Collective Victimhood and Social Prejudice: A Post-Holocaust Theory of Anti-Semitism

Anti-Semitism represents one of the most penetrating forms of prejudice, yet social research has failed to address the causal underpinnings of the phenomenon. To this end, we put forward a new theory of anti-Semitism that builds on the legacy of the Holocaust. Standing as the benchmark for collective suffering, the Holocaust creates competition over recognition of the status of the victim. Upward comparisons between victimized in-groups with other victimized out-groups trigger social prejudice. Victimhood, thus, creates an antagonistic view of the Jews that, in turn, fuels anti-Semitic prejudice. We test this theory using data from Greece -- the European nation with the highest proportion of anti-Semites -- leveraging two survey experiments. Our results confirm the observational implications of the theory, while a natural experiment and a comparative analysis shed additional light on the causal mechanism. The findings of our research carry important implications for dealing with anti-Semitism and for combating various forms of out-group prejudice.

We show that in-group/out-group victimhood competition generates anti-Semitism and the high levels of anti-Semitism cannot be attributed exclusively to the country's long-standing financial crisis, but they are product of a widespread latent sense of collective victimhood, which precedes the crisis. Although most of our analyses focus on Greece, we also test the observable implications of our theory with comparative data.

Background
Research on anti-Semitism has identified a strong link between personality traits, ideology, and prejudice against the Jews. Individuals with more authoritarian attitudes appear to be more likely to hold anti-Semitic attitudes. Frindte, Wettig and Wammetsberger (2005), for example, find that latent anti-Semitism in Germany is strongly associated with authoritarianism (as measured by prominent authoritarianism scales). Similarly, Dunbar and Simonova (2003) show that rightwing authoritarianism predicts out-group bias and increases the propensity to express hostility and negativity against minorities. In a recent article, Charnysh (2015) shows that variation in Jewish population size in pre-Holocaust Poland predicts subnational variation in attitudes towards European integration. On this account, populist frames depicting the Jews as having an influence over the EU where disproportionately effective in those areas. 2 According to scholarly wisdom, the prevailing underlying mechanism behind these associations is scapegoating (Glick 2008). Collective frustration over economic outcomes creates a need to explain the world and to assign responsibility for those outcomes. During this process, members of the in-group target an out-group to assign blame for their distress.  found that relative economic deprivation (change in economic outlook) predicts higher levels of anti-Semitism in Poland, yet not in Ukraine (see Bilewicz and Krzeminski 2010). Gibson and Howard (2007) in Russia also failed to find supportive evidence for scapegoating theory. Descriptive plots of economic series and anti-Semitic opinions showed that, if anything, anti-Semitism surges in good economic times. What explains this inconsistency across relatively similar cases? We argue that at least one of the reasons behind this ambiguous evidence is that the theory is incomplete. Anti-Semitic attitudes do not only stem from a need to alleviate relative deprivation generated by economic misfortunes. Instead, they are the result of an underlying sense of collective victimhood.
Although the concept of victimhood is old, the literature on the topic is relatively new, widespread across subfields and rather vague about what victimhood means and what are its consequences (for a review see Vollhardt 2013). For example, in African countries colonialism has been the primary cause of a narrative of in-group victimhood due to arbitrary border drawing and the racialization of the various tribes by the colonizers (Kiwuwa 2012;Mamdani 2001;Vollhardt 2013). One immediate consequence of this sentiment of victimhood was the devastating war in the mid-1990s between Hutus and Tutsis in Rwanda, which ended with the genocide of the latter. Recent survey work in Rwanda, Burundi, the Democratic Republic of Congo and on the Crimean Tatars finds that exclusive victimhood successfully transmits traumas of previous generations to the next ones, prolonging negative intergroup attitudes (Lupu and Peisakhin Forthcoming;Vollhardt and Bilali 2015). Evidence from Kosovo suggests that one way to alleviate competitive victimhood is to establish extended out-group contacts and common in-group identifications (Andrighetto et al. 2012).
Such a task, however, remains particularly challenging. As Bar-Tal and Halperin (2013) put it: ". . . past events, especially chosen traumas, greatly contribute to the definition of group identity and are therefore maintained in the culture and transmitted to new generations, while also occupying a central place in the collective memory of a society involved in intractable conflict" (p. 936). Narratives of victimhood can be so pervasive that they also dominate elite strategies. Volkan (1998), for instance, eloquently describes the predominance of the 1389 battle of Kosovo in the 1994 peace negotiations during the Yugoslav civil conflict. The Serbs were focusing on what happened almost 700 years ago to highlight their victimhood and to explain their duty to protect their people.
Indeed, in cases of intractable conflict like the above (and others), victimhood is crucial to understand in-group/out-group relationships. Victimhood, however, is not necessarily conflict-specific; it plausibly stems from comparisons using the global context as the reference point (Bar-Tal and Antebi 1992). Competing for the severity of an in-group's trauma or level of victimization can be made about groups that have never been perpetrators or at conflict with the in-group (see also Noor et al. 2012;Vollhardt 2013). In such cases, members of the in-group keenly associate themselves with attitudes of deservingness and entitlement (Eidelson and Eidelson 2003). On this account, the in-group's victimhood is portrayed as unique, exclusive and incomparable.
Before moving to the theoretical argument, we need to underline how our theory differs from previous studies. There are, indeed, some correlational studies that find a link between collective victimhood and anti-Semitism. For example,  and Bilewicz and Stefaniak (2013) show that Polish respondents with high levels of perceived victimhood were more likely to report anti-Semitic stances. Whereas these studies lend some confidence to our theoretical motivation, they neither discuss nor test the theoretical underpinnings of the relationship. To be sure, victimhood is conceptually different in the above studies as they consider it a product of the

Romantic-Messianic view of Poland that created antagonistic views between Poles, Jews and also
Germans (see Krzemiński 1993Krzemiński , 2002. Our theoretical proposition is fundamentally different; it builds on the antagonistic views among in-groups towards out-groups with a well-acknowledged status of victim. The Jews and the historical legacy of the Holocaust serve as the benchmark of suffering and drive the competition of victimhood that, eventually, leads to anti-Semitism. In what follows, we set out our theoretical proposition.

Victimhood & anti-Semitism
In an article in Le Monde titled 'La pitié et la raison', Emmanuel Terray noted bitterly that since he and his ancestors have never been victims of any kind of violence he is at the bottom of social existence. This statement echoes the common wisdom among students of trauma, working both at the individual and the collective level. At the individual level, clinical psychologists have argued that the victim stance is a powerful one (Fassin and Rechtman 2009). Focusing on the construction of a collective consciousness, cultural sociologists have exemplified the societal implications of suffering (Eyerman et al. 2004;Moses 2011;Sztompka 2000). Since a given trauma becomes a solid component of individual and collective identity, it can also be socially instrumental (see Cole 2007). Victimhood solidifies identities, encourages national glorification and transforms traumas into a social process with exchangeable value.
Although instances of genocides and mass killings have been far from rare throughout history (Kiernan 2004), it was the WWII experience that attracted scholarly and institutional attention to the fate of victims. The Holocaust gradually became the premium historical analogy or the reference point to any potential victim of any given Genocide (Dean 2004(Dean , 2010. In so doing, it transformed into the archetypical and unequivocal paradigm of victimhood. Its legacy created universal moral, legal and cultural standards that postwar societies had to build their future upon (Alexander 2002).
The hypothesis that all people might become victims because of their ethnicity, race, gender, or sexuality-because of who they are-has created multiple layers of victimhood. For example, Dean (2004Dean ( , 2010 examines how the suffering and trauma in the United States and Western Europe have become central to identity, with victims competing for recognition. According to this view, popular attention to the Holocaust produces a seemingly contradictory set of outcomes. One the one hand, scholars of Genocides have noted that apart from international law the most important factor in preventing genocides is the mobilization of public opinion with the threat of Holocaust repetition (Power 2002). Therefore, transforming the Holocaust into a 'moral universal' can trigger sympathy and mobilize resources for the 'victim'. 3 On the other hand, it produces a "surfeit of Jewish memory" that obscures the suffering of other peoples (Dean 2006). Although seemingly competing, the two processes co-exist and generate implications for victimized collectives. 4 3 An example of the demand for self-interested appropriations of the Jewish experience is a series of Holocaust-victim hoaxes as the infamous Wilkomirski and Defonseca affairs; both 'survivors' were not even of Jewish origin and the author of the best-seller claiming she survived the Holocaust raised by the wolves apologized saying that "Ever since I can remember, I felt Jewish" (Maechler 2001;Vice 2014). Stein (1998) describes a similar trend of symbolic appropriation of the Holocaust among social movements, which extensively employs the "Holocaust" metaphor. Examples include the characterization of slavery as the "Black Holocaust" and abortion mills as "death camps." How does the ambivalent relation of the Holocaust legacy with other victims generate anti-Semitism? We propose a mechanism that builds on the idea that collective traumas produce a culture of national inwardness. Societies with a historical background that favors a victimized perception of their past tend to glorify their suffering and relegate the suffering of others. 5 Works on social psychology confirm this mechanism at the individual level and suggest that when comparisons between in-groups and out-groups are upward, in-group members form negative opinions about the out-group because the comparisons challenge their self-esteem and induce feelings of relative deprivation (Dasgupta et al. 2009;Major, Sciacchitano and Crocker 1993;Pettigrew et al. 2008).
The out-group's superior suffering threatens the social identity of the in-group that was largely built on sentiments of victimization (Noor et al. 2012;Tesser 1988;Vollhardt 2013).
The Holocaust, as the exemplary paradigm of suffering, is evaluated in a similar way. The comparison between the in-group's victimhood and the Holocaust creates out-group prejudice and negativity. Observational manifestations of this type of prejudice include the trivialization of the Holocaust, which can lead up to Holocaust fatigue and Holocaust denial, and the connection of Jews with conspiracy theories. Such theories have been consistently found to be strong correlates of anti-Semitic attitudes (Bilewicz et al. 2013). 6 process. In particular the maintained that ". . . even those individuals and groups who would once have resisted being described as victims now embraced the term -and the political advantages it brought. As always in these cases, this became a competitive, zero-sum game." (Prince and McBride 2011). 6 The case of Poland and Germany illustrate how the widespread postwar victimhood sentiments, dominant in these countries, affected the treatment of Jewish survivors in a direct suffering competition that prioritized their experiences. In Poland, the anti-Semitic pogroms and reactions against the return of Jews relate to the deep belief of local Polish societies that Jews were the instruments of Bolsheviks. This meant that Jews were the perpetrators and responsible for the Polish suffering, which superseded the Jewish one (Kopstein and Wittenberg 2011). Even in Germany, the Prisoners Of War who returned from camps were cleansed of their war actions as their duty and provided them with a double victimhood identity; they suffered from both Hitler and the Soviets (Moeller 2001, 158). As Olick (2013) has argued, a "me too" claim on victimhood became prevalent and characterized German society until the confrontation with their past became the dominant memorial culture.
The increased attention placed by the international community to collective suffering translates trauma prioritization into a need for the widest possible recognition of one's suffering. In so doing, it converts in-group/out-group comparisons into a spiral of victimhood (see e.g. Bar-Tal and Halperin 2013). This tendency is imminent in international campaigns for the recognition of alleged or real genocides. 7 Any failure in recognizing such proclamations creates a hierarchy of victimhood that produces grievances. As the Holocaust represents the ultimate benchmark of victimhood, this quest for recognition might generate anti-Semitic feelings among the populations who consider themselves to have suffered the same, without their suffering having been sufficiently acknowledged by the international community. Even anecdotally, it is still indicative that in the Anti-defamation league survey we cited in the introduction the rates of anti-Semitic attitudes in Armenia rise to 58% of the population, the highest in the region. 8 Another example comes from France, where the main opposition to the introduction of the Holocaust in the secondary education curriculum came from pupils of former French colonies background (e.g. Algeria). They juxtaposed the inclusion of this topic with the absence of a subject dedicated on their suffering by the French during the colonial period (Corbel et al. 2004). (from an advancing peripheral European power) into a humiliated status quo country, whose only concern was to heal the territorial and social trauma caused by the defeat (Leontaritis 1990). 9 The remedy employed was the attribution of blame to foreign factors, the so-called "Great Powers." This tendency was repeated in all other major events of the last century, such as the civil war (Marantzidis and Antoniou 2004), the military coup (Kornetis), or even the current debt crisis (kalyvas?). These narratives have enabled the development and priming of self-centered victimization, accompanied by a mindset of historical injustice (Demertzis 2011). These components have been succinctly summarized in Diamantouros' portrayal of Greek culture as an "underdog culture", characterized by a mixture of feelings of cultural uniqueness and grievance with social discourse taking place through sentimentally-rich historical references (Diamantouros 1983).
The socialization agents that help the intergenerational transmission and recycling of these beliefs are typically public education, national commemorations, public history accounts and family narratives. 10 All these operate during the early years of socialization and they turn those narratives into identity components. 11 The identity becomes so strong that any challenge to the dominant narrative meets severe criticism. One such example was the 6th-grade textbook project debacle that took place in 2008. The project met a wide front of resistance that rejected the presentation of some of the major moments in Greek history, the "Asia Minor Catastrophe", in a less sentimental Ataturk. Obviously, such a revision meant that a prerequisite to respect Jewish suffering was the simultaneous recognition of Christian and Greek suffering and the elevation of the latter to the legal and symbolic status of the former.
12 As shown in the concluding section, almost 90% of Greek public opinion believes that the teaching of the Pontiac Genocide should be mandatory.
13 A typical example of how political elites have not only embraced but have also primed feelings of victimhood competition comes from a speech of the government representative in an inauguration of a Holocaust monument in the Northern part of Greece, in June 2015. As he argued, "the Greek people are sensitive to such acts of violence because they have experienced the Genocides of Pontiacs, the Asia Minor Greeks, the Armenians and the Assyrians [. . . ] WWII barbarity has not stopped ever since. It resurfaces in the Ukrainian crisis, the Turkish Occupation of Cyprus and the Gaza isolation."

Empirics
Our empirical analysis is divided into four parts. We start by visually inspecting the relationship between perceptions of collective victimhood and anti-Semitism. We then move to the experimental evidence, which serves to a) shed light on the underlying mechanisms; and b) rule out competing explanations for our results. In the third part, we make use of a natural experiment to showcase the observable implications of the theory. The last part goes beyond Greece, delving into the origins of collective victimhood and examining its consequences on anti-Semitism in a comparative setting.

Main Results
We use data from two national representative surveys conducted in Greece during 2014 and 2015. 14 We measure anti-Semitic attitudes with two indicators. The first draws on one of the facets of modern anti-Semitism relating to the tendency to trivialize the Holocaust and/or to accuse the Jews of using it instrumentally (Bergmann 2008;Gibson and Duch 1992;Stone 1980). In particular we used agreement and disagreement with the following statement: Jews exploit the Holocaust to gain influence in the international arena. According to the European Monitoring Center on Racism and Xenophobia and the later Agency for Fundamental Rights of the European Union, such attitudes are considered as anti-Semitic (EU-Report 2013). 15 The second indicator reflects a more explicit manifestation of racial discrimination, asking respondents if they agree or disagree that It should be denied from Jews to buy land in Greece. Taken together, the two items allow us to test our theory across different degrees of intensity in terms of anti-Semitic sentiments.
In measuring collective victimhood we are guided by our theory in three key respects. First, we invite respondents to think about the Greeks as a collective identity rather than themselves individually. Second, we use benchmark groups against which we compare the Greeks. Third, we 14 More information about the surveys can be found in the Online Supplementary Information. 15 Other elements are a) a tendency to equalize Israel with Nazi Germany; and b) a belief that the Jews are overrepresented in powerful economic and political institutions Stone 1980). The Online Appendix shows the descriptive statistics for these indicators as well.
vary the degree of recognition of the outgroup's status of victim. In particular, respondents are asked how much they agree or disagree with the statement that Greeks have suffered more than X. X varies according to the treatment. In The first survey, X is a) the Jews; b) the Armenians; c) Others' and d) the Carthenians. The first two groups represent well-known cases of collective suffering. 16 The third treatment made no reference to a specific group, thus leaving the window open for comparisons.
The fourth treatment uses a fictional group (the Carthenians). The rationale for this condition is to use an uninformative benchmark, with no established status of victim. 17 Comparisons with groups that lack recognized victim status leave little room for victimhood competition. If such competition matters, we expect victimhood to exert a weaker effect on anti-Semitism in this condition compared to all others. We repeated this exercise in the second study, in which we only changed the dependent variable (now measuring explicit racist attitudes-It should be denied from Jews to buy land in Greece) and one of the comparison groups. In particular, we replaced the Armenians with a group also well-known to the Greek public but in which public discourse has not been as positive as in the case of the Armenians: the Bosnians (Greeks have suffered more than the Bosnians). 18 All descriptive statistics for these measures are shown in the Online Appendix. high and remain so irrespective of the comparison group. Indeed, the percentage of respondents either agreeing or strongly agreeing that Greeks have suffered more than the benchmark group ranges between 65% in the case of the Jews up to 82% in the case of "Others." Second, there 16 Seen as victims of the same historical "enemy", i.e. the Ottoman Empire and the newly established Turkish state, Armenians consist of a fairly strong benchmark due to the favorable perceptions harbored among the Greek public opinion towards this group.
17 For the usage of fictional groups in survey research see also Gibson and Duch (1992). 18 The Yugoslavian war was very salient among the Greek media, which held a predominantly pro-Serbian stance. We thus expect Greek public opinion to be more likely to downplay the suffering of Bosnians against their collective suffering, compared to either the Armenians or the Jews. Greeks suffered more than: Others The Jews exploit the Holocaust Greeks suffered more than: Carthenians The Jews exploit the Holocaust Greeks suffered more than: Armenians The Jews exploit the Holocaust Greeks suffered more than: Others Denied from Jews to buy land Greeks suffered more than: Carthenians Denied from Jews to buy land Greeks suffered more than: Bosnians Denied from Jews to buy land Greeks suffered more than: Jews Denied from Jews to buy land Note: The blue curves denotes the local linear regression estimate (span=1), with the shaded area surrounding capturing the point-wise 95% confidence intervals. All scatterplot dots are jittered to enable the visualization of the conditional distribution of anti-Semitic attitudes, given the level of collective victimhood.
is remarkable variation in the levels of agreement with each of the anti-Semitic statements. As expected, the percentage of respondents agreeing with the statement that the Jews exploit the Holocaust is much larger, 59%, than those agreeing that it should be forbidden for Jews to buy land (21.5%). Third, regardless of which outcome and which benchmark group is used, there is a largely monotone positive relationship between victimhood and anti-Semitic attitudes. The only exception to this rule is the treatment using the fictional group of Carthenians. Consistent with our expectations, when the comparison group does not signal a well-known case of victimization the relationship between victimhood and anti-Semitism weakens considerably.  of victimhood are associated with significantly higher levels of anti-Semitism. The similarity in the magnitude of the effects is important because it helps to address endogeneity concerns. If we only found an effect when using the Jews as a comparison group, such effect could be also attributed to a feedback loop from anti-Semitic attitudes to victimhood. The fact that the effect appears to be indistinguishable between Jews, Others, Armenians and Bosnians helps to rule out the possibility that the relationship is simply an endogeneity artefact. The only condition for which no significant effect is found is the group with no reputation of suffering. 20

Unpacking the Mechanism I: Upward Comparisons
The results thus far support the idea that victimhood predicts anti-Semitism. As a next step we try to delve into the mechanism driving our effects. We do this by designing an experiment tailored to test the logic of victimhood competition. Table 1 presents the treatments for the first experiment.
We use the WWII as the historical setup to induce variation in victimhood. We find this setup useful for three reasons. First, it alludes to a salient event of which there is ample recognition of in-group victimhood. Second, the WWII goes sufficiently back in time to ensure that it does not bring the financial crisis in mind. Third, it offers useful leverage to vary the degree of suffering competition. We do that by using first the Jews and then both the Poles and the Jews as groups against which the Greeks are compared. Thus, whereas in the first treatment condition Greeks appear as those who suffered most from the WWII, conditions two and three induce competition, with the third treatment inducing more competition than the second one. If our theory is correct, the level of anti-Semitism should increase monotonically from the control condition and up to the third treatment condition. The results appear in the first panel of Figure 3 and confirm this expectation. Compared to the control condition, victimhood triggers anti-Semitism and does so in a steadily ascending fashion as competition over the magnitude of collective suffering intensifies. Is the mechanism driving these effects an increased sentiment of collective suffering? The second panel of Figure 3 tests this interpretation of the results. It presents a simple manipulation check, other. We try to account for this possibility by combining the two treatments. Finally, we also ask the question without any prior adjustment. The results appear in Figure 4. As the first panel of the Figure shows, the highest percentage of agreement with this statement is found for those who received the victimhood treatment, whereas the lowest level of agreement is observed for the scapegoating treatment. 21 The victimhood category also outweighs by nearly ten percentage points the control category, in which the question about the Jews was not preceded by the treatment. The second panel of Figure 4 directly compares victimhood against the other three conditions. Our main interest lies in the comparison between victimhood and scapegoating. We find a significant difference, with respondents in the victimhood treatment being approximately 17% more likely to agree with the Holocaust statement (95% CIs Note: Each bar represents the percentage of respondents who agree or strongly agree with the statement that the Jews exploit the Holocaust. The vertical bars represent the parametric 95% CIs. [0.079-0.261]). 22 Victimhood is also likely to lead to a higher percentage of agreement compared to the control condition, although this difference marginally fails to reach statistical significance (95% CIs [-0.018-0.158]). Given the results thus far, the third entry comes as no surprise: when combined the two treatments seem to cancel each other out. If anything, victimhood seems to operate better alone than in conjunction with scapegoating, even though this difference is not statistically significant (95% CIs: [-0.047-0.128]). This uncertainty notwithstanding, it seems that the only treatment that induces higher levels of anti-Semitism is the one used to tap victimhood with the scapegoating treatment having the reverse effect (as in Gibson and Howard 2007).
One last concern could be that our treatment works not because of activating victimhood but hather because it activates scapegoating. Our manipulation check suggests that this is not the case.
We find that the treatment of victimhood sparked higher rates of "Strongly Agree" responses in the victimhood item, while it had no discernible effect on the scapegoating question. 23

Observable Implications: Evidence from A Natural Experiment
Taken as a whole, the results thus far point to two conclusions. First, victimhood increases anti-Semitism and this is not due to scapegoating. Second, upward comparisons exacerbate this link: when competition increases so does anti-Semitism. In this section, we make use of a natural experiment to test one observable prediction of the theory. We ask what happens when a random shock changes perceptions of victimhood. We address this question by making use of an event that took place during the period of data collection for the first study. Interviews for the first survey With this victory, Greece qualified for the first time in its history to the last 16 of the tournament.
Given that football is by far the most popular sport in the country and given that the World Cup represents the most salient and prestigious international tournament, this victory became -hardly unexpectedly-headlines in the national news. One-third of our respondents (n = 354, 33.88%) were already interviewed before this game, and two-thirds (n = 691, 66.12%) were interviewed between Wednesday and Friday, i.e. after knowing the result of this football match.
As previous literature has shown, national athletic events can have significant effects on public opinion (Abell et al. 2007;Beck 2003;Ohmann, Jones and Wilkes 2006). Sports achievements, in particular, raise national pride and Greece should be no exception. 24 A boost in national pride strengthens nationalist sentiments and exacerbates in-group favoritism (Bertoli 2013). This predicts Note that the equivalent manipulation check for scapegoating also does not influence victimhood. It also does not influence the corresponding scapegoat manipulation check. People have very strong opinions about who is responsible for the Great Recession; it should not come as a surprise that we are unable to move that item with the scapegoating treatment.
higher levels of prejudice against out-groups. Being one such group of outsiders, Jews, are expected to be perceived more negatively right after the game than before.
Our victimhood hypothesis, however, leads to a different expectation. If negative judgments against the Jews stem from self-categorizations of the nation as the underdog, a sports victory of this magnitude and saliency, coming with an ambiguous favorable decision by the referee, should, if anything, reduce victimhood. If victimhood is a driving force of anti-Semitic sentiment, then we should see reduced levels of anti-Semitic judgments.
We exploit this quasi-random variation in perceived victimhood to test the two divergent expectations. In particular, we use the date of the interview as an instrument of victimhood. Respondents interviewed from Wednesday to Friday are given the score of one whereas those interviewed on Monday and Tuesday (the last Tuesday interview had ended before the start of the game) are given a zero. Our treatment is also binary, switching on for those respondents who agree or strongly agree that the Greek nation has suffered worse genocides than the Jews. We use as outcome the Holocaust Instrumentalization item.
Before moving to estimation, we try to briefly examine the two major assumptions upon which our IV estimation is based: ignorability and exclusion. We think ignorability is unlikely to be violated because in telephone surveys that take place amidst the electoral cycle there is little scope in choosing when to be interviewed. Exclusion, however, can be violated for two reasons. First, different people might be more accessible at the beginning of the week than towards the end of the week. These differences might have an effect on anti-Semitism without being driven by victimhood.
Second, the game might not have resulted in lower anti-Jew prejudice because of a decline in the sense of victimhood but due to an increase in the feel-good factor generated by the victory. Table 3 tries to address the first concern by examining baseline differences between respondents interviewed before and after the match. We include a series of demographics and two attitudinal predictors (political knowledge and ideological self-placement), which we believe are unlikely to have been influenced by the match. We find no significant difference between the two groups. The overall level of balance approaches the standards of a randomized experiment. 25 To examine whether the results are due to a more generic feel-good factor we use the instrument as a predictor of the level of trust towards various groups and institutions: the Jews, the Americans, the Turks, the LGBT community, immigrants, the Church, the Parliament, the EU, and interpersonal trust. The test is based on the idea that victimhood should matter only with respect to the Jews, who are also perceived as victims. It is not related to social or ethnic groups or institutions that lack this characteristic. The results of this exercise appear in Figure 5. As it is seen, the only group for which we find a significant change in trust after the match is the Jews. Table 4 presents the results. The coefficient of the first column denotes the difference in the average value of the outcome before and after the game. In so doing, it provides an estimate for the Intent-To-Treat (ITT) effect of victimhood on anti-Jew prejudice. 26 In line with the victimhood 25 We employed an omnibus test, proposed by Hansen and Bowers (2008), which tests the null of no systematic differences between groups and assesses balance not only on the set of covariates listed in Table 3, but also on all their linear combinations. The Chi 2 statistic is 6.53, which with 9 degrees of freedom gives a p-value of 0.686. 26 The difference remains significant when controlling for all covariates shown in Table 3. Note: The black dots indicate the OLS coefficient from regressing each trust indicator on the dummy distinguishing respondents according to whether they were interviewed before or after the game. The horizontal spikes denote the 95% CIs. All trust items are measured using a 0-10 scale, except from Interpersonal trust, which is originally measured with a 5-point scale but has been recoded into a 0-10 scale.
hypothesis, levels of anti-Semitic attitudes drop significantly after the match. The second column shows the first stage of the IV analysis. We find a negative coefficient which implies that respondents interviewed after the football match were less likely to agree with this statement. Although the football match seems to work towards the expected direction, it does not seem to be a very strong instrument. Essentially, the design represents a conservative test of the victimhood hypothesis.
The second column shows that this effect largely holds when all variables shown in Table 3   together, the findings provide suggestive evidence for the victimhood hypothesis. Becoming "Kings for a Day," Greek citizens felt fewer victims than usual, and this led to a more favorable posture towards the Jews.

Anti-Semitism Beyond Greece
Is this story only applicable to Greece? Although the scope conditions of the theory extend to any polity with high levels of collective victimhood, testing the generalizability of the argument beyond this case requires a comparative measure of perceived victimhood, which is particularly difficult to find. We can still gauge the external validity of the theory, however, by focusing not on victimhood itself but the extent to which the teaching of history in a given history is introvert or extrovert. We posit that this emphasis on the national (rather than the international) is at the core of victimization. One example might serve to clarify this point. We conducted an experiment, in which we asked people whether a given topic should be made obligatory in the history school curriculum. In total four topics were discussed and each topic was randomly asked to a portion of the sample. The topic with the highest acceptance rate (95%) was the "Asia Minor" Catastrophe, the archetypical episode of victimization within contemporary Greek history. The lowest percentage (50%) was the Holocaust. 27 This finding signals an inward-looking stance towards history and the world in general. It is this tendency that retains high levels of anti-Semitism in a country with an infinitesimal number of Jews.
To move beyond the Greek case we use the 1997 data from a rare cross-national survey of high-school pupils in 23 countries, targeting their views about how the subject of history should be taught at school (Angvik and von Borries 1997). In particular, students were asked how much they want to be taught about their own country's history versus the world history. Although not a perfect approximation of victimhood, it clearly depicts inwardness across countries. We take the average national score in that question and plot it against the 2014 ADL score of anti-Semitism, as shown in the upper-left panel of Figure 6. The blue curve captures the local mean of ADL score, conditional on a country's average score in the way pupils want the subject of history to be taught. Imposing no structure in the data, the pattern that emerges denotes a strong, monotonically ascending relationship, indicating that inward-looking stances towards one's history help us predict anti-Semitic attitudes. The lower-left panel displays the same relationship this time controlling for a series of potential confounders: population density, GDP per capita, ethnic, linguistic and religious fragmentation (using the Alesina et al. (2003) indices), a binary indicator for post-Communist regimes, and a proxy of Ethnocentrism, obtained from the same survey (through the national averages in the question "How important is [countryname] to you personally? "). The x-axis depicts the residuals from the regression of National History on these covariates, whereas the y-axis uses the residuals from the regression of ADL score on the same control variables: The relationship remains robust when controlling for this set of covariates. 28 A different way to examine the same pattern is to study the teachers' views on the subject. Part of our argument with respect to the sources of victimization suggests that teaching history in a 27 The second and third topics were the Greek civil war and the Pontiac genocide, which received approximately 80-85% acceptance rates. 28 When the x-axis is replaced by the ratio between interest in national history and interest in International History, the results are almost identical, as shown in Table A.3 of the Online Supplementary Information. Note: The left column presents the relationship between the 2014 ADL score and National History; the upper graph uses a local linear regression to capture the bivariate relationship between these two variables, whereas the lower graph uses the residuals from both variables, after each has been regressed on the following list of covariates: population density, GDP per capita, ethnic, linguistic and religious fragmentation, post-Communism, and ethnocentrism. The right column presents the same results but using the teachers' assessment of the frequency in which history is portrayed as good or bad or black and white. In all plots, the blue curve denotes the local linear fit between the two variables, whereas the shaded area encompasses the 95% confidence intervals.
single-minded manner-with historical events being evaluated as right or wrong-drives perceived victimhood. In the same 1997 survey, teachers were asked to report on the character of a usual history lesson. The measure examined how frequently history was portrayed as "good or bad, black or white" and it was based on a 5-point scale (very seldom-very often). 29 The upper-right panel of Figure 6 denotes the bivariate relationship between the two variables, whereas the lower-right graph displays the same relationship with controls. For the most part, both graphs confirm the positive relationship between the tendency to teach history in a Right/Wrong fashion and anti-Semitic attitudes. 30 To be sure, this exercise represents only suggestive evidence, as future work is clearly needed to fully explore the comparative statics of the theory. That said, this finding speaks to the heart of our argument and lends itself to the policy implications of our theory.

Discussion
Anti-Semitism, "The Longest Hatred" as Robert Wistrich (1991) eloquently described it, neither started nor ended with the Holocaust. Masqueraded in various forms prejudice against the Jews has been with us for millennia. The Holocaust experience and various attempts to combat this phenomenon after the end of the WWII have resulted in a decrease in the levels of anti-Semitism compared to the interwar period. That said, anti-Semitic attitudes are still worryingly high. Importantly, although modern anti-Semitism represents a more latent and less explicit attitude than in the past, it still has important behavioral implications. Attacks against Holocaust memorials, Synagogues, and Jewish properties are far from rare. The puzzle is thus still vivid. What accounts for this prejudice?
We provide ample survey and experimental evidence that anti-Semitic stances are strongly related to a sense of victimhood. Our micro-level explanation relied upon works in social psychology and highlighted the role of competition in suffering as the key mechanism. In line with the literature, 29 The X-axis on the right panel is based on the following question: "Students are told what is good or bad, right or wrong in history". This question was part of a number of items that were introduced with the prompt, "What actually happens during your History lessons? ".
we also maintained that victimhood stems from both formal and informal means of socialization.
Socialization agents as diverse as the family, the school, and the media transmit the predominant national narrative that emphasizes collective traumas, breeds national self-centeredness and creates a sense of national entitlement. Our comparative evidence in the previous section illuminates this pattern and lends confidence to the generalizability of our theory.
The findings also touch upon a more general trend after the WWII. The Holocaust experience has activated the international community in becoming more alert when human rights violations are observed. Over the last decades, various groups have been trying to motivate the international community to recognize past or more recent genocides. This is not surprising, as the status of the victim has nowadays more concrete and symbolic benefits than in the past. Based on these findings, we should expect that this trend will have counter effects, boosting competition between groups and thus leading to higher levels of anti-Semitism. Future research on this question might be useful in shedding light on the unintended consequences of the internationalization of victimhood.
Reflecting upon our results, it is clear that victimhood is a necessary and sufficient condition for the development of anti-Semitic sentiments. As the example of interwar Germany indicates, the sense of victimhood might be the fundamental element in both anti-Semitic attitudes and the scapegoating narrative. Although this sense of collective victimhood (stemming from the Versailles treaty and its devastating terms) was conflated with scapegoating (with the appearance of the "stab in the back" myth in the Weimar Republic or the then popular conspiracies about Jews) most scholars would agree that the former brought the latter and not vice versa. Importantly, the way future generations learned about the WWII and the Holocaust had serious implications for post-war anti-Semitism. By the late 1950's, anti-Semitism was -again-on the rise and the Germans decided that an alternative pedagogical approach to their recent past was necessary to move forward. This shift included a direct confrontation with their own responsibilities. As Puaca (2011) suggests, "Post-War textbooks left no doubt that the war had created many victims, but their authors also had no qualms about placing the German people high on that list." (p.136). A decade later, the History curriculum and the general attitude toward the WWII changed dramatically to address the nation's responsibility for WWII and the extermination of the Jews. This shift from portraying the German people as (one more) victim of the Third Reich to a nation that bears responsibility for war crimes altered the course of socialization for a generation of young Germans (see also Adorno et al. 1950).
There Israel: Israel treats Palestinians in the same way they were treated by the Germans in WWII.
Business: Jews have too much power in the business sector. 31 Responses to all three items follow the classic 5-point Likert scale, ranging from Strongly Disagree (1) to Strongly Agree (5). 31 The actual wording of the third sentence differs slightly because we randomized the subject of the sentence. Instead of Jews, for one random third of the sample the sentence used "Zionists" and for the last third, "Jews" were replaced by Israel. No discernible difference is observed. Consequently, we simply add up the three parts of the sample.  Respondents were then asked to assess how positive they think this development is for their area on a 0-10 scale. Treatments were defined by the identity of the investor. A random third of the sample was presented with each scenario. We examine the role of perceived victimhood in respondents' evaluations of the investment opportunity across the three scenarios.  pattern, we interact our indicator of victimhood with each of the two treatment conditions: 32 A possible critique to the design could be that we compare a religious stimulus to a national stimulus. Our exercise using the "business" question suggest little room for such a confounding effect: As already mentioned, we found no difference between the three stimuli: "Jews", "Zionists" and "Israel." We are thus confident that we capture prejudice against the Jews.
Y represents the investment evaluation score on a 0-10 scale and V stands for victimhood. JCA and F CA denote respondents who were told that the investor is the Jewish and the French Commerce Association respectively. The EU condition is treated as the reference category. Thus, β 1 + β 4 V provides the estimate of the difference in the investment assessment between the Jewish Association treatment and the EU treatment, conditional on respondents' victimhood score. Similarly, we expect an equivalent gap between the JCA and the EU conditions, which is given by the following The results of this analysis are presented in the first two panels of Figure   The overall difference found initially in Figure A.4 seems to be mainly due to those who believe that Greeks have suffered more than the Jews. The last panel presents a placebo test: when comparing the French and the EU treatments, we find a considerably smaller gap in the investment evaluations as the level of victimhood increases. 33 Consistent with our expectations, victimhood can explain why the JCA is less preferred to the EU and the FCA, but cannot explain why the EU is more preferred to the FCA. Taken as a whole, the results provide at least suggestive evidence that victimhood predicts anti-Semitic attitudes not only when standard observational measures are used to tap into the concept but also when such attitudes are measured in a more implicit, experimental set-up.  Note: Entries in the first three columns are group averages whereas the fourth column shows the p-value from a F-test testing the null of no difference between groups.

Full Randomization-Inference Results
We present the distribution of treatment effects under the sharp null, i.e. under the assumption the effect is zero for all respondents. The solid red vertical line denotes the sample treatment effect.
The position of this line indicates how likely this estimate is given the null. The dashed vertical lines provide the 95% confidence intervals, stemming from this procedure. Confidence interval construction follows the path described in (Gerber and Green 2011): we impute the missing potential outcomes (Y 1 ) for the control groups by adding the estimated ATE to the observed Y 0 . For treated observations, missing "untreated" values are imputed by subtracting the estimated ATE from the observed treated values. We then list the estimated ATE from each permutation in ascending order, where the 2.5th percentile marks the bottom of the 95% confidence interval, and the estimate at the 97.5th percentile marks the top.     Figure 8 of the main text. We use three indicators, which we believe that reflect, albeit imperfectly, the socialization of perceived victimhood. All three measures represent national averages in surveys of random sample of pupils and teachers, conducted in 23 countries as part of the State of History Education in Europe project, published in 1997 (Angvik and von Borries 1997).
• Ethnic, Religious & Linguistic Fragmentation: source Alesina et al. (2003). All estimates shown in Table A.2 stem from OLS models, with heteroskedasticity-robust standard errors in parentheses. In the last two columns, we lose one country (Croatia), because no History Teaching data were made available for this question.