ABSTRACT
How does ethnic violence affect the perpetrator group’s attitudes towards the victim group? Combining the security threat and ultimate attribution error, I expect that perpetrator group members react to security threats caused by ethnic violence to form negative attitudes towards the victim group. This study tests this hypothesis using the case of ethnic violence against refugees who came to Germany during the “migrant crisis” in 2015. Results show that a higher rate of anti-refugee violence in a German state is associated with more negative attitudes towards refugees among German natives. The effect of ethnic violence is more prominent among Germans with higher concerns for security, pinpointing the threat mechanisms. A series of additional tests, including instrumental variable strategy, Bayesian analysis, and robustness checks support the conclusion. The results of this study indicate that ethnic violence induces negative attitudes among the perpetrator group towards the victim group.
Acknowledgement
An earlier version of this article was presented at the 41st annual meeting of the International Society of Political Psychology (ISPP) in San Antonio. I would like to thank Kikuko Nagayoshi, Takuma Kamada, Gong Shun, Nobuo Suzuki, and anonymous reviewers for suggestions and comments. I take full responsibility for all infelicities and errors of judgment or interpretation.
Disclosure statement
No potential conflict of interest was reported by the author(s).
Notes
1 Since 2015, European countries have experienced a massive influx of immigrants and refugees. This so-called “migrant crisis” generated over 1.2 million asylum applications in 2015 and in 2016, historical record numbers in European countries (Eurostat Citation2017). During this period, 47.3 per cent of those asylum applications were registered in Germany (ibid.); this is primarily because of the stability of the German economy and the tolerance of the German government, which attempted to accommodate and protect a large proportion of the asylum applicants (Crage Citation2016; Juran and Broer Citation2017).
2 I should note that ultimate attribution error has been effective in situations of intense intergroup conflict, such as in Burundi’s incidents. Nevertheless, psychological mechanisms of attribution error may operate in Germany due to perceptions of heightened security threats.
3 I should note that excluding demonstrations from the anti-refugee violence variable does not alter the main results presented in .
4 I conducted exploratory factor loading, and the three questions are summarized as one factor. The factor loadings for the questions about bogus refugees, government judgement, and family reunification are .554, .644, and .656 respectively.
5 This study uses anti-refugee violence rates rather than the total number of violent events as the main independent variable. The mechanism, security threat perceptions, is expected to be higher when regional security is perceived to be threatened. The perception of threats may stem from the appearance of anti-refugee violence or the extent to which anti-refugee violence seems to be present. Thus, counts of violence are standardized by population.
6 Models 1–6 are estimated with the mixed command, and Model 7 is estimated with the xtivreg command in Stata version 14. Model 8 is estimated with the R package brms, based on the NUTS algorithm.
7 An important omitted variable in this case is precluded resentments (e.g. negative attitudes towards refugees in 2013, before the migrant crisis occurred), and estimation with IV can isolate these effects.
8 I use the t-value instead of the F-value because I use only one IV.
9 Using Bayesian estimation to overcome small-N problems at level-2 of a multilevel analysis is becoming common. Research by Manatschal and Stadelmann-Steffen (Citation2014) is a good example.
10 I should note that although it is possible that attitudes induce violence, this possibility does not deny the causal direction in our study, because causality is, by definition, “comparison (e.g. difference) between the value of the outcome if the unit is treated and the value of the outcome if the unit is not treated” (Angrist, Imbens, and Rubin Citation1996, 444) and not determining the direction of effect.