Abstract
Four studies investigate how perceptions that one’s social group has been victimized in society—that is, perceived group victimhood (PGV)—influence intergroup trust. Jewish and politically conservative participants played an economic trust game ostensibly with “partners” from their ingroup and/or a salient outgroup. Across studies, participants dispositionally or primed to be high in PGV revealed greater trust behavior with ingroup than outgroup partners. Control participants and those dispositionally low in PGV did not display such bias. Study 3 revealed, moreover, that high PGV enhanced ingroup trust even after an overt betrayal by an ingroup partner. Results were not explained by fluctuations in group identification, highlighting the novel, independent role of PGV in shaping an important aspect of intergroup relations—that is, trust. Implications of PGV for intergroup relations are discussed.
In 2009, Bernie Madoff pleaded guilty to bilking investors out of an unprecedented US$65 billion in a massive Ponzi scheme. Although less publicized, Madoff’s success was in part the result of an affinity fraud—the exploitation of one’s own social group membership to gain fellow members’ trust. Of the thousands of people that Madoff scammed, a significant portion were Jewish, like Madoff himself, and were frequently recruited from social clubs originally founded as safe havens from anti-Semitism (Leamer, 2008). Perhaps this history of group victimization and shared group membership allowed Madoff to gain and maintain the trust of so many, even when signs pointed to trouble. The present work investigates this possibility. Specifically, does perceived group victimhood (PGV) enable an unearned and particularly resilient form of ingroup trust?
PGV and Intergroup Trust
Trust is an inherently risky undertaking. It involves making oneself vulnerable to another, in pursuit of some important outcome, despite being unable to control that other party (Mayer, Davis, & Schoorman, 1995). Given the inherent vulnerability associated with trusting, it is not surprising that a variety of self-report, implicit, and behavioral measures reveal that people typically trust ingroup members more than outgroup members (Brewer & Kramer, 1986; De Dreu et al., 2010; Insko & Schopler, 1998).
Members of groups that have suffered from discrimination or intergroup violence may be particularly likely to display such biases, even when the victimization seems distant from the individual. Discrimination or oppression can lead individuals to develop a deep sense of victimhood (Shaw, 2003), even if they were not directly affected by intergroup conflict or generations have passed since the event. For example, American Indians may relate current personal victimization to the historical trauma suffered by their ancestors, and even members of tribes whose history did not include the practice of forced boarding schools may still identify with this particular form of victimization (see Evans-Campbell, 2008). In addition, PGV can be “passed down” to future generations through narratives emphasizing injustices the ingroup has suffered (Noor, Brown, & Prentice, 2008), promoting potentially chronic states of threat. PGV can also be induced situationally through reminders of past victimization or through social identity threat (Sullivan, Landau, Branscombe, & Rothschild, 2012; Wohl & Branscombe, 2008). Historically dominant groups can also experience PGV. Men accused of harming women (a moral identity threat) are more likely to see men as victims of sexism and claim such victimhood is more prevalent than discrimination against women (Sullivan et al., 2012). This “competitive victimhood” in dominant groups highlights how easily PGV can arise.
Although PGV may reduce the trust granted to members of the offending outgroup, those who view the ingroup as highly victimized may also strategically show enhanced trust in fellow ingroup members (Noor et al., 2008). Research has found, for instance, that members of culturally devalued groups often engage in ingroup bonding to cope with discrimination (Jetten, Branscombe, Schmitt, & Spears, 2001; Schmitt & Branscombe, 2002). When the power, status, or success of the ingroup is at stake, attempts are made to increase group solidarity (Grant & Brown, 1995; Hogg, 1992). Ingroup homogeneity is pursued to buffer positive social identity for the group and in preparation for mounting a defense of the ingroup (Turner, Hogg, Turner, & Smith, 1984; Wohl, Branscombe, & Reysen, 2010). Because discrimination and (sometimes violent) victimization is a clear threat to the ingroup in this regard, the resulting increase in group cohesion may lead to interesting intragroup trust behavior. Specifically, perceived group victimization might lead individuals to be particularly concerned with their relations with fellow ingroup members and, potentially, particularly trusting of fellow ingroup members even when such trust is not warranted.
Consistent with this possibility, perceived threats to one’s group have been found to increase individuals’ desire to strengthen their ingroups. Reminders of past victimization can lead to individuals defensively legitimizing ingroup actions, even when such actions are harmful to another group. For example, reminders of the 1941 Japanese attack on Pearl Harbor caused American participants to feel less collective guilt regarding American offenses in Iraq (Wohl & Branscombe, 2008). Perceived ingroup victimization also promotes direct intentions to aid the ingroup. Specifically, in Wohl and colleagues’ (2010) research, Jewish individuals prompted to think about the Holocaust reported greater willingness to perform behaviors that enhance the ingroup’s vitality (e.g., donating to Jewish organizations, marrying someone Jewish) than participants who thought about being a member of the Jewish community. Although insightful, these studies did not examine the effect of perceived group threat on actual behavior or on trust specifically—perhaps a particularly important ingroup-strengthening behavior. Furthermore, Wohl et al. did not examine whether these ingroup-strengthening motivations come at any expense to the outgroup, or, rather, whether ingroup-relevant threats primarily shape relations with fellow ingroup members. Most importantly, although the Wohl et al. studies suggest that threats to the ingroup promote pro-ingroup intentions, they do not touch on whether or not such intentions are observed if the actions come at a cost to the self.
Indeed, certain conditions can cause people to value helping the group over personal gain (Barreto & Ellemers, 2000)—This principle can even go as far as sacrificing one’s life to protect the group or its individual members (Swann, Gomez, Seyle, Morales, & Huici, 2009). It is possible that PGV is one such factor, inspiring both a high level of cohesion and solidarity (Grant & Brown, 1995; Hogg, 1992) and promoting a desire to act in the interest of the group over the self. Thus, individuals who perceive the group as highly victimized may commit themselves to aiding fellow ingroup members, behaving as though another ingroup member is trustworthy regardless of the other’s actual behavior. In other words, PGV may “blind” individuals to ingroup members’ trustworthiness and evoke increased trust behavior even when ingroup members behave treacherously. Consequently, the present research examines the implications of PGV for trust in ingroup compared with outgroup members. We predicted that PGV engenders enhanced trust in ingroup members that, in turn, results in a particularly resilient intergroup trust bias.
Present Research
Four studies investigate how PGV affects trust behavior toward ingroup, compared with outgroup, members. Using modified versions of the “investment game” (Berg, Dickhaut, & McCabe, 1995), Jewish and politically conservative participants decided whether to trust ingroup and/or outgroup partners with monetary endowments. We predicted that participants primed to be high in PGV, or those dispositionally high, would display a greater ingroup-favoring trust bias than participants not so primed. Study 3 pursued potential boundaries of this ingroup-favoring trust bias by examining trust after an outright betrayal by an ingroup member. We predicted that participants high in PGV would demonstrate enhanced trust in ingroup members even after a blatant betrayal.
Study 1a
Study 1a aimed to examine whether minority group members’ prior perceptions of the ingroup’s victimhood (PGV) predict trust behavior toward ingroup and outgroup members. We explored these processes among Jewish participants, with the majority Christian community as the relevant outgroup. We predicted that participants high in PGV would reveal a greater ingroup-favoring bias relative to those low in PGV.
Participants
Twenty-four male, Jewish participants were selected from an introductory psychology course. During a mass-testing session held at the beginning of the academic quarter, participants rated their perceptions of ingroup victimization on two items (“I believe that Jews have been victimized, discriminated against, or treated unfairly” and “I believe that Jews have been victimized, discriminated against, or treated unfairly to a greater extent than other groups”) using a 7-point Likert-type scale. Participants were selected from those scoring in the highest and lowest quartiles on PGV. 1
Materials
Explicit outgroup trust
To determine whether PGV is related to explicit beliefs about the trustworthiness of outgroup members, and whether such beliefs predict actual trust behavior, participants reported their level of outgroup trust during a pretesting session. Explicit outgroup trust was measured with six items from Noor et al.’s (2008) outgroup trust scale, such as “Most Christian people try to be fair” (reverse scored), rated on a 1 (strongly disagree) to 7 (strongly agree) Likert-type scale. Items were averaged to form a composite measure of explicit outgroup trust (α = .62).
Group identification
To examine whether any effects of PGV that emerge are independent of group identification, we used the identity and private subscales (4 items each) of the Collective Self-Esteem Scale (CSES; Luhtanen & Crocker, 1992). Items such as “Being Jewish is an important reflection of who I am” and “In general, I’m glad to be Jewish” were rated on 1 (strongly disagree) to 7 (strongly agree) Likert-type scales.
Behavioral trust task
Trust was measured using a modification of Delgado, Frank, and Phelps’s (2005) version of the investment game (Berg et al., 1995). 2 Participants were motivated to earn “money” that would be converted into lottery entries for a gift certificate. The task involved one ingroup (i.e., Jewish), one outgroup (i.e., Christian), and one filler player of unspecified background—each actually controlled by a computer program. In each trial, participants saw one player’s name and received an endowment of US$1 that they could either keep or invest in that player. If the money was invested, the amount was tripled and the partner received US$3. The partner could then decide either to keep the entire US$3 (behave untrustworthily—that is, defect) or return US$1.50 (behave trustworthily—that is, cooperate). Investment, therefore, reflects participants’ trust that their partner will cooperate.
Participants played 72 trials (24 per partner) presented in 6 blocks of 12 trials each, on a computer. Trials were presented randomly across blocks so that participants could not predict which partner they would play in any given trial. Participants made their decisions, and after a brief delay, received feedback on the trial’s outcome (e.g., “[Your partner] has decided to share/keep the money”). Importantly, partners were programmed to behave equally, albeit only moderately, trustworthily. Specifically, they reciprocated by returning money on half of the trials in which participants invested the endowment (i.e., 50% reinforcement rate).
Procedure
Several weeks after completing the PGV items and explicit outgroup trust measure, participants arrived at the laboratory and were informed they were going to play an economic decision making game with three “partners.” Participants then completed a demographics questionnaire to “introduce players to one another,” on which they reported their name and religious affiliation (among other filler items). They then viewed the same questionnaires ostensibly completed by their partners for the game. Specifically, they were introduced to “Adam Cohen” and “Christopher Schroeder,” who indicated that they were Jewish (ingroup) and Christian (outgroup), respectively, and to “Dan S.,” 3 who did not indicate his religion. “Dan” was included to obscure the study’s focus on ethnoreligious group membership. Hence, trials on the behavioral measure with Dan were not included in the main set of analyses, as we cannot interpret what religion participants’ believed Dan to be and whether this varied by PGV level (but see Note 6). After receiving this partner information, participants completed the behavioral trust game followed by the group identification items. Participants were then debriefed and credited for their participation.
Results
Explicit outgroup trust
Participants’ reported explicit outgroup trust was high, with an average score of 5.2 out of 7 (SD = 0.80). More importantly, explicit outgroup trust did not differ significantly as a function of PGV level, t(22) = 0.20, p = .84, nor did it predict participants’ actual trust behavior toward the outgroup partner during the investment game (r = .18, p = .40).
Group identification
To examine whether the effects of PGV on trust were merely the result of variations in level of group identification, we compared the identity and private CSES scores of high- and low-PGV participants. Results revealed no reliable effects on either subscale, ts < 0.97, all ps = ns. Moreover, participants’ continuous PGV scores were also not significantly correlated with their CSES (all rs = −.02, p = ns). Taken together, these results suggest that any effects of PGV on behavioral trust found in the present work are unlikely to be due to differences in participants’ levels of group identification.
Behavioral trust
Behavioral trust was assessed by the number of “share” (i.e., trust) decisions made for each target. Specifically, a 2 (PGV: low, high) × 2 (Partner Type: ingroup, outgroup) mixed-model ANOVA was conducted on the number of trust decisions participants made.4,5 Results revealed that the PGV × Partner Type interaction was significant, F(1, 22) = 5.37 p < .05, η2 = .1. 6 Although low-PGV participants did not differ in their trust of ingroup (M = 9.14, SD = 5.16) and outgroup partners (M = 9.21, SD = 5.41), t(13) = 0.05, p = ns, high-PGV participants made significantly more trust decisions with the ingroup partner (M = 11.50, SD = 7.12) than with the outgroup partner (M = 6.60, SD = 4.40), t(9) = 2.86, p = .019. Furthermore, although the pattern of means suggests that high-PGV participants tended to trust the ingroup partner more and the outgroup partner less than did low-PGV participants, neither difference was statistically reliable, ts < 1.26, ps > .05.
To ensure that these behavioral trust effects were not due to decisions made early in the game and without the benefit of observing the ingroup and outgroup partners’ equally trustworthy behavior, we compared the number of trust decisions for each partner in the first and last quarters of the game. There were no significant main effects or interactions of quarter, partner type, or level of PGV (all Fs < 2.6, all ps = ns). In other words, high-PGV participants continued to display ingroup favoritism in trust during the final quarter of investment game trials, despite the fact that the ingroup and outgroup partners had actually behaved equally trustworthily throughout the game.
Discussion
The results of Study 1a provide preliminary evidence that individuals with high levels of PGV trust ingroup members more than outgroup members. Although the relative roles of ingroup favoritism and outgroup derogation are unclear, this may be because PGV was measured as a chronic trait, and no actual information regarding group victimization was presented—an issue tested in Study 1b. It is also interesting to note that the high- and low-PGV participants did not differ in their self-reported trust of the outgroup, nor did explicit outgroup trust predict behavior during the game with the outgroup partner. These findings indicate that people may be unwilling to express a lack of trust toward outgroups, particularly one that is the dominant, majority in society (i.e., Christian majority). Importantly, group identification did not differ across PGV levels, suggesting that PGV is a separate construct from group identification and that our results are not merely due to victimhood’s relation with group identification. Taken together, the results of the present study provide initial evidence that perceiving one’s group as highly victimized is associated with an ingroup-favoring trust bias. In Study 1b, we explore the possibility that state, rather than chronic, PGV is similarly associated with intergroup trust bias.
Study 1b
The results of Study 1a suggest that the extent to which one perceives one’s social identity group as victimized in society (PGV) influences trust across ingroup–outgroup boundaries. The results of Study 1a are limited, however, due to the fact that PGV was measured as a chronic individual difference variable, and, consequently, it is not yet possible to know whether PGV actually causes the ingroup trust bias observed in Study 1a or, rather, whether some related, unmeasured third variable is really at play. Similarly, it is not clear from Study 1a whether situations that temporarily increase individuals’ perception that their group has been victimized have similar effects as chronic PGV on willingness to trust ingroup, compared with outgroup, members. To address these concerns, in Study 1b we manipulated rather than measured PGV. We expected participants primed to be high in PGV to reveal a greater ingroup-favoring trust bias than participants not so primed.
Participants
Thirty male, Jewish students participated in exchange for partial course credit or US$8.
Materials
PGV manipulation
PGV was manipulated via primes adapted from Voci (2006). Participants viewed histograms ostensibly representing responses to the question “What most comes to mind when thinking about the Jewish community?” The high-victimhood prime contained two statements invoking past Jewish victimization, specifically the Holocaust and European pogroms, embedded among four neutral statements, such as “Orthodox Jews keep kosher.” The control prime consisted of the same four neutral statements as well as one negative, nondiscriminatory, statement regarding the likelihood of having a child with Tay-Sachs—a fatal disease more common among Ashkenazi Jews. This statement was included to ensure that both conditions presented negative, group-relevant information.
Behavioral trust task
Behavioral trust was again measured using a modified investment game (Berg et al., 1995; Delgado et al., 2005). Again, participants played 72 trials (24 per partner) presented in 6 blocks of 12 trials each. Unlike in Study 1a, each block contained 4 trials per partner (to minimize chances that a large number of trials could occur without the presentation of any particular partner), and trials were presented randomly within blocks.
Estimates of partner reciprocation frequency
In Study 1a, we did not assess whether high- and low-PGV participants accurately recognized that each partner behaved equally trustworthily (50% reciprocation rate). Hence, it is possible that high-PGV individuals pay less attention to the behavior of the outgroup and/or ingroup partner, resulting in their persistent ingroup favoritism. To examine this possibility, participants indicated on a 0% to 100% scale the frequency with which each partner returned money.
Group identification
Group identification was again measured using the identity and private CSES subscales (Luhtanen & Crocker, 1992).
PGV manipulation check
Participants rated their perceptions of group victimization on the two items used to select participants based on PGV level in Study 1a (i.e., “I believe that Jews have been victimized, discriminated against, or treated unfairly” and “I believe that Jews have been victimized . . . to a greater extent than other groups.”).
Procedure
The procedure for Study 1b was the same as in Study 1a, except that PGV was primed before participants completed the behavioral trust game. Participants were first asked to help provide feedback on another study and randomly assigned to view either the high-PGV or control condition prime. Participants rated whether the information was accurate and familiar to them on 1 (completely unfamiliar/ inaccurate) to 7 (completely familiar/accurate) Likert-type scales. Participants then completed the demographics questionnaire, reviewed the questionnaires ostensibly completed by the partners, and then completed the behavioral trust game. After, participants reported their estimates of partner reciprocation frequency and completed the group identification and manipulation check measures. Participants were then debriefed and credited for their participation.
Results
PGV manipulation check
Responses to the PGV items (α = .72) revealed that participants primed to be high in PGV did indeed perceive the ingroup as significantly more victimized (M = 5.86, SD = 1.10) than participants exposed to the control prime (M = 4.38, SD = 1.59), t(28) = 2.01, p < .05, η2 = .07.
Group identification
An independent-samples t test performed on participants’ ratings on the CSES’s private subscale (α = .78) revealed no significant differences as a function of PGV condition, t(28) = 1.16, p > .25. There was, however, a marginal trend for control participants (M = 4.21, SD = 0.73) to report higher identity CSES (α = .89) than high-PGV participants (M = 3.81, SD = 0.38), t(28) = 1.85, p > .06. Given that this trend should counteract the predicted effect of PGV on intergroup trust, any evidence suggesting that PGV leads to greater intergroup trust bias is not easily attributable to it.
Behavioral trust
The total number of investments made was, again, used as our behavioral index of trust. The number of “share” (i.e., trust) decisions for each partner type was subjected to a 2 (Victimhood Prime: control, high) × 2 (Partner Type: ingroup, outgroup) mixed-model ANOVA. Results revealed the predicted interaction between victimhood condition and partner type, F(1, 28) = 4.26, p < .05, η2 = .13. As shown in Figure 1, trust in the control condition did not differ between partners, t(13) = 0.89, p = ns. High-victimhood condition participants, however, made more trust decisions with the ingroup partner than with the outgroup partner, t(15) = 2.09, p = .05. In addition, high-victimhood condition participants displayed more trust in the ingroup partner than did control participants, t(28) = 3.42, p < .05, but trust in the outgroup partner did not differ, t(28) = 0.48, p = .64. 7
Trust decisions by partner type and condition
Estimates of reciprocation frequency
Estimated reciprocation frequencies were subjected to the same 2 × 2 mixed-model ANOVA. Participants perceived that partners reciprocated to an equal extent (respective Ms = 35.59% and 41.16%), irrespective of prime, partner type, or their interaction (all Fs < 2.05). Hence, the ingroup favoritism displayed by participants in the high-victimhood condition persisted despite awareness that partners behaved equally trustworthily.
Discussion
Consistent with the results from Study 1a and our predictions, participants primed to be high in PGV trusted ingroup members more than outgroup members. In other words, higher PGV resulted in intergroup trust biases regardless of whether high PGV was a preexisting, measured dispositional variable or was experimentally induced in the lab. Specifically, participants who perceived the ingroup as highly victimized demonstrated a persistently higher level of trust toward ingroup than outgroup members. In contrast, those either dispositionally low in PGV or those who received the control prime (making them relatively lower in PGV compared with the experimental condition) showed no such bias. Such results appeared even as participants’ Jewish identity (and negative, group-relevant information) was made salient in both conditions.
The results of Study 1b are also intriguing in that they offered clear evidence that heightened PGV appears to have increased participants’ ingroup favoritism, rather than their outgroup derogation. In other words, high-PGV participants in Study 1b revealed greater ingroup trust, but not less outgroup trust, compared with control condition participants. It is possible that the victimization threat presented to participants in the high-PGV condition may have caused them to seek out bonds with fellow ingroup members (see Jetten et al., 2001; Schmitt & Branscombe, 2002) as a coping mechanism. A strong, stable group is better equipped to provide such comfort, and increasing one’s trust in the ingroup is certainly a reasonable ingroup-strengthening behavior (Wohl et al., 2010).
Although we expected to observe an intergroup trust bias among high-PGV participants, the robustness of the bias revealed in the present work is nevertheless somewhat surprising. Specifically, high-PGV participants continued to display ingroup favoritism at the end of the behavioral trust task in both Studies 1a and 1b, despite both ingroup and outgroup partners being programmed to “behave” equally trustworthily (reciprocating trust 50% of the time). Recall, furthermore, that both high- and low-PGV participants correctly reported the partners’ equal rates of reciprocation. Hence, high-PGV participants’ quite persistent trust in the ingroup does not appear to be due to their inattention to their partners’ behaviors. Indeed, this awareness makes the results for high-PGV participants particularly striking; despite being aware that ingroup and outgroup partners were behaving equally trustworthily, high-PGV participants continued to favor the ingroup partner.
Much like these findings regarding participants’ awareness of the partners’ behavior, the present work provides compelling evidence that differential group identification among high- and low-PGV participants does not account for the observed patterns of intergroup trust. Indeed, group identification did not vary among high- and low-PGV participants whether measured (Study 1a) or manipulated (Study 1b). That said, perhaps the lack of correlation in Study 1b should not be surprising given that group identity was made salient in both the high-PGV prime and the control prime conditions. Nevertheless, the group identification findings (or lack thereof) suggest that PGV is a construct that is distinguishable from group identification.
Taken together, therefore, the present results suggest that perceiving one’s group as highly victimized in society engenders a particularly robust and resilient bias to trust in ingroup members. Before exploring potential boundary conditions of this bias, it is important to consider the generalizability of the effect to different types of societal groups. Specifically, it is possible that the desire to bond with and strengthen a victimized ingroup may only occur for members of groups with historical precedents of actual widespread group danger and marginalization. If so, we may not expect people to display this kind of enhanced favoritism for groups that feel threatened or marginalized only in relatively local, circumscribed contexts or for which intergroup violence is highly unlikely and/or unprecedented. However, previous research indicates that reminding dominant groups of rare instances of victimhood can lead to the same ingroup-favoring responses as for historically devalued groups. Specifically, Jews reminded of the Holocaust and Americans reminded of the Pearl Harbor or 9/11 attacks were both more likely to legitimize current ingroup offenses against current adversaries (Wohl & Branscombe, 2008). Study 2 examines the possibility that the mere perception of group victimization is sufficient to increase ingroup trust, considering the generalizability of PGV’s effects on intergroup trust among groups that have not traditionally been marginalized or socially oppressed.
Study 2
Although Studies 1a and 1b provided initial evidence of the role of PGV in shaping intergroup trust, we examined a participant sample with a long history of victimization, and, thus, it is unclear whether the effects generalize to groups with more recent and/or limited forms of marginalization. In addition, it is possible that the fact that the behavioral trust measure requires monetary investments may have influenced the results. Due to negative cultural stereotypes regarding Jewish individuals and money, it is possible that concerns other than trust may have been activated, affecting participants’ behavior with ingroup and/or outgroup partners. Consequently, Study 2 examines the effect of PGV on intergroup trust among a population not traditionally viewed as victimized, but, nevertheless, like all groups, is susceptible to feelings of group victimization in certain contexts such as in the climate of a politically liberal university. Specifically, we recruited politically conservative students, a minority sociocultural identity on campus, to complete a modified version of the investment game with either an ingroup (conservative) or outgroup (liberal) partner. We again predicted that high-PGV participants would demonstrate greater ingroup favoritism than control participants.
Participants
Fifty-eight (24 female) politically conservative students participated in exchange for partial course credit or US$8. Participants were selected from among those who, during a mass-testing session several weeks prior, marked a “6” or “7” on a Likert-type agreement scale of their endorsement of “politically conservative ideology” and selected a “1” or “2” on a similar scale of their endorsement of “politically liberal ideology.”
Materials
PGV manipulation
PGV was primed as part of an ostensibly separate study on memory. Participants read two filler articles before the PGV prime, each followed by recall questions. The control article described a social event hosted by a politically conservative campus group. The high-victimhood article described how polls showed fewer people openly identifying as conservatives amid growing public hostility toward conservatives.
Behavioral trust task
Participants played a single-round version of the investment game (Berg et al., 1995) with either an ingroup or outgroup “partner,” although they were led to expect multiple rounds. Participants were given a US$10 endowment of which they could invest any portion in their partner. As in Studies 1a and 1b, investments were tripled before the partner received it, and the partner could return any amount to the participant. Rather than a dichotomous measure, therefore, trust was operationalized as the amount of money invested. Again, “money” earned was converted into lottery entries.
Group identification
Group identification was again measured using the identity and private CSES subscales (Luhtanen & Crocker, 1992), with “conservatives” as the relevant group.
PGV manipulation check
Participants rated their perceptions of group victimization on the two items from Study 1b (with reference to conservatives) and one item on perceptions of group discrimination in the media, on 7-point Likert-type scales.
Procedure
Participants first completed the ostensibly separate “memory study,” through which the PGV prime was administered. They were next randomly assigned to play the investment game with either an ingroup (conservative) or outgroup (liberal) partner. Similar to the procedures outlined in Studies 1a and 1b, partner ingroup/outgroup status was communicated via exchanged demographic questionnaires that included ratings of political ideology. In addition, the partner demographic questionnaires (manipulated by the experimenters) included an open-ended question about campus activities in which the partner described joining either a politically conservative or a liberal group. After the game, participants completed the group identification and manipulation check measures, then were debriefed and credited for their participation.
Results and Discussion
PGV manipulation check
PGV item ratings (α = .76) were averaged and subjected to a 2 (Victimhood Prime: control, high) × 2 (Partner Type: ingroup, outgroup) ANOVA. Only the main effect of PGV condition was reliable, F(1, 54) = 5.39, p < .05, η2 = .087. High-PGV prime participants (M = 5.05, SD = 1.21) perceived conservatives as significantly more victimized in society than did control prime participants (M = 4.28, SD = 1.33).
Group identification
Ratings on the CSES identity and private subscales (αs > .61) were subjected to the same 2 × 2 ANOVA. No main effects or interactions were reliable (all Fs < 3.0), suggesting group identification was equally salient in both conditions.
Behavioral trust
The amount of money participants invested (in dollars) was subjected to a 2 (Victimhood Prime: control, high) × 2 (Partner Type: ingroup, outgroup) ANOVA. 8 Results revealed the predicted interaction between victimhood condition and partner type, F(1, 54) = 5.59, p < .05, η2 = .09 (see Figure 2). Participants in the control condition did not differentially invest in (i.e., trust) ingroup and outgroup partners, t(30) = 0.77, p = ns. However, high-victimhood condition participants invested more with the ingroup than outgroup partner, t(24) = 2.12, p = .05. In addition, high-victimhood condition participants invested more with the ingroup partner than did control participants, t(28) = 2.08, p < .05, but investments with the outgroup partner did not differ across conditions, t(26) = 0.61, p = ns.
Investment by partner type and condition
The present results, therefore, offer compelling evidence that PGV promotes intergroup bias by increasing ingroup favoritism and that this effect is not limited to sociocultural groups with long histories of marginalization. For the conservative students in Study 2, their perceived victimization may extend beyond campus, but it is likely more salient due to the campus demographics being primarily liberal and causing participants (perhaps for the first time) to feel like members of a minority group. Thus, the present results suggest that perceptions of group victimization motivate such ingroup-strengthening behavior.
Study 3
Using two different participant samples and versions of the investment game, Studies 1a, 1b, and 2 revealed that PGV leads to enhanced trust in ingroup members. However, PGV may only cause people to give ingroup members the benefit of the doubt initially or when behavior is somewhat ambiguous, but not in cases of more extreme behavior. In Studies 1a and 1b, the cost of partner betrayal was relatively low (a loss of US$0.50 compared with the decision not to trust), and a binary option, perhaps making these “betrayals” less obvious or damaging, whereas in Study 2 participants never actually learn whether or not their enhanced trust in ingroup members was well placed. It is possible that more overt or costly betrayals would mitigate the impact of PGV. Hence, Study 3 considers the potential limits of the ingroup-boosting effect of PGV. Specifically, Study 3 examines whether PGV predicts ingroup loyalty even in the face of overt betrayal of trust.
The behavioral manifestation of trust after betrayal is loyalty—continuing to trust even when given a signal that such behavior may not be warranted. Research suggests that PGV inspires beliefs that ingroup members should practice loyalty. Indeed, threats to the ingroup have been found to increase demands and expectations that members practice group loyalty (Branscombe, Wann, Noel, & Coleman, 1993). Branscombe and colleagues (1993) found that when sports fans thought their team was on a losing streak (a value threat), they evaluated a loyal fan more warmly than a disloyal fan, but not if the team was winning. Acute threats also increase actual ingroup loyalty (Stern, 1995; Van Vugt & Hart, 2004), suggesting that PGV might increase expectations of loyalty from and the practice of loyalty toward ingroup members.
If individuals expect ingroup loyalty, betrayal should be perceived as a violation of group norms. Research suggests that misbehaving and nonprototypical ingroup members are evaluated and dealt with more harshly than outgroup members—a phenomenon known as the “black sheep effect” (Marques, Yzerbyt, & Leyens, 1988). If ingroup members betray trust and are perceived as “black sheep,” then other group members should withdraw their trust. Indeed, betrayal may lead those high in PGV to reduce their trust in ingroup members more than in outgroup members. However, not all deviant ingroup members or contexts elicit “black sheep” derogation (Morton, Postmes, & Jetten, 2007; Pinto, Marques, Levine, & Abrams, 2010). Individuals strategically forgo the black sheep effect if doing so benefits their group (Morton et al., 2007)—essentially, the black sheep effect may not occur if the relative position of the group is at stake. Because PGV is fundamentally a concern about group standing, those high in PGV may turn a blind eye to betrayal by ingroup members, thus maintaining ingroup trust and, by extension, intergroup trust biases. Study 3 examined these competing predictions.
Participants
Sixty Jewish students (42 female 9 ) participated in exchange for partial course credit.
Materials
PGV manipulation
PGV was primed in a manner similar to that described in Study 2. The control prime was an article that described a social event, unrelated to Jewish stigmatization, hosted by the campus Jewish group. The high-victimhood prime was an article that described the appointment of a Holocaust survivor’s daughter to an anti-Semitism taskforce.
Behavioral trust task
Participants were randomly assigned to play the investment game with either the ingroup or outgroup “partner” from Studies 1a and 1b. During each of 11 trials, participants could invest any portion of a US$10 endowment, which was subsequently tripled. The partner could return any amount to the participant, after which each person’s earnings were displayed. The game was programmed such that the partner behaved extremely trustworthily through the first 4 trials, returning amounts that ensured equitable payouts. In the fifth trial, however, he defected with the entire investment—betraying participants’ trust—before resuming equitable divisions for the remaining trials (see Molden, Olson, & Lucas, 2011). Trust was operationalized as money invested per trial. Again, earnings were converted into lottery entries.
Group identification and manipulation check
Group identification and PGV were assessed as in Study 1b.
Procedure
The procedure was the same as in Study 2, although partner ingroup/outgroup status was communicated via the same questionnaire exchange described in Study 1 (introducing either the ingroup or outgroup partner), and the behavioral trust game included 11 rounds rather than just 1.
Results
PGV manipulation check
Ratings on the PGV items (α = .67) were averaged and subjected to a 2 (Victimhood Prime: control, high) × 2 (Partner Type: ingroup, outgroup) ANOVA. Only the main effect of PGV condition was reliable, F(1, 52) = 5.74, p < .05, η2 = .09. High-PGV participants (M = 6.30, SD = 0.99) did indeed perceive the ingroup as significantly more victimized than did control participants (M = 5.58, SD = 1.15).
Group identification
Ratings on the CSES identity and private subscales (αs > .85) were subjected to the same 2 × 2 ANOVA. No main effects or interactions were reliable (all Fs < 2.3), again suggesting group identity was equally activated in both prime conditions.
Behavioral trust
A 2 (Victimhood Prime: control, high) × 2 (Partner Type: ingroup, outgroup) × 11 (Trial) mixed-model ANOVA was conducted on investment amount (in dollars). No main effects or two-way interactions were significant (all Fs < 1), except for the main effect of trial, F(10, 560) = 10.76, p < .001,η2 = .15. On average, the amount of money participants invested increased over time, dropped immediately after the betrayal, and gradually rebounded as fair play resumed (see Figure 3). Importantly, the predicted omnibus three-way interaction was significant, F(10, 560) = 1.95, p < .05, η2 = .03. To consider our a priori predictions, we further inspected trials of primary interest—(a) the initial trial and (b) the trials just prior to and after betrayal (Trials 5 and 6).
Investments across trials
Initial trust
A 2 (Victimhood Prime: control, high PGV) × 2(Partner Type: ingroup, outgroup) ANOVA on investments in Trial 1 revealed only the predicted interaction between victimhood prime and partner type, F(1, 56) = 5.70, p < .05, η2 = .09. Replicating the results of Studies 1 and 2, high-victimhood condition participants invested more with (i.e., trusted) ingroup partners (M = 7.40, SD = 0.69) than with outgroup partners (M = 5.25, SD = 0.77), t(25) = 1.98, p = .058, but not control participants (ingroup: M = 4.67, SD = 0.69; outgroup: M = 5.83, SD = 0.63), t(31) = 1.3, p = ns. High-PGV condition participants, furthermore, trusted ingroup partners more than did control condition participants, t(28) = 2.65, p < .05. Trust in outgroup partners, however, did not differ by PGV condition, t(28) = 0.63, p = ns.
Betrayal-related trust
Investments just before and after betrayal were subjected to a 2 (Victimhood Prime: control, high) × 2 (Partner Type: ingroup, outgroup) × 2 (Trial: 5, 6) mixed-model ANOVA. Results revealed a main effect of trial, F(1, 56) = 20.97, p < .01, η2 = .25, modified by a significant three-way interaction, F(1, 56) = 5.86, p < .05, η2 = .07 (see Figure 4). To decompose this complex interaction, we first examined changes in investments as a function of PGV condition. A 2 (Partner Type: ingroup, outgroup) × 2 (Trial: 5, 6) mixed-model ANOVA on trust pre- and postbetrayal among control participants revealed only a significant interaction, F(1, 31) = 5.21, p < .05. Consistent with the “black sheep effect,” control participants decreased their investments with the ingroup partner following betrayal, t(14) = 4.12, p < .01, but the trust afforded (i.e., investments in) the outgroup partner did not change, t(17) = 0.60, p = ns. Although the partner type by trial interaction was also reliable among high-PGV condition participants, F(1, 25) = 4.26, p < .05, the pattern of means was strikingly different. Specifically, high-PGV condition participants significantly reduced their trust toward outgroup members postbetrayal, t(11) = 3.10, p < .05, but the trust afforded ingroup partners did not change, t(14) = 1.5, p = ns. Taken together, these patterns suggest that high-PGV can attenuate and indeed reverse, the “black sheep effect” in response to ingroup betrayal.
Investments just prior to (Trial 5) and following betrayal (Trial 6), by partner type and condition
To examine the effect of PGV on postbetrayal trust further, we also decomposed the three-way interaction by examining trust behavior with ingroup versus outgroup partners, by high- versus low-PGV participants, separately during Trials 5 and 6. Neither PGV condition, nor partner type, nor their interaction significantly predicted investments during Trial 5 (all Fs < 2.07), suggesting that betrayal caused the trust differences revealed in Trial 6. Not surprisingly, a 2 × 2 ANOVA on investments in Trial 6 revealed a significant interaction between PGV and partner type, F(1, 56) = 5.11, p < .05, η2 = .083. Following betrayal, high-PGV participants trusted ingroup partners significantly more than outgroup partners, t(25) = 2.08, p < .05. Postbetrayal trust among control participants, however, did not differ across partners, t(31) = 1.38, p = ns, although the pattern of means suggested a trend toward outgroup favoritism. Taken together, these analyses suggest that betrayal appears to have restored the effect of PGV on intergroup trust revealed in participants’ initial trust behavior and in Studies 1 and 2.
Discussion
Study 3 offers striking evidence that PGV causes an enhanced and resilient form of ingroup trust. Replicating Studies 1 and 2, high-PGV participants initially exhibited ingroup favoritism, but not outgroup derogation, whereas control participants expressed no bias in their trust behavior. However, Study 3 also demonstrated that people are not immune to learning about others’ trustworthiness. Throughout the game’s early trials, the partners displayed extremely high levels of trustworthiness, behaving unequivocally fairly at first, unlike the questionably trustworthy behavior of the partners in Studies 1a and 1b (50% reciprocation rate). In the present study, therefore, the reliable trustworthiness of participants’ partners eventually elicited the trust of the participants, who by the fifth round of the game displayed no differences in investments irrespective of their level of PGV or partner group membership—that is, ingroup and outgroup partners were being trusted equally by high-PGV and control participants. Thus, the results of Study 3 reveal that participants are sensitive to overtly positive, trustworthy behavior regardless of their PGV level or their partner’s group membership.
The development of trust, however, was not the only way in which participants in the present study demonstrated their sensitivity to partner behavior. The overt betrayal by their partner had a major effect on participant behavior, but in this case, PGV modulated such responses. Consistent with the “black sheep effect” (Marques et al., 1988), participants in the control condition punished ingroup betrayers more than outgroup betrayers. High-PGV condition participants, however, punished outgroup betrayers but not ingroup betrayers. In other words, high-PGV condition participants appear to have responded to ingroup members’ betrayals by remaining loyal themselves. Taken together, the present findings suggest that PGV may lead individuals to display ingroup favoritism not only in the absence of more diagnostic information about whom to trust but also even after overt betrayal by ingroup members.
General Discussion
Four studies explored the consequences of PGV for intergroup trust. The findings suggest that PGV engenders a level of ingroup trust that is particularly resilient. Indeed, high-PGV participants’ enhanced trust in ingroup members was strong enough to lead them to turn a blind eye to relatively large betrayals by a fellow ingroup member. These findings, therefore, replicate and extend recent work by Wohl and colleagues (2010) suggesting that perceived threats to a group’s existence motivate intentions to enhance the group’s vitality. Specifically, the present studies demonstrate that such intentions are borne out in actual behavior toward ingroup members, strong enough to withstand overt betrayal by ingroup members, and do not necessitate the derogation of outgroup members. Last, the present studies extend previous research by suggesting that group members who are high in PGV are unlikely to subject deviant ingroup members to black sheep derogation.
Interestingly, these effects were observed among members of a group with a long history of victimization as well as among member of a group that have not experienced widespread violence or victimization (due to the group membership) and, further, is a numerical minority only in a quite limited, local context. This suggests that some effects of group victimhood truly lie in perceptions rather than solely in demographics or historical patterns of group treatment. Furthermore, group identification was not related to PGV in any of our studies, suggesting that the observed effects of PGV on trust are not simply an artifact of heightened group identification.
The lack of correlation between PGV and group identification may be somewhat surprising, given Branscombe and colleagues’ rejection identification model (RIM) wherein perceived discrimination is expected to increase ingroup identification (Branscombe, Schmitt, & Harvey, 1999). However, the primes used to manipulate higher versus lower PGV included group-relevant information and, thus, are all likely to increase group identification and identity salience. Furthermore, several factors distinguish the current research from that examining the RIM. Specifically, the RIM predicts changes in identification following discrimination perceived as pervasive across time and contexts (Branscombe et al., 1999); in contrast, our studies focus on past discrimination (Studies 1 and 3) or discrimination in limited contexts (Study 2). In addition, the RIM primarily concerns individuals’ personal experiences of discrimination (Schmitt, Branscombe, & Postmes, 2003). Participants in the present work were not exposed to any direct discrimination (indeed, participants perceived that both ingroup and outgroup partners treated them similarly). Consequently, the procedures used in the present work allowed us to investigate the effects of PGV on trust without differentially increasing group identification across PGV conditions.
Limitations and Future Directions
Although the present work provides compelling evidence for the effect of PGV on intergroup trust, there are several limitations. Most importantly, we only explored trust behavior using an economic paradigm. The money at play in the present studies was a windfall and eventually converted into points for a lottery—It is possible that when betrayal of trust has greater consequences than losing a chance to win a gift certificate (such as entrusting one’s real property, or disclosing private and personal information), ingroup biases may be even stronger and, perhaps, observed even among participants with relatively low levels of PGV. Future studies should, therefore, investigate the role of PGV in how trust is formed and maintained across various contexts and stakes.
It is presently unclear how specific these effects of PGV are, relative to those triggered by the myriad of other types of group threat. Preliminary evidence suggests, however, that different types of group threat may not universally engender the ingroup-favoring trust biases observed in the present work. In Study 1b, both PGV conditions were exposed to at least some ingroup-threatening information. Specifically, participants in the control condition were reminded of Jewish individuals’ higher likelihood of the universally fatal Tay-Sachs disease, a threat of arguably greater relevance to our participants than the Holocaust or European pogroms (the high-PGV condition threats), because it is a current-day issue that could affect them personally. Nevertheless, only high-PGV condition participants displayed intergroup trust biases; thus, it is possible that PGV, but not all other types of group threat, leads to these ingroup-boosting behaviors. Again, future research is needed. Furthermore, although we found little evidence that group identification moderates the effects of PGV on intergroup trust, given the fairly small number of participants associated with these analyses, future research should investigate this possibility more thoroughly.
Finally, future research should investigate possible mediators of the relationship between PGV and intergroup trust. PGV may increase anxiety related to the ingroup’s jeopardized future, or collective angst, which has previously been linked to intentions to strengthen the ingroup (Wohl et al., 2010). PGV may also highlight ingroup disadvantages, promoting feelings of ingroup entitlement. Individual-level victimization can increase personal entitlement, which in turn increases the incidence of selfish behavior (Zitek, Jordan, Monin, & Leach, 2010). Perceived group victimization may similarly evoke group-level entitlement, causing people to protect or strengthen the ingroup by displaying increased trust and loyalty. Indeed, Wohl and Branscombe (2008) found that reminders of past ingroup victimization caused participants to experience less collective guilt in response to current ingroup harm-doing toward a third party—suggesting that perceived ingroup victimization results in a licensing of otherwise morally questionable behavior.
PGV and Outgroup Derogation
Intriguingly absent from the current studies is clear evidence of outgroup derogation. One explanation for this may be that the outgroup partners with whom participants interacted were not the direct perpetrators of the group threats. The ostensibly Christian collegiate in Studies 1b and 3 could not possibly have committed the acts in the primes, and in none of the studies did the outgroup partner make any suggestion that he held negative attitudes toward the participant’s ingroup. Furthermore, Study 1a’s results indicated that self-reported trust toward the majority, Christian outgroup was relatively high among the Jewish participants and did not differ by PGV level—suggesting that participants could differentiate between the bad acts of individuals and the trustworthiness of an entire group.
In addition, for members of numerical minority or socioculturally stigmatized groups, antagonizing members of dominant majority groups may not necessarily be a productive response. This is true both in general and, perhaps especially so, during times when the ingroup is vulnerable to being victimized. When PGV is high, perhaps the more adaptive response for lower power individuals is not to pull away from individual outgroup members but rather to turn toward ingroup members (Jetten et al., 2001; Schmitt & Branscombe, 2002) and support the strengthening of the ingroup as a whole (Wohl et al., 2010). Such an approach would help satisfy desires to help the ingroup without risking provoking conflict with more numerous and powerful outgroups.
This is not to say that outgroup derogation never occurs in response to PGV. Despite our surprise at not observing much evidence of outgroup derogation, high-PGV participants in Study 3 reacted swiftly to outgroup betrayal, punishing such betrayers with reduced trust. Ingroup trust and loyalty may simply be of paramount importance under conditions that trigger high levels of PGV in the absence of direct outgroup provocation. Consistent with this sentiment, work by De Dreu and colleagues (2010) found that the hormone oxytocin drives a “tend-and-defend” mentality toward the ingroup, increasing ingroup favoritism and defensive (but not offensive) aggression toward a competing outgroup. PGV may act similarly, primarily focusing individuals on ingroup strengthening rather than on outgroup aggression, unless directly provoked by outgroup members.
Conclusion
The present research reveals that perceiving one’s ingroup as highly victimized can lead to betrayals of trust or missed opportunities—ingroup members were favored by high-PGV participants, even when their behavior suggested they may not be trustworthy. The potentially devastating perils of such unwarranted trust are readily apparent in the Madoff saga; Madoff convinced his largely ingroup clientele to trust him because he was “one of them.” Making potential investors aware of potential susceptibility to this cruel twist on affinity schemes may help prevent members of groups victimized by society from being further victimized by devious ingroup members. Of course, most people are not operating affinity schemes, and in general, the increased trust from PGV is likely to promote prosocial actions that have beneficial effects for ingroup members such as recommending a group member for a position at one’s company. However, this may result in trustworthy outgroup members being unfairly passed over, possibly breeding feelings of hostility. As such, the consequences of PGV on trust might ultimately make group members vulnerable to future victimization by both outgroup and ingroup members.
Footnotes
Acknowledgements
We thank Jalisha Tolbert, Ida Samuel, and Michelle Rheinschmidt for assistance with data collection.
Author’s Note
Portions of this research were included in K.N.R.’s master’s thesis, and presented at the 2010 and 2011 meetings of the Society for Personality & Social Psychology and Midwestern Psychological Association.
Declaration of Conflicting Interests
The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.
Funding
The author(s) disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: This research was partially supported by NSF grant BCS-0843872 and MacArthur Foundation Fellowship awarded to J.A.R., Northwestern DRRC grant to M.G.B. and J.A.R., and NSF grants BCS-0720312 and BCS-0722326 to J.Y.C.