How terrorism affects political attitudes: France in the aftermath of the 2015–2016 attacks

Abstract

This study investigates what impact the terrorist attacks in Paris (2015) and Nice (2016) had on political attitudes in France. Drawing on nine cross-sectional surveys, it tests the premises of three major theories of opinion change that predict contrasting shifts in opinion among ordinary citizens according to their ideological position in the aftermath of terrorist attacks: the Reactive Liberals Hypothesis (RLH), the Terror Management Theory (TMT), and the Bayesian Updating Theory (BUT). In line with both RLH and BUT, the findings show that left-wing sympathisers shifted toward the right following the attacks. However, the results suggest that, in line with BUT, the attacks only had a significant impact on attitudes toward security, while they had no effect on attitudes toward immigration, or toward moral and socio-economic issues.

Keywords:

• France
• terrorism
• attitudes

Acknowledgements

We are especially grateful to Will Jennings and Christopher Wlezien as well as the anonymous reviewers for helpful comments and suggestions. A previous version of the paper was presented at the 2016 CEVIPOF ‒ LIEPP workshop of the Policy Priority Barometer. We also thank all the participants for their comments and suggestions. The usual disclaimer applies.

Notes

1. In the post-regional wave in December 2015, the sample size (4827) was substantially higher than 1500 individuals.

2. Cronbach’s alpha for these four items is 0.72. A factor analysis of the four items displays only one significant factor.

3. Cronbach’s alpha for these two items is 0.76. A factor analysis of the two items shows that all items load on one significant factor.

4. Cronbach’s alpha for these nine items is 0.71. A factor analysis of the nine items shows that all items load on one factor.

5. Cronbach’s alpha for these two items is 0.71. A factor analysis of the two items displays only one significant factor.

6. Details about this segment of the sample are provided in the Appendix A.

7. Whether or not control variables are included does not affect the substantive results on the key variables. We also tested media preference variables and interactive terms between the media variables and Wave ID. The results remain unchanged. Control variables are not displayed for reasons of space.

8. We generated the predicted values on the security scale for the pre- and post-attack waves associated to each terrorist attack.

9. Two types of analyses were carried out. First, we ran t-tests on the mean left‒right self-positioning according to the date of the survey wave: the mean left‒right self-positioning has never been significantly higher – i.e. more on the right ‒ since December 2014. Second, we specified a regression analysis with the left‒right self-positioning as a dependent variable and survey wave, age, gender and occupation as independent variable: field surveys after December 2014 were never associated with a positive and significant coefficient that would have indicated a drift to the right after December 2014.