What’s left after right-wing extremism? The effects on political orientation

. Does radical right political violence favour or hinder public support for right-wing stances? Numerous existing studies have demonstrated that Islamic terrorism provokes a conservative shift, increases nationalism and induces negative sentiments towards immigration. However, little is known about the consequences of far-right terrorism, despite its incidence in Western societies. We leverage four waves of the British Election Study (BES) and use a quasi-experimental design to analyse individual political orientations shortly before and after terrorist attacks. We ﬁnd that respondents distance themselves from the ideology associated with the perpetrator and shift away from ideological positions at the right end of the political spectrum. Furthermore, respondents are less likely to report nationalistic attitudes and immigration skepticism, core tenets of extremist right-wing political ideologies. Our ﬁndings suggest that the characteristics of the perpetrators and their driving goals are crucial factors shaping the impact of terrorism on public sentiments.


Introduction
In recent years, radical right parties have secured increasing public support in many advanced democracies and gained electoral ground, following political campaigns that exploited public anxieties around globalization and migration (e.g., Norris and Inglehart, 2019).At the same time, several countries have experienced political violence motivated by extremist rightwing ideologies (Piazza, 2017;Ravndal, 2018).Whereas these two phenomena are certainly distinct -and radical right parties use conventional democratic means to influence politics whereas extremists openly reject democracy (Mudde, 2007;Ravndal, 2018) -violence on the right often intersects with traditional politics.For one, the two groups embrace the same rhetoric characterized by frustration with the political establishments, support for authoritarian values, concerns about a dilution of national identity, as well as prejudice and intolerance towards ethnic and racial minorities (Mudde, 2007;Ravndal, 2018). 1lthough political activists seldom engage in actual violence and often condemn violent methods (Bjørgo and Gjelsvik, 2017), little is known about the extent to which radical right political violence favours or hinders public support for the ideas they advance.Extremist individuals engaging in such activities often believe that their views are enjoying growing social legitimacy and acceptance (Perliger, 2017); and far-right movements express antiimmigrant and culturally protectionist positions that are often popular among moderate sectors of the electorate and promoted by mainstream conservative parties (Van Spanje, 2010;Wagner and Meyer, 2017;Abou-Chadi and Krause, 2018).
We ask whether right-wing terrorism affects political orientation and the support for ideas that are shared to varying degrees by conservative parties.Recent studies have investigated how terrorism affects public opinion and have demonstrated that terrorist incidents lead to more support for right-wing parties (see, e.g., Berrebi and Klor, 2008;Kibris, 2011;Getmansky and Zeitzoff, 2014;Aytaç and C ¸arkoglu, 2019).Yet, with the exception of studies on the 2011 terrorist attack in Norway (e.g., Fimreite et al., 2013;Jakobsson and Blom, 2014;Solheim, 2018),they have largely focused on Islamist extremism or assumed homogeneity in the characteristics of perpetrators and in their driving goals.
The perpetrators of radical right violence often express ideas that are becoming prevalent within some sectors of wider society.If anything, diluted or more moderate versions of their driving ideology are shared by some segments of the public, such as a decreasing tolerance towards immigration or increasing concerns for the protection of national identities (Colantone and Stanig, 2018).Yet, there is an important moral dimension associated with violence; that is, its incompatibility with the values and needs of the general population.
Violent tactics make it more likely that movements and political actors are perceived as unreasonable by citizens, who identify less with, and ultimately support less, the very ideas they promote (Simpson et al., 2018).Violence is particularly unlikely to resonate well in democracies, even when some of the demands may be considered legitimate by a segment of society (Muñoz and Anduiza, 2019).2As such, negative views of the attack may lead people to dissociate not only from the terrorist, but also from their ideas (Jakobsson and Blom, 2014;Solheim, 2018).This simple argument leads to the theoretical expectation that right-wing extremism pushes the public to distance itself from the attack and the perpetrator and thus makes it less likely to support far-right ideologies.
We investigate whether right-wing extremism affects citizens' political orientation in the UK, an interesting and timely case study as a third of all terror plots since 2017 were driven by extreme-right causes. 3In December 2016, the UK Home Secretary proscribed the first extreme right-wing group, National Action, under the Terrorism Act 2000 (UK Home Office, 2018).And in 2019, "extreme right-wing" terrorism has been included in official threat-level warnings for the first time. 4We leverage four waves of the British Election Study (BES) and employ a quasi-experimental approach that exploits the occurrence of an unexpected event (terrorism) during the fieldwork of a survey to assign respondents into treatment and control groups as good as randomly (Muñoz et al., 2020).We complement this strategy by exploiting the panel dimension of the data, and control for individual baseline information to further address issues of unobserved confounding factors and differential response rates (Nussio, 2020).
We focus on two well-known incidents that were incited by extreme-right propaganda: the 2016 murder of MP Jo Cox by a 53-year-old gardener with far right views and links to British and American far-right political groups; and the 2017 Finsbury Park attack, when Muslim worshippers were attacked in north London by a man influenced by the propaganda of far-right anti-Muslim organizations, such as the English Defence League. 5e find that exposure to far-right terrorism shifts citizens' ideological positions away from the 'right' ideology; that is, following these incidents, individuals are less likely to place themselves at the right end of the political spectrum.We also show that respondents are less likely to report nationalistic attitudes that have been shown to affect radical right voting, such as national identification and immigration skepticism (Rydgren, 2008;Lubbers and Coenders, 2017).This result suggests that there are some genuine shifts in the underlying at-titudes typically associated with far-right electoral support, to the extent that these attitudes are less prone to social desirability bias.To lend further credibility to our causal claims, we implement the Muñoz et al. (2020)'s best practices in the form of falsification tests, placebo treatments, and several other sensitivity and robustness checks, and the results carry over.

Data and Methodology
We employ individual-level data from the British Election Study (BES) -an internet panel survey with a stratified random probability sample of citizens living in England, Scotland, and Wales.BES includes questions about the respondents' socio-demographic attributes and political preferences, as well as questions designed to capture their attitudes on key issues, such as immigration and national identity.Exploiting information from the survey waves that coincide with terrorist attacks allows us to examine the causal effect of terrorism on people's responses based on an 'Unexpected Event during Survey Design' (UESD, Muñoz et al., 2020). 6Our identification strategy relies on the assumption that the timing of attacks is exogenous (unexpected) and largely randomly assigned relative to that of the interviews, and thus individuals interviewed after the attack can be defined as the 'treatment' group whereas those interviewed before the attack can be defined as the 'control' group.
We consider two terrorist attacks that were motivated by far-right extremism: the murder of MP Jo Cox (16 June 2016) and the Finsbury Park attack (19 June 2017).Both attacks overlapped with recent BES waves (8 and 13, respectively) and received widespread coverage in national media outlets, which makes them particularly salient and impactful. 7The latter 6 The UESD shares some properties with other research designs commonly used in social sciences, such as the Regression Discontinuity Design, the Interrupted Time Series design, or event studies.At the same time, it has a range of unique features and particularities that warrant some specific methodological considerations.For further discussion on this, please see Muñoz et al. (2020).
ensures that, regardless of where each attack occurred, individuals from all over the UK were potentially exposed to them.Indeed, as captured by Google Trends (Figure 1), the relative search volume for the term 'terrorism' went drastically up 24 to 48 hours after the two attacks, suggesting a notable nationwide interest in these events. 8The similar patterns of the relative search volume for the term 'far right' also indicate that the two attacks were correctly classified -and perceived by the large audience -as far-right terrorism.
[Figure 1 about here] Our empirical model specification takes the following form: where 'Right Orientation' irw is a binary indicator capturing ideological self-placement at the right end of the political spectrum (values 8 or more on the 0-10 left-right scale) for individual i, living in region (government office region) r, and interviewed in survey wave w; 'Post-attack' irw is a binary indicator that takes value 1 if the individual was interviewed after the day of the attack, and 0 before the day of the attack;9 Z irw is a vector of variables that includes gender, age, age squared, level of education (low, medium, high) and employment status (employed, retired, student, and not working); λ rw represents region-by-wave fixed effects; and, u irw is an error term, clustered at the region level.10Our parameter of interest, γ, measures the effect of terrorism on political orientation, with a negative value indicating that that these were the only two 'major' far-right terrorist attacks that occurred between 2016 and 2019 and coincided with recent BES waves.As such, the choice of the two attacks is as random as their timing.exposure to terrorism drives the population away from a 'right' political stance.We employ short-range time windows before and after the attack (1-3 days' bandwidths) to substantiate the as-if random treatment assignment assumption and to minimize the possibility of other events driving the estimated effects (Nussio et al., 2019;Muñoz et al., 2020;Nussio et al., 2021). 11 An important identification concern is that our γ estimate captures pre-existing trends in political preferences, which are unrelated to the two terrorist attacks.To address this possibility, we focus on the sub-sample of survey participants who are interviewed twice (once during the attack's survey wave and once more during the previous wave) and augment Eq.
(1) with the lagged value of the dependent variable; that is, individuals' right orientation as observed in the previous wave.This set-up enables accounting for the baseline level of our outcome variable (similar to a difference-in-differences design), and also controls for biases arising from the potential omission of unobserved characteristics (Nussio, 2020;Bove et al., 2021).This also means that our estimates are relatively more conservative as a lot of variation in the outcome variable is absorbed by the lagged value.
Figure 2 shows the share of respondents classifying themselves as 'right' (based on the definition above) before and after our sampled far-right terrorist attacks.A visual inspection of this figure provides some first evidence that, after this type of attacks, people are more likely to distance themselves from the 'right' ideology.We now turn to a more systematic analysis of the causal relationship.
[Figure 2 about here] 11 See Table B.1 for summary statistics and definitions of all variables included in Eq. ( 1 3 Empirical Findings

Main results
Panel (a) of Figure 3 presents the results from estimating Eq. ( 1) using a linear probability model, based on 1, 2 and 3 days' bandwidths; that is, when we compare individuals interviewed within 1 day, 2 days and 3 days after the attacks with those interviewed within 1 day, 2 days and 3 days before the attacks, respectively.The results confirm that exposure to far-right terrorism shifts citizens' ideological self-placement away from the 'right' ideology.Specifically, the point estimates suggest that, after the attacks, 'Right Orientation' decreases by 0.02 to 0.04 points on a 0-1 scale; i.e., a decrease that amounts to 10% of the standard deviation of the variable.As expected, the estimates are less precisely estimated for the 1-day set-up due to the small sample size (low statistical power), which is one of the downsides of using very narrow bandwidths (Muñoz et al., 2020).For this reason, and also to account for the fact that the terrorism effect may take more than 24 hours to unfold, our inferences are based on the 2-and 3-days' bandwidths.Panel (b) of Figure 3 investigates the sensitivity of these (baseline) results to excluding the control variables in vector Z irw , and to augmenting Eq. ( 1) with the lagged dependent variable (LDV).The effect of terrorism remains negative, statistically significant, and stable in size across specifications, which is quite reassuring as regards to biases arising from pre-existing trends and the omission of unobserved characteristics.12 [Figure 3 about here] A threat to our identification strategy is that individuals with specific characteristics may respond to the survey at different points in time, and these characteristics may be predictive of the outcome.To further support our causal claims, we rely on entropy balancing to pre-process the data and produce covariate balance between the treatment and control groups (Hainmueller, 2012). 13As shown in Figure 4, re-weighting the sample through entropy balancing produces very similar results as in Figure 3 (panel (b)).

Further insights
Our theoretical expectation in Section 1 relies on the argument that individuals of right-wing ideology share some diluted versions of the same ideas promoted by far-right terrorists.As such, the (observed) terrorism-induced effect on 'Right Orientation' should be exclusively driven by individuals with broader right-wing positions.In order to test for this, we split the sample of respondents into a left-wing constituency and a right-wing constituency (values 0-4 and 6-10, respectively, on the left-right scale), and run the same regression set-up as in Eq. (1).To ensure that the results are not subject to post-treatment bias, we create the two groups based on information from the previous survey wave.The corresponding evidence, presented in Figure 5, lends support to the above interpretation: the treatment estimates for the right-wing constituency (panel (b)) are highly statistically significant and twice as large (compared to the baseline results), whereas those for the left-wing constituency (panel (a)) are very close to zero and fail to reach statistical significance.
[Figure 5 about here] We next examine the impact of far-right terrorism on attitudes which are typically associated with far-right electoral support. 14To do so, we exploit information on the respondents' 13 Table B.3 illustrates that this method corrects for all imbalances observed in our data (see Table B.2).
14 According to a survey by ICM Research (Boon, 2010), the two main reasons given by respondents to explain their support for the British National Party are the party's views of immigration and its interest in British people.
self-placement on a 1-7 'Britishness' scale and their views towards immigration, as captured by their answers to 'whether immigration enriches or undermines cultural life', 'whether immigration is good or bad for the economy' and 'whether the families of immigrants should be allowed (or not be allowed) to come to the UK' -ranging from either 1 to 7 or 1 to 10, with higher values indicating more pro-immigration attitudes.We then construct four binary indicators of right orientation -taking value 1 for the top 25 percentile of the responses in the 'Britishness' question and the bottom 25 percentile of the responses in each one of the immigration questions -and run the same analysis as before. 15The corresponding results, reported in Figure 6, suggest that exposure to far-right terrorism does not only affect individuals' self-positioning towards the right end of the political spectrum, but also makes them less likely to attest strong national identity and to report immigration scepticism. 16  These results run counter to previous studies on how (Jihadi-inspired) terrorism generates negative perceptions of immigrants (see, e.g., Legewie, 2013;Böhmelt et al., 2019;Ferrín et al., 2020).17Given the "tendency in academic scholarship, just as in popular discourse, to focus on negative attitudes towards immigrants" (Schwartz et al., 2019, p.18), our findings here highlight some of the conditions under which attitudes towards immigrants can actually become less hostile.
The results in Figure 6 can also address, to some extent, the possibility that our key outcome variable is susceptible to misreporting; i.e., survey participants falsify their preferences 15 We do not control for the LDV, as the corresponding questions were not asked in the previous waves.
Note that the effects persist when we employ composite measures based on the responses to all four questions (available in our replication files).
16 Similar effects are obtained when we explore the treatment effect on citizens' beliefs about traditional British values based on data from BES wave 13; see Figure B.6.Unfortunately, other questions that can be linked to extreme-right views (e.g., on ethnocentrism, authoritarianism and populist attitudes) are not available in the two waves of the attacks.
after the attacks to look as if they don't hold a far-right political identity.Even though it is not possible to distinguish between 'honest' and 'dishonest' answers -and rule out the possibility of social desirability being one potential mechanism18 -the fact that we find consistent effects across a range of different outcomes suggests that there are at least some genuine shifts in citizens' ideological stance following the attacks.This is also corroborated by the fact that we employ data from an online survey, a survey mode that has been shown to reduce social desirability bias in responses on sensitive topics compared to face-to-face interviews (Kreuter et al., 2008;Blinder et al., 2019).
Finally, one may ask whether exposure to far-right extremism can affect ideological scores through other channels.For instance, one may argue that, in the wake of the attacks, citizens move away from both extremes of the political spectrum and take more moderate positions -and thus converge towards more centrist attitudes.In Section C.1 of the Appendix, we experiment with alternative specifications of the outcome variable to test for the existence of other types of terrorism-induced shifts in political orientation.The estimates of 'Postattack' fail to reach statistical significance when we employ a binary indicator capturing ideological self-placement at the left end of the political spectrum and when we consider the continuous version of the outcome variable based on the full left-right scale.This suggests that neither the effect is 'across-the-board' -i.e., far-right terrorism causing similar shifts within the entire political spectrum -nor there seems to be evidence of "deradicalization" with individuals on the left taking more moderate positions.
[Figure 6 about here] political spectrum.At the same time, it does not affect the political orientation of individuals on the left.One popular theoretical perspective on how terrorism affects political beliefs is offered by the terror management theory.This theory predicts that death-related threats motivate individuals to seize upon whatever political ideology has served their needs in the past, be it liberal or conservative, thus reinforcing extant political beliefs.Our results do not seem to offer empirical support to this perspective.
As the general public has moral issues with violence tactics and is unlikely to justify violence, we argue that citizens are expected to be less likely (after the attacks) to support or identify with the goals of the perpetrator.As such, the use of violent tactics can erode the popular support for the ideology associated with the perpetrator of violence.If this is the case, citizens should also distance themselves from some of the core values of the perpetrators and their political views.As radical right-wing movements claim that nonnative elements are threatening the homogeneity of the nation state, we show that terrorist violence can actually decrease nationalistic attitudes and trigger more tolerance, rather than more hostility, towards out-groups.Previous studies have usually focused on one attitude and shown that (Islamic) extremism benefits conservative parties, increase nationalism or leads to more negative attitudes towards immigration.We find effects in the the opposite direction, and following a far-right terrorist incident, respondents are more likely to voice political views in opposition of those of the perpetrator.In other words, far-right political violence leads to a relatively wide reduction in support for central right-wing extremist attitudes.Taken together, our results demonstrate that the characteristics of perpetrators and their driving goals cannot be ignored for a more nuanced understanding of the audience's reception and consequences of terrorist attacks.
Important avenues for future research can emerge out of our work.For one, the UK lacks a viable far-right party as the radical right has traditionally been associated with failure. 19 19 There is not even consensus among academics about the correct ideological label for the UK Indepen- As this might point towards some sorts of uniqueness of the UK case, it would be interesting to investigate whether the results would be similar in countries with more successful far-right parties, like France or Germany.Our study suggests that far-right political violence could actually provoke positional shifts away from some of the most contentious political views associated with popular parties, such as the National Front in France.There are specificities to the UK case, as in all single-case studies, and we hope that future studies will be able to leverage a more extensive dataset and embed a quasi-experimental approach in a comparative design, so as to increase external validity.
dence Party (UKIP).For example, Mudde (2016) excludes UKIP from the family of "populist radical right" as UKIP's core ideology does not contain the nativism used to define the group.

Days
Relative interest 'terrorism' q q q q q q q q q q q q q q −3 −2 −1 0 1 2 3

Days
'far right' Attack: q q Jo Cox MP Finsbury Park Notes: The figure reports Google Search intensity for the keywords 'terrorism' and 'far right' within the UK three days before and after each event.Relative interest is comparable within each keyword and attack.The day with the highest search intensity takes the value 100 and all other points in the series are relative to this value.
Figure 2: Share of respondents classifying themselves as 'right' .12 .13 .14 .15 .16 .17 Mean of 'Right Orientation'  Notes: Estimates are balanced using entropy weights that match the mean, variance and skewness of covariates across the treatment and control units.See also notes of Figure 3.   4,056 (7,123); and 3,919 (6,881).Standard errors are clustered at the region level.Fat (thin) lines signify the 90% (95%) confidence interval.

8
As further evidence of the salience of the two attacks for the British public, we provide examples of national newspaper front pages covering the attacks (Figure B.3), and show that these incidents displaced attention from other key concerns and made individuals more likely to report terrorism as the top national problem (Figure B.4).
), and Figure B.5 for histograms of the outcome variable.
The figure reports the daily mean of 'Right Orientation' before and after the two sampled attacks.

Figure 5 :
Figure 5: The effect of far-right terrorism on people's right orientation: left-wing vs right-wing constituencies

Figure 6 :
Figure 6: The effect of a far-right terrorism on people's perceived national identity and attitudes about immigration