This is an open access article distributed under the terms of the Creative Commons
Attribution License (Recently, undocumented immigration1 has become a highly moralized and a politically changed topic in American politics. One of the most embattled controversies revolves around the fate of the 11.1 million undocumented immigrants in the United States, many of whom have lived there for decades (Felter, Renwick, & Cheatham, 2022). The controversy is fueled by the particularly negative attitudes people have toward undocumented (compared to documented) immigrants (Murray & Marx, 2013). Animosity could be explained by moralization. The presumed unlawful entry by undocumented immigrants into a country is perceived as harmful (Nittle, 2021).
Harming vs. Helping
The policy debate emphasizes different sets of approaches. Therefore, it is critical to distinguish between policies that harm and help undocumented immigrants. This cannot be captured with a unidimensional measure. Past research has typically focused on antecedents of anti-immigration perspectives (Louis, Esses, & Lalonde, 2013; Stephan et al., 1998), but little research has examined predictors of pro-immigration policies. People may be compelled to avoid harming undocumented immigrants, but that does not mean they will be motivated to offer them help either. To address this gap in the research, I examined support for helping and punishing undocumented immigrants.
A Theory of Dyadic Morality Perspective on Undocumented Immigration
Moral psychology provides insights that may be useful to better understand people’s policy positions toward undocumented immigrants. Moral behavior is perceived in terms of actors who commits a good or bad act (agents) and actors who are impacted by this behavior (patients; i.e., Theory of Dyadic Morality, TDM; Schein & Gray, 2018). Moral agents are perceived as exercising control over their behavior and being blameworthy for harm they cause (Alicke, 2000). Moral patients are seen as victims who can experience pain (Bernstein, 1998). The TDM argues that agents and patients are generally perceived as mutually exclusive (Gray, Waytz, & Young, 2012; Gray & Wegner, 2009; Gray, Young, & Waytz, 2012).
According to the TDM, agents possess the intention and the ability to cause harm (Schein & Gray, 2018). Many people illustrate undocumented immigration with images of menacing individuals climbing border fences (Nittle, 2021), and this belief about the agency of undocumented immigrants led millions of Americans to vote for Donald J. Trump for president (Felter et al., 2022). Conversely, undocumented immigrants can be perceived as experiencing suffering and lacking control over their circumstances.
Moral typecasts may be valuable in understanding support for different types of policies toward undocumented immigrants. Perceiving them as agents of harm is likely to lead to support for punishing because punishment is a mechanism to curtail the harm. According to the TDM, where there is an agent, there is also a patient. Thus, if people perceive undocumented immigrants as agents, they are contrasting them with citizens who are presumably harmed (Guskin & Wilson, 2017). If viewed as patients, they will be contrasted with the employers or authorities (agents) causing harm to them and may increase helping. Undocumented immigrants often work in physically demanding jobs and are subjected to predatory workplace practices (Spiggle, 2019). Recent debates have also focused on the risk of deporting undocumented immigrants who are fleeing violence and of separating parents and their children (Felter et al., 2022).
The Moral Dyad vs. Integrated Threat Theory
The Integrated Threat Theory (ITT; Stephan & Stephan, 1996; Stephan, Ybarra, Martínez, Schwarzwald, & Tur-Kaspa, 1998) has been applied to immigration attitudes (Scheibner & Morrison, 2009; Stephan et al., 1998). Although the current research does not empirically test the TDM (Gray & Wegner, 2009; Schein & Gray, 2018) vs. the ITT, it is worth comparing them. The ITT defines different types of perceived threats that lead to out-group prejudice. Realistic threats appear to jeopardize one’s safety or wellbeing. Symbolic threats come from infringement on one’s values or identity (Stephan & Stephan, 1996). Anxiety about social interaction and negative stereotypes can also lead to outgroup prejudice (Stephan et al., 1998). The TDM is about perceived harm, immorality, and moral typecasts (Schein & Gray, 2018). Perceived harms and perceived threats overlap conceptually. What is harmful could be viewed as threatening and vice versa. Both harm and threat predict outgroup animosity. However, the ITT does not provide guidance to predict helping, whereas the TDM does. The current research contributes uniquely by examining the effect of victim beliefs on pro-immigrant perspectives.
Most research applying the ITT to immigration examines immigration generally, not undocumented immigration. For example, Stephan et al. (1998) found that all four types of threat predicted attitudes toward Moroccan, Russians, and Ethiopian immigrants in Spain and Israel (without specifying their status). While intergroup anxiety and negative stereotypes were more consistent predictors. Similarly, Scheibner and Morrison (2009) found that realistic, symbolic, and intergroup anxiety were related to evaluations of Polish immigrants in Ireland. However, beliefs about undocumented immigrants specifically are important to examine because their lack of legal status has been moralized (Nittle, 2021). In addition, Tannenbaum, Uhlmann, and Diermeier (2011) found that people who commit one harm are expected to commit other harms. Some argue that undocumented immigrants have already broken the law; therefore, they are more likely to harm Americans (Lopez, 2018).
Mediating Role of Moral Emotions
Each moral dyad elicits a constellation of emotions (Gray & Wegner, 2011) that can help explain support for policies toward undocumented immigrants. For example, agents make observers feel disgusted and angry while victims garner sympathy and sadness. Disgust and anger are moral emotions, and they are felt in response to negative evaluations of someone else’s behavior and involve condemnation of that behavior (Haidt, 2003). Disgust (Hodson et al., 2013) and anger (Haidt, 2003) lead to negative prejudice. Although disgust and anger have different antecedents and outcomes, research has found that they can amplify each other (Salerno & Peter-Hagene, 2013). Perceiving undocumented immigrants as harmful agents may result in feeling disgust and anger, which then may explain greater support for punishment (Hodson et al., 2013).
Conversely, perceiving that someone is suffering is associated with presumed innocence (Wegner, 2010). Empathy is critical for helping behavior (Zaki & Cikara, 2015). Empathy, however, is complex, and includes affective and cognitive elements (Miklikowska, 2018). Viewing undocumented immigrants as victims of harm may elicit feelings of empathy (affect), which is a potential psychological mechanism to explain more helping. Theory on the moral emotions proposes that people cannot feel disgust, anger, and empathy at the same time (Weiner, 1980).
Current Research
This research set out to apply the TDM to views about undocumented immigration and distinguished between punishing and helping. Study 1 collected open-ended data to capture the nature of college students’ beliefs about undocumented immigrants. College students have historically been the demographic group in the U.S. with the lowest voter turnout, but they changed that trend in the 2018 and 2020 elections (Thomas, Gismondi, Gautam, & Brinker, 2020). College students’ involvement in political issues is growing. Undocumented immigration became moralized during the 2016 Trump campaign (Felter et al., 2022; Nittle, 2021). Thus, understanding the perspectives of college students about this specific political issue is increasingly important.
Gray and colleagues’ moral typecasts (Gray & Wegner, 2009, 2011; Gray, Young, & Waytz, 2012) informed the coding of themes describing undocumented immigrants, resulting in two overarching themes. Controlling for political orientation, I hypothesized that descriptions of undocumented immigrants as agents would be indirectly associated with punishment (H1). Conversely, above and beyond political orientation, victim typecasts would be indirectly associated with helping (H2). Gray and Wegner (2011) posit that an agent typecast will elicit disgust and anger while a victim typecast will elicit empathy-related emotions, thus I hypothesized that the link between agent beliefs and punishment would be mediated by disgust/anger (H3) and the link between victim beliefs and helping would be mediated by empathy-related emotions (H4). Varela et al. (2013) found that political conservatism was correlated with negative attitudes toward immigrants. I hypothesized that political orientation would also independently predict people’s policy responses. However, moral typecasting beliefs would predict policy support beyond political orientation.
Study 2 included closed-ended items measuring beliefs about undocumented immigrants as agents or victims that mapped onto the themes that emerged in the qualitative data from Study 1. Mediation models tested the indirect links from the agent and victim beliefs to policy support (punishing or helping) through the emotions (disgust/anger or empathy-related emotions). Study 2 analyses were run controlling for political orientation.
Study 1
Method
Participants and Design
One hundred thirty-six undergraduate students were recruited from the Psychology Department prescreening pool at a public university in New England. All undergraduate students enrolled in a psychology course had the option to complete the prescreening survey during the first two weeks of the semester. I included questions about support for policies that punish and help undocumented immigrants. To ensure a diversity of views, scores on punishment were divided into three groups (low, medium, high based on the distribution) and an equal number of participants were randomly selected for recruitment from each group. Students were emailed in batches until the desired number of participants was reached.2 The response rate of those contacted was 24.78%. The Study 1 sample included an approximately equal number of people from each group based on punishment (i.e., low, n = 49; medium, n = 45; high, n = 45) and based on helping (i.e., low, n = 41; medium, n = 47; high, n = 51). Study 1 participants did not differ on punishment, helping, political orientation, age, ethnicity, or race from those who were invited but did not participate, all ps ≥ .113. However, 13.38% of females and 9.12% of males participated, χ2 = 4.19(1), p = .041. Therefore, there was evidence of response bias for sex. In the analyses, punishment and helping were examined as continuous variables.
Of the 136 participants, 15 did not complete the open-ended measures and were dropped from the analyses. The mean age for the remaining 121 was 19.88 (SD = 2.42); 91 participants identified as female, 29 as male, and 1 did not report sex; 89 participants identified as non-Hispanic and White, 13 as Asian, 7 as multiracial, 6 as Hispanic and non-White, 3 as Hispanic and White, 3 as Black, 1 as indigenous American, and 1 did not report race or ethnicity. The study was a two-part online cross-sectional survey.
Procedure and Measures
Participants completed a prescreening survey during the first two weeks (January 18-February 8, 2011) of the semester that included measures about participants’ support for punishing and helping undocumented immigrants. In an online study completed during the last week of the semester (April 25-27, 2011), participants were asked to provide their own views to an open-ended prompt, which was also used in the email invitation:
“I am a graduate student doing research on illegal immigration. Currently very little is known about how college students view the topic of illegal immigration. We are trying to learn more about undergraduates’ views on illegal immigration in the U.S. Not a lot is known about what people believe and in what ways those beliefs reflect personal experience and or social, political, moral, and economic views.”
I framed my query as exploratory. To encourage participants to think beyond one domain, I mentioned several areas that might reflect participants’ views.
Mixed methods were used to analyze the data. The author completed multiple readings of the participants’ open-ended responses. From the readings of the data, inductive sub-themes about the actions of undocumented immigrants were developed and check by an auditor (Hill, Thompson, & Williams, 1997). The author and auditor agreed that nine sub-themes emerged; five fit with the deductive agent typecast theme and three with the patient typecast theme. The ninth sub-theme about seeking a better life was dropped since it described a motive, not a moral category. Then content coding was conducted by three undergraduate coders for the eight themes (Woike, 2007). Coders read each theme’s definition developed by the author, read each response, and coded them as 1 (present) or 0 (not present) for that theme. Frequencies, interrater reliabilities, and example quotes were reported. The unit of analysis was the full response of each participant, meaning that one response could only receive one code for each theme. Coders completed the coding for each theme one at a time. The final codes used for analyses were determined by the majority (2 of 3) if the coders did not unanimously agree.
Five themes involved perceiving undocumented immigrants as agents of harm, including the beliefs that they freeload by taking undeserved benefits, Light’s Kappa (the mean of three Kappas; Hallgren, 2012), к = .69; poach jobs from Americans, к = .56; do not pay taxes, к = .54; hinder the process for legal immigrants coming to the U.S., к = .38; and threaten the safety of Americans, к = .52. Three themes involved perceiving undocumented immigrants as victims, including the beliefs that they fled bad circumstances, к = .57; are exploited in the U.S., к = .13; and deserve basic human rights they may be denied in the U.S., к = .34. The composite of the agent themes was coded 1 (if any code or multiple codes were present) or 0 (if none of the codes was present), making it a dichotomous variable. The correlation between the agent and victim themes was negative, ϕ = -0.34, p < .001 (see Table 1 for all the correlations).
Table 1
Study 1 Descriptive Statistics and Correlations
| Variable | M | SD | 1 | 2 | 3 | 4 | 5 | 6 | 7 |
|---|---|---|---|---|---|---|---|---|---|
| 1. Agent coding | 0.41 | 0.49 | – | ||||||
| 2. Victim coding | 0.31 | 0.46 | -.34*** | – | |||||
| 3. Disgust/anger | 2.99 | 1.73 | .54*** | -.19* | – | ||||
| 4. Empathy-related emotions | 3.10 | 1.33 | -.17† | .20* | -.07 | – | |||
| 5. Punishment | 3.01 | 1.30 | .54*** | -.26** | .59*** | -.23** | – | ||
| 6. Helping | 2.93 | 1.39 | -.43*** | .23** | -.51*** | .18* | -.59*** | – | |
| 7. Political orientation | 3.36 | 1.19 | .18† | -.15 | .35*** | -.16† | .46*** | -.27** | – |
Note. Of the 85 participants who wrote a response to the open-ended question, 58.82% wrote about agent themes and 43.53% wrote about victim themes.
†p < .10. *p < .05. **p < .01. ***p < .001.
Following the open-ended responses, participants completed items assessing emotions: “When I think about illegal immigrants in the U.S., I feel disgusted, angry, sad, pity, guilty” on a 1 (not at all) to 7 (extremely) scale. Due to the high correlation between disgust and anger, r = .75, p < .001, they were averaged. Sad, pity, and guilty were averaged to measure empathy-related emotions, α = .68. Without guilty, the correlation between sad and pity was, r = 0.43, p < .001 and the reliability was, α = .60. Adding guilty improved the reliability. Guilt and empathy are often both associated with helping or reparations after transgressions (Leunissen, Sedikides, & Wildschut, 2017), and thus, it was reasonable to average sad, pity, and guilty. Analyses examined two emotion composites: disgust/anger and empathy-related emotions.
Policy support was assessed by asking: “Each of the following has been recommended as the appropriate way to deal with illegal immigration in this country.” Then they were asked to rate each policy on a scale from 1 (strongly disapprove) to 7 (strongly approve). Support for punishment was measured with four items (α = .73): “Force illegal immigrants to return to their country of origin, require illegal immigrants to serve time in prison, require illegal immigrants to pay a large fine, and fire illegal immigrants from their jobs (i.e., not allow them to have a job).” Support for helping undocumented immigrants was measured with four items (α = .82): “Allow illegal immigrants to collect welfare, give illegal immigrants the right to vote, permit children of illegal immigrants to receive scholarships or funding to U.S. colleges, and provide free medical care to illegal immigrants.” Items were selected based on popular beliefs not necessarily what policies were the best for undocumented immigrants. Six items were originally measured for both helping and punishing, but two items were dropped from each sub-scale because they were either more general or blurred the distinction between punishing and helping. These constructs were negatively correlated, r = -.59, p < .001. The bivariate distribution of scores on support for punishing and helping showed that scores were low on both variables (see Table 1 for descriptive statistics).
Two items were used to measure political orientation, r = .78, p < .001. “Where would you place yourself politically on the following scale:” from 1 (extreme Democrat; very liberal) to 7 (extreme Republican; very conservative). For all study materials see Supplementary Materials.
Results
First, a confirmatory factor analysis examined the policy support items. Next, examples for the eight coded sub-themes were explained. Of the 85 participants who wrote about one of the coded themes, 50 (58.82%) mentioned the role of undocumented immigrants as agents, and 37 (43.53%) mentioned their role as victims; 6 (7.06%) of the respondents mentioned both. Then a confirmatory factor analysis was run with the coded agent and victim themes. Next, a mediational path model was run in Mplus with the agent and victim themes as predictors, the emotion composites as mediators, and the policy support as the outcome variables, controlling for political orientation.
Confirmatory Factor Analysis of Policy Support
A confirmatory factor analysis (CFA) was run with two factors, support for punishing and helping. Four items loaded onto each of the two factors. The fit statistics were good. The standardized root mean residual was excellent, SRMR = .06 (ideal SRMR ≤ .08). The Comparative Fit Index was excellent, CFI = .90 (ideal CFI ≥ .90). The root mean square effort of approximation was a little high, RMSEA = .12 (ideal RMSEA ≤ .08), but SRMR may be a better fit index than RMSEA (Shi, Maydeu-Olivares, & Rosseel, 2020). The model χ2 was, χ2(19) = 52.96, p < .001. The results of the CFA and the theoretical model justify the examination of support for punishing and helping as distinct variables.
Moral Typecasting Themes
Agent Typecast Themes
The most frequently mentioned agent theme was the belief that undocumented immigrants freeload by taking undeserved benefits, n = 27, 31.76%. For example, “Every time an immigrant is granted funding and other privileges, it takes away from what we as the citizens are entitled to.” The second most frequent agent theme was the belief that undocumented immigrants poach jobs from Americans, n = 24, 28.24%. “When illegal immigrants come into our country they take away from the limited job opportunities.” Participants claimed that this harm to “honest citizens” further exacerbates “what is going on with our economy and people getting laid off.” The third most frequent agent theme was the belief that undocumented immigrants do not pay taxes, n = 16, 18.82%. “It’s extremely infuriating that my family's tax money (etc.) goes to people that have not earned their right to stay in this country.” Another agent theme was the belief that undocumented immigrants hinder the process for legal immigrants coming to the U.S. n = 13, 15.29%. Such as, “I think that illegal immigrants are taking away from the immigrants who are actually coming to this country legally.” The final agent theme was the belief that undocumented immigrants threaten the safety of Americans, n = 7, 8.24%. “Without the ability to screen these people… we are not able to find out if these people are violent,” responsible for “drug-related incidents,” “crime and gangs,” or “terrorism”.
Victim Typecast Themes
The most frequently mentioned victim theme was the belief that undocumented immigrants fled bad circumstances in their countries of origin, n = 21, 24.71%. The “desperation” and “dire situation[s]” of undocumented immigrants included “circumstances of poverty, social or political distress,” and “religious persecution.” The second most frequent victim theme was the belief that undocumented immigrants are exploited in the U.S., n = 12, 14.12%. For example, “Due to their status as ‘illegal,’ these people often get taken advantage of by corporations and individuals.” They get hired to do “unethical” “dirty work.” The final victim theme was the belief that undocumented immigrants deserve basic human rights, n = 10, 11.76%. For example, “I don't think we as American[s] should ban these immigrants of their basic human rights.”
Confirmatory Factor Analysis for the Agent and Victim Typecasts
I hypothesized that the eight typecast themes would load onto two factors (Gray & Wegner, 2011). A confirmatory factor analysis was conducted with the five agent themes loading onto one factor and the three victim themes loading onto another factor in Mplus (CFA with categorical factor indicators; Muthén & Muthén, 1998-2017). The model was estimated using maximum likelihood estimation. The model fit was acceptable, χ2(19) = 23.95, p = .198, RMSEA = .05, NFI = .84, weighted root mean square residual (WRMR) = .72. The factor loadings of each indicator were also acceptable, agent r = .53-.98, and victim r = .61-.71. The results of the confirmatory factor analysis justified creating two dichotomous (present vs. not present) composite variables, agent and victim typecasts.
Support for Punishing Policies
Mplus was used to test a path model. Agent beliefs and victim beliefs were set as predictors in the model, controlling for the direct effect of political orientation. In addition, the indirect effects of agent and victim beliefs through the emotions were tested. Support for punishment and helping were tested simultaneously as outcome variables. While controlling for political orientation, the indirect effect of agent beliefs on punishment through empathy-related emotions was not significant, b = 0.02, SE = .02, p = .327, 95% CI [-0.01, 0.05], but as hypothesized the indirect effect through disgust/anger was significant, b = 0.23, SE = .07, p < .001, 95% CI [0.07, 0.36] (all results were standardized coefficients; see Table 2 and Figure 1). There was also a direct effect from agent beliefs to support for punishment, b = 0.25, SE = .09, p = .009, 95% CI [0.05, 0.42]. There were no indirect effects between victim beliefs and punishment, either through empathy-related emotions, b = 0.02, SE = .02, p = .222, 95% CI [-0.05, 0.01], or through disgust/anger, b = 0.01, SE = .03, p = .698, 95% CI [-0.07, 0.04]. Nor was there a direct effect of victim beliefs on punishment, b = 0.04, SE = .07, p = .628, 95% CI [-0.18, 0.11]. As hypothesized, political conservatism was positive associated with punishment in addition to the agent beliefs, b = 0.22, SE = .09, p = .016, 95% CI [0.04, 0.39]. The model explained 48.8% of the variance in punishment, SE = 0.07, p < .001.
Table 2
Study 1 Links in the Path Model
| Effect | a paths
|
b paths
|
Indirect Effects
|
Direct Effects
|
||||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| ab paths
|
c paths
|
|||||||||||||||
| Disgust and anger
|
Empathy-related emotions
|
Punishment
|
Helping
|
Punishment
|
Helping
|
Punishment
|
Helping
|
|||||||||
| Est. | SE | Est. | SE | Est. | SE | Est. | SE | Est. | SE | Est. | SE | Est. | SE | Est. | SE | |
| Agent beliefs | .56*** | .07 | -.12 | .09 | .25** | .09 | -.15 | .09 | ||||||||
| Victim beliefs | -.03 | .07 | .14 | .07 | -.04 | .07 | .07 | .09 | ||||||||
| Political orientation | .22* | .09 | -.06 | .10 | ||||||||||||
| Disgust/anger | .41*** | .10 | -.41*** | .11 | ||||||||||||
| Empathy-related emotions | -.13* | .06 | .21* | .07 | ||||||||||||
| Agent beliefs x disgust/anger | .23*** | .07 | -.23*** | .07 | ||||||||||||
| Agent beliefs x empathy-related emotions | .02 | .02 | -.02 | .02 | ||||||||||||
| Victim beliefs x disgust/anger | -.01 | .03 | .01 | .03 | ||||||||||||
| Victim beliefs x empathy-related emotions | -.02 | .02 | -.01 | .02 | ||||||||||||
*p < .05. **p < .01. ***p < .001.
Figure 1
Study 1 Path Model
Note. For simplicity, only significant paths were represented.
*p < .05. **p < .01. ***p < .001.
Support for Helping Policies
There was not an indirect effect of victim beliefs on helping through empathy-related emotions, b = 0.02, SE = .02, p = .327, 95% CI [-0.01, 0.06], counter to my hypothesis, and there was not an indirect effect through disgust/anger, b = 0.01, SE = .03, p = .699, 95% CI [-0.04, 0.07]. There was also no direct path from victim beliefs to helping, b = 0.07, SE = .09, p = .463, 95% CI [-0.11, 0.25]. There was not an indirect effect of agent beliefs on helping through empathy-related emotions, b = -0.01, SE = 0.02, p = .357, 95% CI [-0.05, 0.11], but there was an unexpected indirect path from agent beliefs to helping through disgust/anger, b = -0.23, SE = .07, p = .001, 95% CI [-0.36, -0.10]. There was not a direct effect of agent beliefs on helping, b = -0.15, SE = .09, p = .098, 95% CI [-0.33, 0.03]. Political liberalism was not associated with support for helping, b = -0.06, SE = .10, p = .506, 95% CI [-0.25, 0.12]. The model explained 32.4% of the variance in support for helping, SE = 0.08, p < .001.
Model Fit
The Standardized Root Mean Squared Residual was good, SRMR = 0.06, which should ideally be less than 0.08 and is a better measure of model fit than Root Mean Square Error of Approximation (RMSEA) when items are measured on different scales (Hooper, Coughlan, & Mullen, 2008). The non-normed fit index could have been better, NFI = 0.59; a good fit is considered ≥ 0.95. NFI is a better fit index than the comparative fit index (CFI) for small sample sizes. Ideally, model chi-square should not be significant, but for this model it was significant, χ2(3) = 16.59, p = .001. Model chi-square is worse when correlations in the model are larger (Shi et al., 2020); 9 of the 21 correlations reported in Table 2 were moderate to large (on Cohen’s scale).
Discussion
The coding of the open-ended responses revealed two major themes, beliefs about undocumented immigrants as either agents or victims of harm. There were five agent sub-themes and three victim sub-themes. The low co-occurrence of these themes (7.06%) fit with theorizing by Gray, Young, and Waytz (2012) that perceptions of agents and victims are mutually exclusive. However, no previous qualitative research has confirmed this idea.
Gray and Wegner (2011) proposed that viewing someone as an agent elicits disgust and anger and that viewing an individual as a victim elicits sadness and sympathy. As hypothesized, the link between agent beliefs and punishment was mediated by disgust/anger. Those that mentioned agent roles were more likely to feel disgust/anger toward undocumented immigrants. The disgust/anger composite was positively related to punishment. There was also a direct path (not hypothesized) between agent beliefs and punishment. There are myths in public discourse that immigrants hurt the economy or take jobs (Guskin & Wilson, 2017), which might explain this direct link. These results fit with both the TDM and the ITT and do not provide evidence, as I had expected, that the TDM offers a better explanation.
The indirect path from agent beliefs to lower helping mediated through disgust/anger was not predicted. I originally hypothesized that victim beliefs would be associated with helping. I also hypothesized that this link would be mediated through empathy-related emotions. I believed the TDM would help explain helping, but the results from Study 1 do not provide support for this idea. Rather the results were more in line with the ITT. Another shortcoming of Study 1 could be the measure of empathy. Research on empathy and perspective taking finds that the latter is a better predictor of people’s attitudes toward immigrants (Miklikowska, 2018). Thus, a measure of perspective taking may have been more effective. Also, I measured the presence or absence of these beliefs in participants’ responses, not the strength of the beliefs. Thus, the presence of agent beliefs may simply have been a more powerful antecedent. In both mediation analyses, disgust/anger mediated the link between agent beliefs and policy support while empathy-related emotions did not.
Analyzing the free responses of participants was a unique aspect of Study 1. However, due to the nature of count variables with low frequencies, I cannot determine the strength of the agent and victim beliefs. Therefore, in Study 2, I wrote close-ended items to measure the themes from Study 1. Another shortcoming of Study 1 was that I used time 2 variables (beliefs and emotions) to predict time 1 variables (policy support). Thus, Study 2 was cross-sectional. Only punishing (H1 and H3) but not helping was predicted as hypothesized (H2 and H4). Thus, the results of Study 1 offered only limited support for the TDM and are more in line with what the ITT would predict.
Study 2
Study 1 and Study 2 data were collected at different times (2011 and 2018, respectively). Between 2011 and 2018 debates about immigration in the U.S. became increasingly politicized. In 2011, neither Republicans nor Democrats in office were trying to help undocumented immigrants (Draper, 2018). In 2016, Trump ran on an anti-immigrant platform, including increasing deportation, building a boarder wall, and banning Muslim immigrants from the U.S. (Lopez, 2018). In the face of anti-immigrant rhetoric, some Democratics started advocating for pro-immigrant policies. Hence, for Study 2, I hypothesized that agent beliefs would indirectly predict punishment (H1) and that this path would be mediated through anger/disgust (H3) and that victim beliefs would indirectly predict helping (H2) and that this path would be mediated through empathy-related emotions (H4; Gray & Wegner, 2009, 2011).
Method
Participants and Design
Eighty-three undergraduate students at a mid-Atlantic public university participated for extra credit (from April 11-30, 2018). Sixty-three identified as female, 18 as male, 1 did not report biological sex, and 1 preferred not to answer. Fifty-eight identified as White, 13 as Black, 3 as Asian, 3 as Latino, 3 as multiracial, 1 as Native American, and 1 as unspecified (1 had missing data). One participant identified as undocumented, and one did not respond. Four participants reported having a family member who was undocumented, and one did not respond. Twenty-seven participants (32.53%) reported knowing someone who was undocumented, 1 preferred not to say, and 1 did not respond. The study was an online cross-sectional survey.
Measures
Closed-ended items were created to represent the coded variables from Study 1. Five statements represented agent beliefs (α = .84) such as, “Undocumented immigrants: freeload by taking undeserved benefits from Americans, take jobs from Americans, do not pay taxes, threaten the safety of Americans, hinder the process for legal immigrants coming to the U.S.” Three statements represented victim beliefs (α = .68) such as, “Undocumented immigrants: fled bad circumstances in their countries of origin, are exploited in the U.S., deserve basic human rights they may be denied in the U.S.” Each belief was measured on a scale from 0 (not at all) to 100 (very much).
Participants reported emotions toward undocumented immigrants including anger, disgust, sadness, guilt, pity, and empathy. Angry and disgusted were positively correlated, r = 0.64, p = .001, and averaged. The other emotions were averaged to form an empathy-related emotions composite (α = .68). Sadness, guilt, and pity were also used in Study 1. I added the word “empathy” in Study 2 given that it is the underlying emotion of interest.
Study 1 items were used to assess punishment (α = .83), including topics such as deportation, prison, fines, and termination. Study 1 items were used to measure helping (α = .82), including statements about allowing undocumented immigrants to collect welfare, giving them the right to vote, allowing them to receive scholarships, and providing free medical care.
Study 1 items were used to measure political orientation, r = .78, p < .001. (See Table 3 for all correlations.)
Table 3
Study 2 Descriptive Statistics and Correlations
| Variable | M | SD | 1 | 2 | 3 | 4 | 5 | 6 | 7 |
|---|---|---|---|---|---|---|---|---|---|
| 1. Agent beliefs | 40.40 | 24.02 | – | ||||||
| 2. Victim beliefs | 61.44 | 22.18 | -.10 | – | |||||
| 3. Disgust/anger | 2.33 | 1.47 | .62*** | -.25* | – | ||||
| 4. Empathy-related emotions | 3.57 | 1.48 | -.37*** | .48*** | -.20† | – | |||
| 5. Punishment | 2.53 | 1.32 | .66*** | -.40*** | .70*** | -.45*** | – | ||
| 6. Helping | 3.06 | 1.51 | -.42*** | .46*** | -.37*** | .52*** | -.53*** | – | |
| 7. Political orientation | 3.77 | 1.30 | .53*** | -.30** | .40*** | -.43*** | .53*** | -.63*** | – |
†p < .10. *p < .05. **p < .01. ***p < .001.
Results
Support for Punishing Policies
A path model was tested using Mplus with agent and victim beliefs were set as predictors, controlling for the direct effect of political orientation. The emotion composites were set as mediators. Punishment and helping were tested simultaneously as outcome variables. The indirect effects of agent and victim beliefs were tested in addition to the direct effects. As predicted, the indirect effect from agent beliefs to punishment through disgust/anger was significant, b = 0.24, SE = .06, p < .001, 95% CI [0.12, 0.39] (all reported results include standardized coefficients; see Table 4 and Figure 2), but the indirect effect through empathy-related emotions was not, b = 0.05, SE = .04, p = .230, 95% CI [-0.02, 0.13]. The direct path from agent beliefs to punishment was significant, b = 0.37, SE = .08, p < .001, 95% CI [0.19, 0.52]. Victim beliefs were not related to punishment through disgust/anger, b = -0.04, SE = .04, p = .233, 95% CI [-0.12, 0.02] or empathy-related emotions, b = -0.06, SE = .05, p = .191, 95% CI [-0.16, 0.02]. There was not a direct association between victim beliefs and punishment, b = -0.13, SE = .08, p = .091, 95% CI [-0.28, 0.02]. Political orientation was not related to punishment (see Table 4). This model explained 68.1% of the variance in support for punishment, SE = .06, p < .001.
Table 4
Study 2 Links in the Path Model
| Effect | a paths
|
b paths
|
Indirect Effects
|
Direct Effects
|
||||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| ab paths
|
c paths
|
|||||||||||||||
| Disgust/anger
|
Empathy-related emotions
|
Punishment
|
Helping
|
Punishment
|
Helping
|
Punishment
|
Helping
|
|||||||||
| Est. | SE | Est. | SE | Est. | SE | Est. | SE | Est. | SE | Est. | SE | Est. | SE | Est. | SE | |
| Agent beliefs | .64*** | .08 | -.32*** | .09 | .37*** | .08 | -.08 | .14 | ||||||||
| Victim beliefs | -.11 | .09 | .42*** | .07 | -.13 | .08 | .20* | .09 | ||||||||
| Political orientation | .06 | .10 | -.41*** | .09 | ||||||||||||
| Disgust/anger | .37*** | .08 | -.08 | .11 | ||||||||||||
| Empathy-related emotions | -.14 | .10 | .21* | .09 | ||||||||||||
| Agent beliefs x disgust/anger | .24*** | .07 | -.05 | .07 | ||||||||||||
| Agent beliefs x empathy-related emotions | .05 | .04 | -.07 | .03 | ||||||||||||
| Victim beliefs x disgust/anger | -.04 | .04 | .01 | .02 | ||||||||||||
| Victim beliefs x empathy-related emotions | -.06 | .05 | .09* | .04 | ||||||||||||
*p < .05. **p < .01. ***p < .001.
Figure 2
Study 2 Path Model
Note. Only significant paths were included for simplicity of the figure.
*p < .05. **p < .01. ***p < .001.
Support for Helping Policies
As predicted, the indirect path from victim beliefs to support for helping through empathy-related emotions was significant, b = 0.09, SE = .04, p = .041, 95% CI [0.01, 0.18] but the indirect path through disgust/anger was not, b = 0.01, SE = .02, p = .618, 95% CI [-0.03, 0.05]. The direct path from victim beliefs to helping was also significant, b = 0.20, SE = .09, p = .027, 95% CI [0.03, 0.38]. The indirect path from agent beliefs to helping through empathy-related emotions was not significant, b = -0.07, SE = .04, p = .051, 95% CI [-0.14, -0.01], and nor was the link through disgust/anger, b = -0.05, SE = .07, p = .488, 95% CI [-0.19, 0.10]. The direct path from agent beliefs to helping was not significant, b = -0.08, SE = .14, p = .562, 95% CI [-0.37, 0.18]. Political conservatism was negatively associated to helping, as expected, in addition to victim beliefs (see Table 4). This model explained 51.2% of the variance in support for helping, SE = .06, p < .001.
Model Fit
The results reported above were simultaneously tested in one model. The model fit statistics were good. The Standardized Root Mean Square should be less than .08, and SRMS = .032, indicating that the model fit was good. The non-normed fit index was close to what would be considered the good cut-off of .95, NFI = .913. Chi-square for model fit should be non-significant, as it was for this model, χ2(3) = 6.12, p = .106.
Discussion
The findings of Study 2 closely fit the hypotheses that agent and victim beliefs were related to support for punishment and helping respectively (H1 and H2). The link between agent beliefs and punishment was mediated by disgust/anger (H3). In addition, agent beliefs were directly associated with punishment. Harmful acts are often punished in society to deter crime and serve justice (Durrant, 2014). The link between victim beliefs and helping was mediated by empathy-related emotions as Gray and Wegner (2011) suggested (H4). Likewise, there was also a direct link between victim beliefs and helping. Research on helping behavior finds that those who can be blamed for their plight will not elicit help, but those who are perceived as innocent victims will be helped (Brickman, Rabinowitz, Karuza Jr., Coates, Cohn, & Kidder, 1982).
Study 2 developed closed-ended items to assess victim and agent beliefs. The results of Study 2 were more in line with the TDM (Schein & Gray, 2018) unlike the Study 1 results because the Study 2 results on helping go beyond what the ITT could predict. These assessments of people’s judgments and emotions provide valuable insight into ongoing debates about immigration policies. Many nations are currently debating what types of policies to implement in response to climate migrants (Davidson & Carr, 2010) and those who are seeking asylum (Felter et al., 2022). These research insights about moral typecasts are relevant to border security and policies toward undocumented immigrants already in a country, and they may also be critical to policies regarding future immigration.
General Discussion
This research found only partial support for the TDM (H3 in Study 1 and H2, H3, and H4 in Study 2) and some support for the ITT (H1 in Studies 1 and 2). Thus, the contribution of the current research is modest. However, this is the first research that has applied insights from the TDM (Schein & Gray, 2018) to policy perspectives about undocumented immigrants. Prior research on attitudes toward immigration generally has focused on examining the ITT (Scheibner & Morrison, 2009; Stephan et al., 1998). Schein and Gray (2018) claimed that the TDM is a general framework for thinking about moral judgments that should be tested and applied to positive or negative behaviors; this research contributes to these scholarly efforts. The present research also uncovers understanding of people’s responses to real situations, going beyond vignettes, which are typical in moral psychology research (Gray & Wegner, 2009). The application of the TDM in the current research contributes to the understanding that victim perceptions influence helping tendencies toward undocumented immigrants, something the ITT cannot explain as it was developed to elucidate prejudice (Stephan & Stephan, 1996; Stephan et al., 1998). Future research could directly test both theories to predict helping.
The results for punishment (H1 and H3) were consistent across both studies (conducted in 2011 and 2018, respectively) likely because punishment is viewed as a fundamental response to people who are perceived as harm doers (Durrant, 2014). On the other hand, the results for helping were not as consistent. Acceptance of helping could have changed across time (2011 vs. 2018). In 2011, there were low perceptions of undocumented immigrants as victims, and thus, helping was a low policy priority (Draper, 2018). However, during the 2018 mid-term campaign, a wider array of helping policies were suggested. The degree to which undocumented immigration has been raised in the public consciousness may have also increased the appeal of helping as a viable and perhaps morally necessary political option.
Alternatively, the poor support for H2 and H4 in Study 1 could be due to the low interrater reliabilities for the victim beliefs. Another potential shortcoming is that the studies used small largely White female college samples. The generalizability of the results is limited as a result. However, college students are a demographic with historically low voter turn-out. The 2018 midterm election was an important turning point in this trend and a time during which immigration became an important political issue (Thomas et al., 2020). Another weakness was that typecast beliefs were measured and not manipulated (for the manipulation of typecasts in a criminal context see Robbins & Litton, 2018). However, typecast beliefs were measured in two ways (open-ended and closed-ended).
Research on undocumented immigrants could not only benefit from further examination of moral typecast beliefs but from measuring anti- and pro-immigrant policies independently. My research helped demonstrate that helping and harming are unique dimensions, not opposite ends of one variable (Louis et al., 2013; Stephan et al., 1998). A useful framework for understanding the motives for harming or helping immigrants comes from research on moral prescriptions (should nots, e.g., hurting) vs. proscriptions (shoulds, e.g., helping; Janoff-Bulman, Sheikh, & Hepp, 2009). Just as prescriptions and proscriptions are not opposites, harming and helping are not opposites. People may be compelled to avoid harming undocumented immigrants, but that does not mean they will be highly motivated to offer them help either.
The mutually exclusive nature of moral typecasts perceptions (Gray & Wegner, 2009; Gray, Young, & Waytz, 2012) and accompanying moral emotions (Weiner, 1980) that I found in the data might help explain the ongoing debate about undocumented immigration. Recognizing that people cannot see undocumented immigrants as harmful agents and potential victims simultaneously could open political dialogue and address diverse national interests related to immigration reform. Those working for more inclusive policies toward undocumented immigrants should consider that sharing stories of their victimization (the current approach) and getting their opponents to think of undocumented immigrants as harmless might go far to build support for pro-immigrant policies.