Are we all Charlie? How media priming and framing affect immigration policy preferences after terrorist attacks

Abstract

Terrorist attacks negatively affect support for immigration policy, and this has been linked to the extensive media coverage of terrorism. Yet, this coverage may also have a moderating effect. This article uses the timing of the fielding of the European Social Survey, which took place during the Charlie Hebdo and Hyper Cacher attacks, as a natural experiment. Because the media coverage of the attacks varied between France and other European countries, it is possible to study how differences in the media framing and priming of the attacks affected attitudes. The expected negative effect on immigration policy preferences is found outside France, but not within France. This study’s findings lend support to a moderating effect of the media coverage of terrorist attacks, both as a framing effect that influenced the perceived relevance of the attacks to immigration attitudes and a priming effect that primed the public with tolerant French Republican values.

Keywords:

• Terrorist attacks
• immigration attitudes
• priming
• framing
• Charlie Hebdo

Notes

Acknowledgements

The author wants to thank the two reviewers for valuable comments and input. In addition, the participants at the panel ‘With Malice towards Some: The Political Psychology of Prejudice, Discrimination and Acceptance’ at the European Consortium for Political Research (ECPR) General Conference 2017 at the University of Oslo, Norway, 6–9 September 2017, and at the Politikkseminar at Institute for Social Research gave helpful comments and suggestions. He is grateful for comments from Henning Finseraas, Marte Winsvold, Nanna Fredheim, Atle Hennum Haugsgjerd, Andreas Kokkvold Tveit, Dina Heider Hov, Helge Renå, Rune Karlsen, Bernard Enjolras, Knut Heidar and Kari Steen-Johnsen.

Disclosure statement

No potential conflict of interest was reported by the author.

Notes

1 The model of framing presented here differs somewhat from the one Chong and Druckman (2007) present. In their model, an attitude is the sum of the products of the weights (w) and the valence (v) of other considerations. Thus, terrorist attacks increase the weight of terrorism and affect attitudes. Their model does not explicitly discuss which issues are included in the consideration itself. In the model used here, the weight (w) is seen as consisting of two dimensions, accessibility (a) and relevance (r) (i.e. applicability).

2 This study differs somewhat from most studies of framing effects in that it is not the issue at stake (immigration policy) but, rather, a separate consideration (the terrorist attacks) that is being framed. Accordingly, some parts of the media coverage that could lead to framing effects of terrorism (i.e. the connection to French Republican values) are assumed to have priming effects in the context of attitudes towards immigration policy. Immigration policy may be framed as connected to terrorism (e.g. the study by Lahav and Courtemanche (2012)) or to Republican values (e.g. the study by Nugier et al. (2016)), but that is not the case here.

3 This argument mirrors the findings by Albertson and Gadarian (2015) on the effects of anxiety. They find that anxiety caused by perceptions of threat increases policy support for safety measures, but only if these measures are perceived as relevant to the threat.

4 This became a problem for Prime Minister Aznar as it soon became evident that Islamic terrorists connected to Al Qaeda perpetrated the attacks and not Basque separatist terrorists connected to Euskadi Ta Askatasuna (ETA).

5 The brothers were not immigrants themselves, but their parents were born in Algeria. In coverage from the English-speaking France 24 during the first two weeks after the attacks, the attackers’ French citizenship was downplayed, and they were described as foreign or Algerian (Połońska-Kimunguyi and Gillespie 2016).

6 This study uses the seventh round of the European Social Survey (ESS). The data that support the findings are available without restrictions for not-for-profit purposes from Norwegian Centre for Research Data (NSD) for European Social Survey European Research Infrastructure (ESS ERIC) at http://www.europeansocialsurvey.org Replication R-code is available from the author’s website: http://www.oyvindsolheim.com.

7 The code from the rdlocrand R-package (Cattaneo et al. 2016) is used and modified to allow for blocked sampling and blocked measurement of the means. Replication R-code is available online.

8 The tests are run for gender, age, national income decile, dummy for parents born in country, dummy for being in paid work and five education dummy variables. None of these variables could be affected by the attacks (Montgomery et al. 2018). Recent papers have controlled for ideology (Castanho Silva 2018) or used ideology (left or right self-placement) as an independent variable (Brouard et al. 2018). This may be problematic here as placement on this scale could be affected by the attacks (for example, because of increased conservatism under threat (Nail et al. 2009)); Castanho Silva (2018: Table 1) does seem to find a right-wing shift after the attacks.

9 All the samples are balanced in the aggregate as well; see the results of the balance tests in Table 9 in the online appendix (supplementary material).

10 The interviews were not conducted in one region at a time, but most individual interviewers only interviewed in one region. Thus, there is reason to believe that the random selection of respondents answering before and after the attacks happened at the regional level. This study uses the highest regional level where there is more than one region (i.e. The Nomenclature of Territorial Units for Statistics (NUTS) 1 or NUTS 2). NUTS is a standard for geographical areas in the European Union.

11 This gives a weighted mean difference, in which each country weighs the same, and each region’s weight in each country is based on its number of respondents. Pooling all respondents (no weights) and weighting by region so that each region weighs the same inside the countries does not give very different results; see supplementary material, Tables 6–8.

12 This varies a bit between the different analyses as the balance tests are run for each of the dependent variables separately, and there are differences in the amount missing for each dependent variable.

13 The date of the attacks, ‘1715’, is set as a seed to facilitate the replication of the results.

14 Controlling for political interest by including a similar interaction as the one with TV news does not affect the results either (not printed).

15 The similarly phrased question on allowing immigrants from poorer countries inside Europe was not asked in the Czech Republic, and, therefore, it is not included in the analysis.

16 The indexes have a high degree of internal consistency, with a Cronbach’s alpha of 0.88 for the general index, 0.86 for the non-European index and 0.76 for the European index.

17 See Table 10 in the online appendix (supplementary material) for the regression results.

Additional information

Funding

This article was made possible by funding from the Research Council of Norway’s program for Societal Security (SAMRISK II), project no. 238118.

Notes on contributors

Øyvind Bugge Solheim

Øyvind Bugge Solheim is currently an adviser in the secretariat of the Norwegian Government Commission on Electoral Law. His research examines the effects of terrorism on political attitudes. He is primarily studying the effects on attitudes towards immigrants and minorities and how such effects are contingent on other factors, such as the societal and political reactions to terrorist attacks. [[email protected]]