns the relevance of identity. The measure of having an American identity increased perceptions of wrongfulness and harmfulness and support for lengthier prison terms. Previously, Wolfe and McLean (2021) found that American identity was positively associated with belief in police legitimacy. It is possible that a strong identity as an American encourages support of the nation’s governmental institutions and subsequently serves to endorse their use of crime control policies. Regardless, the analysis indicates that understanding punishment preferences needs to move beyond the study of partisanship and ideology and consider a broader range of political attitudes, such as American identity and allegiance to political leaders, which heretofore have been under researched.

Prior to drawing any conclusions, two limitations of the study should be discussed in more detail. The first limitation concerns the sample—a national opt-in online sample matched on age, race, and sex—that was used. There are several benefits to this type of sampling strategy. For example, it is relatively inexpensive and data collection takes place over a short period of time. But there are also costs. Perhaps most concerning is that the sample is not entirely representative of the national population of adults. As was noted, the sample underrepresented homeowners, foreign-born individuals, and Latinos. It overrepresented college-educated individuals and those who had resided in their home for at least 1 year. Given these differences, care should be exercised when generalizing the research findings to the adult population in the USA until future research that uses national probability sampling replicates the results. The second limitation concerns measurement. Specifically, the five dependent variables used in this study were measured using single survey items. While these items certainly possess face validity, it is not possible to subject these variables to more robust scrutiny. Future research that addresses similar questions that use multi-item scales will help determine the extent of this potential limitation.

Public opinion undoubtedly plays a role in driving punitive crime control policies (Enns, 2016; Pickett, 2019). Criminological research has demonstrated that a variety of factors shape perceptions about both the seriousness of white-collar crime and beliefs about how it should be punished. Faith in Donald Trump in particular and allegiance to political leaders more broadly is one of the key factors that has largely been ignored. Elected officials can and do have the power to shape public opinion, a point that must be considered not just in future research on punishment perceptions, but also when it comes to efforts to educate the public about criminal justice policy and for broader societal concerns. In autocratic nations, dictators use institutions like the criminal justice system as a mechanism for controlling dissidents (Braithwaite, 2017). Political elites frequently manipulate public opinion for their own advantage, but the potential for harnessing support and using available technology (e.g., social media) to control information and exploit the citizenry for tyrannical purposes also exists (Kendall-Taylor et al., 2020; Keremoğlu & Weidmann, 2020). Moving forward, public awareness campaigns should continue to educate individuals about the factors and circumstances that increase risks of victimization, and would also be well-served by greater efforts to combat disinformation about the causes of crime and about the potential available sanctions that fall within the law.

Below is the link to the electronic supplementary material.

Supplementary file1 (DOCX 36.6 KB)