Parochial Empathy Predicts Reduced Altruism and the Endorsement of Passive Harm.

Empathic failures are common in hostile intergroup contexts; repairing empathy is therefore a major focus of peacebuilding efforts. However, it is unclear which aspect of empathy is most relevant to intergroup conflict. Although trait empathic concern predicts prosociality in interpersonal settings, we hypothesized that the best predictor of meaningful intergroup attitudes and behaviors might not be the general capacity for empathy (i.e., trait empathy), but the difference in empathy felt for the in-group versus the out-group, or "parochial empathy." Specifically, we predicted that out-group empathy would inhibit intergroup harm and promote intergroup helping, whereas in-group empathy would have the opposite effect. In three intergroup contexts-Americans regarding Arabs, Hungarians regarding refugees, Greeks regarding Germans-we found support for this hypothesis. In all samples, in-group and out-group empathy had independent, significant, and opposite effects on intergroup outcomes, controlling for trait empathic concern.

In our everyday lives, we count on empathic people to do altruistic things. In interpersonal interactions, this assumption is supported by studies showing that empathic concern (EC) is associated with prosocial behaviors (Batson, 2014; Eisenberg & Miller, 1987). For example, observers who experience more EC for a needy individual are more likely to engage in costly helping, including volunteering more time, donating more money, and taking on boring or painful tasks (Batson, Chang, Orr, & Rowland, 2002; Batson, Duncan, Ackerman, Buckley, & Birch, 1981; Fultz, Batson, Fortenbach, McCarthy, & Varney, 1986). A large literature has also demonstrated that dispositional EC is associated with prosocial behavior and intentions (Eisenberg, Eggum, & Di Giunta, 2010) and also with a decrease in antisocial behaviors, such as bullying (Jolliffe & Farrington, 2006; Raskauskas, Gregory, Harvey, Rifshana, & Evans, 2010).

Since empathy is involved in providing help and withholding harm toward others, it has been suggested that empathy may play a vital role in preventing intergroup conflict and facilitating reconciliation (Batson & Ahmad, 2009). Many conflict resolution programs have therefore adopted fostering empathy as a primary program goal. However, there is reason to believe that boosting overall EC may not provide a universal palliative in intergroup settings. In fact, interviews with those who engage in (or attempt) extreme intergroup violence indicate that these individuals are characterized not by a lack of empathy but rather by high levels of empathy and communal concern for their in-group (Argo, 2009). Similarly, among Israelis and Palestinians, the willingness to endorse out-group harm is associated with EC toward their own communities (Ginges & Atran, 2009; see also Lickel, Miller, Stenstrom, Denson, & Schmader, 2006). Therefore, in some cases, inducing empathy may not only fail to heal intergroup wounds, but it may motivate out-group hostility. Understanding how different facets of empathy influence intergroup conflicts is critical to practical conflict resolution and peacebuilding efforts. It is also of great theoretical significance, given the prominent role that empathy plays in many models of intergroup relations (e.g., Batson & Ahmad, 2009; Lickel et al., 2006; Stephan & Stephan, 2001).

The opposing predictions about the effect of empathy on intergroup conflict can be resolved if we think of empathy not only as an interpersonal process but also as a group-based emotion—an emotional response that arises because of, or is shaped by, social identities (Smith & Mackie, 2016). Group-based emotions can motivate us to act on behalf of our in-group and/or against hostile out-groups. For example, people experience anger after an in-group insult but only to the extent that they identify with that group (Rydell et al., 2008). Similarly, shame and guilt may be felt more strongly by those who identify with or feel attached to their group: Americans feel more shame when considering the poor treatment of Arabs in the United States in the wake of 9/11 (Johns, Schmader, & Lickel, 2005), and among Americans and Israelis, identification (ID) or attachment to the group, respectively, facilitates group-based guilt about the poor treatment of minorities in their society (Doosje, Branscombe, Spears, & Manstead, 2006; Roccas, Klar, & Liviatan, 2006).

In the current work, we take an intergroup emotions theory (Mackie, Maitner, & Smith, 2009) perspective of empathy and suggest that intergroup empathy has two simultaneous effects: If we feel empathy for individual out-group members, it may motivate action on behalf of that out-group; whereas if we feel empathy for individual in-group members, it may motivate action against the out-group. What types of behavior might intergroup empathy most likely motivate? Whereas moral emotions like shame and guilt have been associated with conciliatory gestures (e.g., support of negotiations or reparations), and negative emotions like anger have been shown to precipitate active harm, empathy has been most strongly associated with altruism. Therefore, we predicted that low out-group empathy and high in-group empathy would be most strongly associated with withholding altruistic behavior toward the out-group, or preventing others from easing out-group suffering (i.e., “passive harm”). From this perspective, the people most likely to withhold out-group altruism or engage in passive harm, would not be those low in trait levels of empathy, such as psychopaths (Baron-Cohen, 2012), but rather those who simultaneously display low out-group empathy and high in-group empathy. Thus, by contrast to interpersonal interactions, intergroup interactions may be best predicted by the parochial distribution of empathy toward in-group versus out-group members, which we term here “parochial empathy” (Bruneau, Cikara, & Saxe, 2015) or “intergroup empathy bias” (Cikara, Bruneau, & Saxe, 2011).

The intergroup emotions perspective on empathy leads to three predictions: First, the amount of empathy felt for out-group members will correlate with social ID with the out-group, and the amount of empathy felt for in-group members will correlate with social ID with the in-group. Second, intergroup emotions will predict out-group attitudes and behavior—that is, greater empathy for individual out-group members will predict greater altruism (and less hostility) toward the out-group and vice versa. Because “in-group love” can be distinct from “out-group hate” (Brewer, 1999), in-group and out-group empathy should predict behavior independently of each other. Finally, parochial empathy may be independent of interpersonal empathy, predicting altruism and antagonism across salient group boundaries, above and beyond trait levels of empathy.

Current Research

The goal of the present work is to demonstrate the consequences of intergroup empathy in a range of real intergroup contexts. Since we predict that in-group empathy and out-group empathy will have opposite effects on out-group attitudes and behavior, a parsimonious way to operationalize intergroup empathy is by calculating the difference in empathy felt toward in-group versus out-group members, within subjects. We take this approach in the current work. However, since a single difference score fails to reveal the individual effects of each component of empathy, and therefore obscures their independent contributions, we also examine in-group and out-group empathy as independent factors. In Experiment 1, we tested whether parochial empathy (1) predicted out-group attitudes and costly altruism over time and (2) mediated the effect of social ID on these outcomes. In Studies 2 and 3, we examined another key question: (3) whether parochial empathy predicted out-group attitudes and costly helping above and beyond trait EC.

Experiment 1: Parochial Empathy in Americans Regarding Arabs

To examine parochial empathy, we presented American participants with fortunes and misfortunes experienced by Americans (in-group) and Arabs (out-group) by adapting a paradigm used previously with novel groups (Cikara, Bruneau, Van Bavel, & Saxe, 2014; Bruneau et al., 2015).[1] Given that the United States has been involved in continual warfare with at least one Arab-majority country since 2001, these groups are characterized by a hostile relationship. We hypothesized that (1) both in-group empathy and out-group empathy would independently predict diminished altruistic (i.e., costly helping) behavior toward the out-group, and given that intergroup emotions are fundamentally shaped by social identities, that (2) the difference in empathy expressed toward in-group versus out-group members (parochial empathy) would mediate the effect of social identity on out-group behavior.

Method

Participants

We recruited a sample of American participants (N = 100) via Amazon’s Mechanical Turk website. Previous studies using a similar paradigm recruited 50–100 participants (Cikara et al., 2014); because the current study focused on individual differences, rather than mean effects, we planned a priori to recruit at the top end of this scale (N = 100). Data collection was terminated automatically after 100 individuals completed the study. Of the participants, 16 missed one of the two attention check questions embedded in the survey, leaving 84 participants (M age = 37.0, standard deviation [SD] = 12.0; 51.2% male).

Experimental Design

Procedure

The procedure was adapted from a paradigm developed by Cikara, Bruneau, Van Bavel, and Saxe (2014). Briefly, after being provided with a cover story for the research (i.e., to examine “directional problem-solving” in people who learned languages that are written in different directions), English-speaking American participants were told that they would be playing as a member of an English-speaking team in a problem solving challenge, against a team of people who speak Arabic. Participants were told that individual scores would be tallied and added to the total team score for both teams and that the first team to 100 points would win the challenge.

Prior to the challenge, participants reported where they grew up, their background, and an event that happened to them in the past week. After being assigned to a team (English speakers) and told that they were (apparently randomly) assigned to compete against Arabic speakers, participants were told that they would read the events that people from their team and the other team had reported.

Participants then read the events of eight English-speaking people on their team (in-group) and the eight Arabic-speaking people on the other team (out-group). The 16 specific events were adapted from previous work (Cikara et al., 2014) and adjusted to be culturally neutral (e.g., “[target] came down with a serious illness”). Each event description included the target’s group membership (e.g., “Beth is from NORTH DAKOTA” or “Salma is from EGYPT”) and the target’s language (“English” or “العربية”). Events were randomly assigned to either an in-group or an out-group target, such that each participant viewed eight in-group events (four positive and four negative) and eight out-group events (four positive and four negative). For each of the 16 events, participants reported how good and how bad it made them feel that the event happened to the target. Not all measures are reported here (for full survey, see Supplemental Materials).

Parochial empathy was measured as the degree to which people felt good about in-group versus out-group fortunes and bad about in-group versus out-group misfortunes. For the purposes of the mediation analysis, parochial empathy was calculated as in-group empathy minus out-group empathy.

Intergroup ID was assessed using the Inclusion of In-group and Out-group in the Self measure (Schubert & Otten, 2002). The degree of overlap with the in-group using this measure has been demonstrated previously to reflect in-group ID (Tropp & Wright, 2001). Participants used their mouse to drag the “self” circle closer or further away from the “group” circle on screen, providing a continuous measure of self/in-group and self/out-group ID.

Following the measures above, participants were presented with eight progressive Raven’s Matrices (which involve directional problem-solving, supporting the cover story).

Out-group altruism

After finishing the Raven’s Matrices, participants were given the opportunity to complete additional problems (up to 20), with each correct problem providing a US$.10 donation to the “Arab Red Crescent Society.” At the bottom of each problem were buttons to “opt out” and skip to the end of the study or “continue” to the next charity problem. The number of problems completed for charity provided a measure of costly altruism.

Follow-Up Survey

One week later, participants were provided with a link to a follow-up survey, which 68 (81%) completed (and passed the check question). Those who did and did not complete the follow-up survey did not differ on age, gender, in-group ID, out-group ID, in-group empathy, out-group empathy, or the number of tasks completed for charity (ps > .25).

The follow-up survey included exploratory measures that paralleled unrelated concurrent cross-cultural research: emotions toward Arabs associated with stereotype content (envy, pride, disgust, and pity), prejudice and dehumanization toward Arabs, and emotional responses to injustices committed against Arabs. These items are not examined here. Also included in the follow-up survey were outcome measures associated with out-group helping designed for the current study (support for Arab immigration, donation to out-group members; for full survey, see Supplemental Materials).

Support Arab immigration assessed the percentage of U.S. VISAS participants thought should be awarded to each of the following groups: Arabs, East Asians, Hispanics, Africans, and Eastern Europeans (Kteily, Bruneau, Waytz, & Cotterill, 2015). We used as our measure the percent of VISAS granted to Arabs.

Out-group donation assessed participants’ altruistic behavior by providing them with a monetary bonus to distribute between an in-group cause (relief fund for victims of the Boston Marathon bombings) versus an out-group cause (relief fund for civilian victims of drone strikes in Afghanistan and Yemen; Kteily et al., 2015). The amount of money participants donated to the out-group cause provided our measure of out-group prosocial behavior.

Mediation Test

To test for mediation, we constructed three separate models using PROCESS (Hayes, 2013). Each model had the difference between in-group and out-group ID as the predictor, the difference between in-group and out-group empathy (i.e., parochial empathy) as the mediator and one of the outcome measures as the dependent variable. To test the specificity of mediation, we examined both the predicted mediation relationship (intergroup ID as independent variable, parochial empathy as mediator) and the reverse (parochial empathy as independent variable, intergroup ID as mediator), as suggested by Hayes (2013).

Results

As predicted, American participants reported stronger ID with the in-group (M = 87.6, SD = 18.1) than the out-group (M = 44.3, SD = 23.4), t(83) = 12.20, p < .001, Cohen’s d = 2.09. We took as our measure of “intergroup ID” the difference between these social identity measures—this measure correlated significantly with all outcome measures (rs > .23, ps < .04). American participants also had stronger empathic responses to Americans (M = 80.7, SD = 12.2) than Arabs (M = 78.3, SD = 15.2), t(83) = 2.10, p = .039, Cohen’s d = .18. We calculated parochial empathy as the difference between the empathy measures. Parochial empathy correlated significantly with all outcome measures (rs > .24, ps < .03), and intergroup ID and parochial empathy were significantly correlated with each other, ρ(84) = .350, p = .001).

The main analyses of interest examined the ability of parochial empathy to predict outcomes at Time 1 (immediately after the in-group and out-group empathy judgments) and Time 2 (1 week after the empathy task), controlling for intergroup ID. We first tested the specific hypothesis that the constituent components of parochial empathy are independently associated with out-group behaviors—that is, that out-group empathy drives prosocial out-group actions, while in-group empathy inhibits such actions. We mean centered all predictors and entered them simultaneously. As predicted, in-group and out-group empathy were independent and significant predictors of immigration support and donations at Time 2, above and beyond in-group and out-group ID (Table 1). In-group and out-group empathy did not predict the Time 1 outcome, beyond in-group and out-group ID.

Table 1.

In-Group (Ing) and Out-Group (Outg) Empathy Independently (and Oppositely) Predict Outcome Measures at Time 2 (1 Week After Empathy Judgments), Controlling for In-Group Identification (Ing ID) and Out-Group Identification (Outg ID).

	Out-Group Altruism (T1) R 2 = .14			Support Arab Immigration (T2) R 2 = .32			Out-Group Donation (T2) R 2 = .25
Independent Variable	β	B	[95% CI]	β	B	[95% CI]	β	B	[95% CI]
Ing empathy	−.15	−.077	[−.249, .095]	−.35*	−.215	[−.412, −.019]	−.63***	−.007	[−.011, −.003]
Outg empathy	.20	.086	[−.059, .231]	.55**	.276	[.113, .439]	.58**	.005	[.002, .009]
Ing ID	.15	.054	[−.024, .132]	.04	.017	[−.083, .118]	.00	.000	[−.002, .002]
Outg ID	.29*	.081	[.017, .145]	.30*	.097	[.020, .174]	.11	.001	[−.001, .002]

Note. T1 = Time 1; T2 = Time 2.

*p < .05. **p < .01. ***p < .001.

We also examined the specific prediction that parochial empathy serves as a direct psychological motivator of out-group behavior through mediation analyses. We found that parochial empathy mediated the effect of intergroup ID on all outcome measures: Out-group altruism (indirect effect: −.0112, SE = .0072, 95% CI [−.0292, −.0014]), Arab immigration support (indirect effect: −.0316, SE = .0164, 95% CI [−.0693, −.0086]), and out-group donation (indirect effect: −.0006, SE = .0003, 95% CI [−.0014, −.0001]; Figure 1). Importantly, the alternate model (i.e., intergroup ID mediating effects of parochial empathy on outcomes) was not significant for any of these outcome measures (95% CI of indirect effects for all models included 0).

Figure 1.

Parochial empathy mediates the effect of intergroup identification on (A) support for Arab immigration and (B) out-group donations to victims of drone strikes. *p < .05. **p < .005.

Therefore, Experiment 1 supported the prediction that parochial empathy is an important predictor of attitudes and behavior in consequential intergroup contexts, mediating the effect of social identity on outcomes. Study 2 sought to replicate and extend these results by examining the predictive validity of parochial empathy in a different cultural context and directly comparing the contribution of parochial empathy and trait EC to out-group attitudes and altruism.

Experiment 2: Parochial Empathy Among Hungarians During the Refugee Crisis

Experiment 1 demonstrated that parochial empathy mediates the effect of intergroup ID on out-group-oriented attitudes and behavior. In Experiment 2, we sought to (1) confirm the role of parochial empathy in predicting outcomes in an independent sample and (2) determine whether parochial empathy predicts outcomes beyond trait EC. In Study 2, we examined Hungarian empathy toward Hungarians and Muslim refugees during the European “Refugee Crisis.”

In early 2015, the Hungarian government launched a major anti-immigration campaign, which included billboards and high-profile comments by Hungarian leaders citing concern for the welfare of in-group members during the influx of Muslim refugees from conflicts in the Middle East. At the same time, empathy toward the refugees was fostered by media coverage of the wars that drove them from home, and the harrowing journey across the Mediterranean, which resulted in thousands of drowning deaths. In particular, in September 2015, the image of Aylan Kurdi, a 2-year-old Syrian boy whose drowned body was collected on a beach in Turkey, was transmitted around the world, resulting in up to 60-fold increases in donations to refugee charities for weeks afterward (Gladstone & Zraick, 2015). Experiment 2 was conducted 2 months after the Aylan Kurdi photo was published and 1 month after the peak (until now) of refugees coming to Europe (220,000 arrivals in October 2015; United Nations High Commissioner for Refugees, 2016). This intergroup context therefore seemed well suited to examine parochial empathy.

We recruited a sample of 604 Hungarian participants (the largest sample we could collect given funding constraints) in late November 2015 through a Hungarian survey company (Kerdoivem); the sample was representative of Hungary in terms of age and gender. Of the participants, 102 missed one of the two attention check questions embedded in the survey, leaving 502 participants (M age = 40.56, SD = 13.15; 46.6% male).

Participants were provided with a survey (in Hungarian), which included measures of parochial empathy, trait empathy, and three outcome measures, as described below. The survey included other measures of interest examined elsewhere (Bruneau, Kteily, & Laustsen, 2017; for full survey, see Supplemental Materials).

Parochial empathy was assessed through eight scenarios involving individual misfortunes (e.g., “[Hungarians/Muslim refugees] whose children don’t do well in school”; for full set of stimuli, see Supplemental Materials). Each scenario had an in-group (Hungarian) target version and an out-group (Muslim refugee) target version. Participants were randomly presented with one set of four misfortunes attributed to the out-group (α = .90), and then the other set of four misfortunes attributed to the in-group (α = .82). For each story, participants indicated empathy for the target (“How much empathy do you feel for [target]?” from 1 “none at all” to 5 “a lot”). Parochial empathy was assessed as in-group versus out-group empathy.

Trait EC was measured with the Empathic Concern subscale of the Interpersonal Reactivity Index (IRI; Davis, 1983; α = .80).

Anti-refugee policies support was assessed by providing participants with four policies that were debated during Europe’s refugee “crisis”: for example, “We should dramatically decrease the amount of aid we provide to refugees in order to deter them from trying to come to our country” (reverse coded). Ratings were made using sliders anchored at 0 (strongly oppose) and 100 (strongly support; α = .84).

Asylum was assessed by asking how many refugees participants would be willing to accept into Hungary: “Of the estimated 1,000,000 refugees who could reach Europe this year, how many do you think Hungary should grant asylum to, allowing them to live there permanently? (range: 0–40,000).” Because of the large range of possible responses, results were log transformed.

Sign prorefugee petition provided participants with the opportunity to sign a petition in support of refugee aid. Participants reported whether they wanted their vote counted against the petition (coded −1), for the petition (coded 1), or if they did not want their vote counted (coded 0).

Hungarian participants reported stronger empathic responses to the misfortunes of in-group Hungarians (M = 3.34, SD = .81) compared to out-group Muslim refugees (M = 3.06, SD = 1.00), t(501) = 9.4, p < .001, Cohen’s d = .31. Trait EC was correlated positively and quite strongly with both in-group empathy, r(502) = .40, p < .001, and out-group empathy, r(502) = .46, p < .001), suggesting that trait EC is a good predictor of empathic responses toward both in-group and out-group targets. On the other hand, trait EC was significantly less strongly correlated with parochial empathy, r(502) = −.21, p = < .001 (Steiger’s Zs > 8.9, ps < .001).

We next tested how strongly trait empathy and parochial empathy correlated with support for outcomes relevant to Muslim refugees: support for anti-refugee policies, the number of Muslim refugees Hungarians thought should be allowed to settle in Hungary, and willingness to sign a petition in support of refugees. Consistent with the American sample, in-group and out-group empathy independently and significantly predicted each of the outcome measures, even when controlling for trait EC (Table 2).[2] Controlling for trait EC, parochial empathy strongly predicted all outcome measures: support for anti-refugee policies (β = .42, p < .001), the number of refugees participants felt should be settled in Hungary (β = −.47, p < .001), and willingness to sign a petition in support of refugees (β = −.33, p < .001); results were similar when we controlled in the regression analyses for age, gender, and conservatism (|β|s > .23, ps < .001).

Table 2.

In-Group (Ing) and Out-Group (Outg) Empathy Independently (and Oppositely) Predict Attitudes and Behavior Toward Refugees Among Hungarians, Controlling for Trait Empathic Concern (EC).

	Support Anti-Refugee Policies R 2 = .36			# Refugees in Hungary (log) R 2 = .35			Sign Pro-Refugees Petition R 2 = .24
Independent Variable	β	B	[95% CI]	β	B	[95% CI]	β	B	[95% CI]
Ing empathy	.24***	0.86	[0.46, 1.25]	−.32***	−1.6	[−2.1, −1.1]	−.16**	−.16	[−.28, −.04]
Outg empathy	−.74***	−2.17	[−2.5, −1.8]	.80***	3.2	[2.8, 3.7]	.59***	.48	[.38, .58]
Trait EC	−.05	−0.30	[−0.82, 0.22]	−.01	−0.05	[−0.77, 0.67]	.03	.06	[−.10, .21]

*p < .05. **p < .01. ***p < .001.

These results again highlight the importance of parochial empathy in intergroup contexts through the independent effects of out-group and in-group empathy. Study 2 also illustrated that parochial empathy is distinct from trait EC and is a stronger predictor of outcomes. In Study 3, we sought to replicate the results of Study 2 in yet another consequential intergroup context, with an outcome measure associated with support for passive out-group harm.

Experiment 3: Parochial Empathy Among Greeks During the Greek Depression

Two years after the global financial crisis that began in 2007, Greece experienced what is now known as the “Greek Depression.” Greek debt ballooned to €30,000 per capita, which was held mostly in German banks. When Greece appealed to the European Union to relieve some of their debt as part of a bailout plan, their appeal was largely rejected, particularly by Germany, and harsh austerity measures were instituted with widespread consequences for individual Greeks that were well documented in the media. Claims of predatory lending by Germany and German hypocrisy were common in Greece, and anti-German sentiments were strong. In the wake of these sociopolitical decisions, we assessed trait empathy, parochial empathy, and anti-German hostility in a large sample of Greeks.

We recruited 545 Greek participants (the largest sample we could collect given funding constraints) to participate in the survey in late September 2015 through a Greek survey company (“The Hellenic Research House”). Of the participants, 78 missed an attention check question embedded in the survey, resulting in 467 participants (M age = 39.24, SD = 10.36; 36% male).

Participants were provided with an omnibus survey (in Greek), which included measures of parochial empathy, trait empathy, and the outcome measures. Other measures (e.g., outcomes related to the Roma minority population) are not reported here (for full survey, see Supplemental Materials).

Parochial empathy was assessed as in Study 2, this time toward German (out-group; α = .86/.74 for Versions 1 and 2) and Greek (in-group) targets (α = .70/.67 for Versions 1 and 2).

Trait EC was measured with the Empathic Concern subscale of the IRI (α = .69).

Save German lives provided participants with a hypothetical scenario in which Greek intelligence could pursue information that would cost considerable resources but could result in the thwarting of a terror attack in Germany (i.e., costly helping). Participants were asked to respond how strongly they supported following the lead (and spending Greek resources to potentially save German lives; for full description, see Supplemental Materials).

Passive harm was assessed with 6 items (e.g., “If Germany were to have a natural disaster, I think Greece should withhold financial support”; α = .76). Responses were provided on unmarked sliders anchored at “completely disagree” (0) and “completely agree” (100).

Greek participants reported stronger empathic responses to the misfortunes of in-group Greeks (M = 4.06, SD = .73) compared to out-group Germans (M = 3.47, SD = 1.00), t(466) = 13.3, p < .001, Cohen’s d = .68. Consistent with Study 2, we found that trait EC was significantly correlated with empathy for in-group targets, r(467) = .29, p < .001, and empathy for out-group targets, r(467) = .28, p < .001. On the other hand, trait EC and parochial empathy were uncorrelated, r(467) = −.07, p = .11.

Consistent with Experiment 2, but this time with a measure of passive harm rather than active help, in-group and out-group empathy independently and significantly predicted the degree to which participants supported passive harm against Germany, controlling for trait EC. Only out-group empathy significantly predicted willingness to engage in costly helping to spare German lives (Table 3).[3] Controlling for trait EC, parochial empathy predicted support for passive harm (β = .36, p < .001) and costly helping to save German lives (β = −.25, p < .001). Results for all regression analyses were similar if age, gender, and conservatism were included as additional regressors. Therefore, parochial empathy not only predicted the withholding of prosocial behavior toward an out-group but also support for passive harm.

Table 3.

In-Group (Ing) and Out-Group (Outg) Empathy Independently (and Oppositely) Predict Support for Passive Harm of Germans among Greeks, Controlling for Trait Empathic Concern (EC).

	Passive Harm R 2 = .16			Save German Lives R 2 = .10
Independent Variable	β	B	[95% CI]	β	B	[95% CI]
Ing empathy	.20***	5.41	[2.8, 8.02]	−.04	−1.86	[−5.95, 2.23]
Outg empathy	−.42***	−8.45	[−10.4, −6.54]	.33***	10.1	[7.15, 13.1]
Trait EC	.08	−3.81	[−8.31, 0.688]	.02	1.17	[−5.89, 8.23]

*p < .05. **p < .01. ***p < .001.

General Discussion

Across three experiments, we examined parochial empathy and its effects on attitudes and behavior toward threatening out-groups. In the context of American/Arab relations (Study 1), in-group and out-group empathy each independently and significantly predicted long-term attitudes and behavior toward Arabs and mediated the effect of intergroup ID on these outcomes. In the context of Hungarian/refugees (Study 2) and Greek/German relations (Study 3), in-group and out-group empathy significantly and oppositely predicted altruistic attitudes and behaviors toward Muslim refugees (in Hungarians) and passive harm of Germans (in Greeks), this time controlling for trait EC.

Across a range of intergroup contexts, we found that while trait EC was correlated with both in-group and out-group empathy ratings separately, trait EC was not correlated with the difference between in-group and out-group empathy (i.e., parochial empathy). More importantly, across all three cultural contexts, parochial empathy was a stronger predictor of altruism and passive harm toward out-groups than was trait empathy. Indeed, the current research indicates that trait EC is a poor predictor of out-group altruism and out-group passive harm. Rather, the research reported here supports an alternative empathy profile for those who withhold aid and advocate passive harm to the out-group: Low enough empathy for the out-group that withholding help or condoning harm is emotionally acceptable, and simultaneously high enough empathy for in-group members that they are sufficiently motivated to withhold help or condone harm on their behalf.

Empathy is one of a number of emotions that are shaped by social identities. For example, moral emotions, such as guilt and shame in response to in-group transgressions, have been shown to drive intergroup forgiveness and foster reconciliation, and positive emotions (e.g., hope) have been shown to similarly ease intergroup aggression (Cohen-Chen, Halperin, Crisp, & Gross, 2013). On the other hand, negative emotions such as anger and hatred drive conflict and prevent reconciliation (in some circumstances; e.g., Halperin, 2008). Obviously, the internal forces that ultimately pull someone toward and away from intergroup conflict are numerous, and mapping the unique contributions of each of these forces will be an important undertaking.

One avenue for future research will be to determine which intergroup outcomes are best predicted by each intergroup emotion. Here, we focused on outcomes that we hypothesized would be most directly affected by intergroup empathy: intergroup altruism and passive harm. On the other hand, previous research has shown that the approach motivated emotion of anger effectively drives support for (or engagement in) more active intergroup harm (Lickel et al., 2006; Spanovic, Lickel, Denson, & Petrovic, 2010). Future studies could test whether increasing moral emotions like shame may foster more indirect, symbolic gestures to the aggrieved out-group, such as a formal apology or an offer of reparations, whereas decreasing parochial empathy may be more likely to foster prosocial interpersonal gestures, such as helping an out-group target. At the same time, future work could also examine how intergroup emotions interact with each other to drive conflict. For example, it would be interesting to determine the potential relationship between in-group empathy, intergroup anger, and active harm of the out-group. This approach could connect the work on parochial empathy, examined here, with related research on “empathic anger” (Vitaglione & Barnett, 2003).

Finally, it is important to note some limitations of the current research. Although we replicated the main effects across a range of cultural contexts, the empathy measures were based on self-report and therefore subject to demand characteristics. Future performance-based or neuroimaging measures could be used to circumvent this issue. Additionally, the measures used in the current research were constrained to one facet of intergroup empathy (i.e., EC). Future research should also examine different forms of intergroup empathy. For example, personal distress (the amount of anxiety or distress one feels in the presence of others’ suffering) is another type of empathy that could be examined in the intergroup contexts. Personal distress has generally been found to be associated with avoidance rather than helping when avoidance is an option (Batson, Fultz, & Schoenrade, 1987). However, if the distressing situation cannot be avoided, personal distress has been shown to predict altruism even better than EC (Batson et al., 1981). It would therefore be interesting to determine the degree to which intergroup empathy as personal distress motivates behavior in the context of intergroup conflict, particularly in protracted conflicts that that may feel inescapable.

Conclusion

Across three distinct intergroup contexts, we show here that the distribution of empathy between in-group and out-group members (i.e., parochial empathy) mediates the effect of “parochial identification” on outcome measures and predicts outcomes independent of (and better than) trait EC. A major implication of this work is that conflict resolution interventions or programs aimed at increasing overall empathy in members of partisan groups, for example by taking others’ perspectives while reading literary fiction (Bal & Veltkamp, 2013; Kidd & Castano, 2013), may have little or no effect on increasing intergroup harmony or decreasing the likelihood of intergroup violence (Zaki & Cikara, 2015). Future work should therefore determine which interventions affect parochial empathy, separately from trait EC.