Violence against politicians is a part of politics, but experimental studies find that its effect on citizens is muted. Rozemarijn van Dijk and Joep van Lit argue those null results are nevertheless meaningful: they should push scholars to study the conditions under which political violence results in (de)mobilisation
Violence against politicians is, regrettably, a part of political life. Politicians are harassed and intimidated on social media, campaign materials are vandalised, and many politicians work under constant security protection. Sometimes this climate of hostility escalates into direct physical attacks on politicians. At its extreme, political violence is fatal. Substantial evidence shows that such experiences change how politicians behave. Experiences of violence influence the ways politicians interact with constituents, what issues they are willing to speak about publicly, and whether they choose to remain in political life.
As such, political violence (which we define as violent acts by individuals, not international violence or terrorism) is directly threatening democracy. The perpetrators intend it not only to silence its immediate targets but also to send a signal to bystanders (‘normal citizens’): stay in line, stay quiet, stay out of politics. This is the demobilising effect of political violence, inducing fear and self-censorship. Conversely, if political violence provokes outrage rather than fear, we might expect it to have a mobilising effect.
A growing body of experimental research shows that while few citizens are willing to tolerate political violence, exposure to it appears to have little or no direct effect on their political behaviour
Yet when we study how citizens respond to political violence, the results are often surprisingly muted. Across a growing body of experimental research, exposure to political violence appears to have little or no direct effect on citizens' political behaviour. Participation does not consistently decline. Attitudes do not shift in meaningful ways. Statistically, nothing much seems to happen.
Based on this, the effects of political violence appear to have very little effect on citizen behaviour. This research was careful and methodologically sophisticated, but we suspect the consistent null findings are telling us something more than ‘nothing is happening’.
We keep finding that political violence does not affect citizens, yet acceptance of political violence is always low. That sits uneasily with how people actually experience it. For example, young women report that witnessing political violence deters them from entering a political career. Party members report that violence is an obstacle to pursuing a political career. And ‘ordinary’ citizens are reluctant to post anything about politics on social media for fear of negative responses. In short, experimental studies may show no effects, but citizens themselves say otherwise.
We suggest two reasons for this discrepancy. First, the proliferation of news reporting on political violence. Citizens are pre-treated: there is no true control group of respondents who do not know about political violence. In a world saturated with news of threats, harassment, and intimidation, political violence is rarely ‘new information’ to respondents. Second, considering that pre-treatment, a single vignette probably isn't enough to push respondents to change their political behaviour – even in a comparatively low-cost environment like a survey. Violence is not a singular incident but a continuous process. This is hard to manipulate in an experimental vignette.
The important question is not whether violence against politicians affects citizens, but when, how and for whom political violence matters
Of course, these are assumptions for which we do not have empirical evidence. Yet that doesn't mean we should not use experiments to study the effects of witnessing political violence. We believe the important question is not whether violence against politicians affects citizens. That much is clear. Instead, we should focus on when, how and for whom political violence matters. We can still use experiments for that.
We believe the most fruitful way forward is to examine the contingencies determining the effects of political violence. Our research looks at victim-characteristics (partisanship, political role, gender) and hypothesises that the more similar a victim of political violence is to a bystander, the more demobilising is its effect. In contrast, however, we find that political violence has a mobilising effect. When the victim belongs to the respondent’s preferred party, the effect is pronounced, across different models. Respondents are more willing to campaign, donate to a party, sign a petition, post on social media, join a protest, or become (more) politically active. The only apparent null effect of this partisan similarity is on respondents’ willingness to discuss politics. The intercept for this, however, was already substantially higher than for the other activities.
Future research could extend this contingency-based approach. One obvious avenue concerns perpetrator characteristics: who commits the violence, and with what motives? Another concerns the form of violence, which may range from online harassment to physical attacks. Context also matters: the same incident may have different implications depending on the political climate, levels of polarisation, or trust in institutions.
The way citizens respond to political violence as bystanders is as individualistic as it is influenced by structural factors
Equally important are the mechanisms through which political violence affects citizens. Violence can trigger fear and anxiety, but also anger, solidarity, or moral outrage. These are emotions with very different (de)mobilising consequences which, in the aggregate, might cancel each other out. Because of this, we need to pay attention to individual-level characteristics, which matter perhaps even more than context, form, or perpetrator characteristics. The way citizens respond to political violence as bystanders is, we find, as much individualistic as it is influenced by structural factors. That does not mean we cannot learn from experimental research or aggregate effect sizes, but we should interpret them carefully with an eye towards the individual.
If we focus only on whether political violence has an effect on citizens, the answer may seem to be ‘no’. But this is misleading. Political violence does matter, just not in uniform or easily detectable ways. Its effects are contingent and mediated.
The question is no longer whether political violence affects citizens, but under which conditions, through which emotions, and for whom.
