How Are Immigration and Terrorism Related? An Analysis of Right- and Left-Wing Terrorism in Western Europe, 1980–2004

Abstract

In this article, I examine the relationship between migration and terrorism in Western European countries from 1980 to 2004. I find that an increase in migration is positively related to an increase in terrorism, but only right-wing terrorism. Immigration has no effect on left-wing terrorism or other non-right-wing terrorism. I also examine the effect of incoming refugees on terrorism and find similar results. I argue that these population flows increase terrorism in part because they aggravate the grievances of those on the radical right. To provide empirical support for this mechanism, I conduct a subnational analysis of right-wing terrorism in Germany. For German states, the percentage of foreign-born immigrants is a bigger predictor of anti-immigrant violence than economic variables such as employment or trade levels. The flow of immigrants from outside of Europe is also positively related with right-wing terror, while no relationship exists for intra-European migration. This analysis serves to qualify the study of terrorism as a strategic choice by showing that increased antipathy toward an out-group, rather than a changing strategic environment, explains variation in levels of terrorism, at least among liberal democracies.

terrorism, migration, Europe, immigration

Introduction

How do immigration flows affect the level of terrorism in a country? The migration-security nexus has received a large amount of attention in both the scholarly and public arenas, yet few studies have examined the relationship between migration and terrorism. While it is an open question as to whether migration and terrorism are causally linked, even less is known about the possible mechanisms by which migration contributes to terrorism. The public debate in Western Europe and the United States has focused largely on border security and ensuring that would-be terrorists do not infiltrate the homeland. This article argues that other mechanisms are at play. Focusing on domestic terror attacks in Western Europe, I find that, indeed, an increase in immigration flows to a country is related to an increase in the number of domestic terrorist attacks that a country experiences. However, this increase is driven almost entirely by an increase in the number of right-wing attacks. No relationship exists for other types of terrorism.

I draw on these empirical results and the body of research in both sociology and political science on the determinants of anti-immigrant attitudes in industrialized societies to argue that, contrary to existing literature and public discourse, migration does not cause domestic terrorism because of the migrants themselves, but because of reactionary elements within the host country. Much of the existing research finds that anti-immigrant attitudes are driven not by national policies but rather by a combination of local interactions and the prevalence of nationalist rhetoric. I theorize that an increase in the flow of migrants increases the saliency of migration as a political issue among the right-wing, strengthens competition between immigrant and native communities in Western Europe, and increases the grievances of right-wing and nationalist groups. Unlike hate groups, which commit hate crimes by attacking immigrants, these right-wing terrorist groups are explicitly political and often respond to increased migration flows with attacks against the state and civilians.

The theorized causes of terrorism are varied, but most modern research has focused on the strategic, economic, or ideological factors that contribute to increased terrorist activity. This article returns to examining the grievances—whether legitimate or not—that terrorist groups harbor and how that affects their behavior. While terrorism is often the result of a strategic choice that is believed to help further the tangible political goals of the terrorist group, this may apply less to politically fringe elements in mature democracies. These groups share few immediately realizable political goals, are rarely politically disenfranchised, and possess chauvinist ideologies where the primary enemy is not the state, but other people. An increase in migration both exacerbates the grievances among the far right and increases the saliency of migration as a political issue. These right-wing groups respond by committing terrorist attacks.

To provide empirical support for this argument, I present a number of statistical models that show a practically and statistically significant relationship between an increase in migration flows and an increase in the number of right-wing terrorist attacks in Western Europe during the period from 1980 to 2004. These models cannot test whether the mechanism linking migration and domestic terrorism is driven by economic grievances or by a general antipathy to immigrants. To do this, I show that an increase in refugees—a population that is on average less likely to compete for the same jobs as the native population—produces a similar pattern. In addition, I present models where immigration is disaggregated by the source country and find that only migration from outside of Europe is linked to terrorism. I also conduct a subnational analysis of right-wing terrorism in sixteen German states from 2008 to 2015. These results show that the percentage of the foreign-born population is a bigger predictor of the number of right-wing attacks than economic variables such as unemployment or trade. This subnational analysis avoids the problems posed by unobserved heterogeneity among nation-states and provides strong support for the hypothesized causal mechanism.

My analysis shows that the causes of an increase in domestic terrorist activity depends on the ideological nature of the group and the goals they are trying to advance. By showing that social and political grievances matter more than economic competition in explaining right-wing terrorism, I suggest that tempering anti-immigrant attitudes, rather than providing economic security, may be a more effective strategy for policymakers looking to combat right-wing terrorism. In addition, these results show that fears about the potential terrorist threat posed by incoming refugees and migrants among the public in the United States and Western Europe are misplaced.

The rest of the article proceeds as follows. I begin by engaging with the broad literature on migration and security, after which I lay out my theoretical argument with empirical support from the sociological and political science research on attitudes toward immigrants in Western Europe. I then describe my research design, data collection, model specification, and estimation strategies. Finally, I present the results and conduct robustness checks.

Migration, Security, and Terrorism

There are four broad literatures that this article engages with: the relationship between population flows and terrorism, the relationship between migration and security, the determinants of public opinion on attitudes toward immigration, and patterns of terrorist attacks in Western Europe.

Population Flows and Terrorism

A number of works have focused on how international population flows are related to terrorist attacks by making it less costly to commit a terrorist attack or by providing incentives for terrorist groups to do so. Choi and Salehyan (2013) argue that the existence of refugees invites terrorist activity because areas with a substantial concentration of refugees have higher levels of lootable humanitarian aid. Choi and Piazza (2016) link internally displaced persons and suicide terrorism via human rights abuses. Milton, Spencer, and Findley (2013) show how refugee flows into a country cause the country to be the target of terrorist attacks. Bove and Böhmelt (2016) connect the inflow of migrants from “terrorist-prone states” to the diffusion of terrorism. They argue that when migrants from these states move abroad, terrorist groups exploit these new connections to spread terrorism to the host country. What is left unexplored, however, is how population flows change the behavior and motivations of terrorist groups operating in the country receiving the incoming population.¹

The Migration-Security Nexus

Those examining the migration-security nexus conceive of migration as a security issue. Studies of migration among international relations scholars became more common with the end of the Cold War and debates about broadening security to include human security (Huysmans and Squire 2016). Analysts of human security and migration study how migration affects the security of the individual. More traditional security studies scholars examine the effects of migration on the security of the state. The latter group of scholars consider the state's migration policies relevant to its security, for example, by changing the level of skilled or unskilled labor or by destabilizing domestic society.

While not all security studies scholars view migration as having a negative impact on a state's security, a particular strand of this literature argues that high levels of immigration can pose a national security threat to the host country. Fears about open borders and infiltration from abroad have been a concern of policy- makers in the United States and Europe since the early 1990s (Karyotis 2007). A prominent survey lists three ways in which migration can increase global instability: “by providing resources that help to fuel internal conflicts; by providing opportunities for networks of organized crime; and by providing conduits for international terrorism” (Adamson 2006, 191). An earlier example offers a similar list: “the threat [from migration] can be an attack by armed refugees; migrants can be a threat to either country's political stability; or migrants can be perceived as a threat to the major societal values of the receiving country” (Weiner 1992, 103). A report by the Nixon Center published after 9/11 erroneously claims that “immigration and terrorism are linked; not because all immigrants are terrorists but because all, or nearly all, terrorists in the West have been immigrants” (Leiken 2004, 6).

Predictably, casting migration as a threat to the security of the state was strongly criticized by a number of scholars operating in the tradition of the Copenhagen School and critical security studies. Generally, these works argue that the security threat of migration to a state is constructed, that securitizing migration actually harms both migrants and democracy, and that policies that restrict migration are partly designed to protect the national identity of European states (Doty 1996, 1998, 2003; Huysmans 2000, 2006; D'Appollonia 2012). These scholars argue that migration policy reflects not just security concerns but also the broader values of the dominant society. Critical security studies scholars contend that viewing migration through the lens of security reifies migrants as a threat (Huysmans and Squire 2016).

While acknowledging that studying the relationship between immigration and terrorism contributes to the securitization of migration, the empirical results in this article are relevant to both types of scholars studying the migration-security nexus. The lack of any relationship between migration and non-right-wing terrorism suggests that, indeed, migrants are not threats to security; elements within the societies themselves are the threats. If migrants into Western European states are not the ones committing terrorist attacks, then the discourse on migration as a source of terrorism reflects something other than a real empirical pattern, be it anxiety, fear of an outside group, or simply posturing from politicians. This article suggests that increasing overall security (of both states and individuals) could be best achieved not by limiting international migration, but by focusing on domestic elements within Western European societies.

Public Opinion Toward Migration

If grievances against immigrants matter for terrorist attacks, what determines those grievances? Immigration into Western European countries is a highly politicized issue, especially among the extreme right wing. The large body of sociological and political science research on attitudes toward immigrants finds a strong relationship between support for right-wing parties and antipathy toward immigration. This relationship has been found to be particularly strong at the local and individual level.

Why is increased immigration related to anti-immigrant attitudes? Much of the existing research argues that this is because of issues related primarily to national or cultural identity, rather than increased economic competition. Increased immigration into Western European countries is associated with a shift rightward among a subset of the population and an increase in support for the radical right. Indeed, the number of non-Western residents in a country is a significant predictor of support for right-wing parties (Lubbers, M'erove Gijsberts, and Scheepers 2002). Semyonov, Raijman, and Gorodzeisky (2006) shows that anti-foreigner sentiment is more strongly associated with the presence of right-wing extreme parties and less so with economic opportunity or the stock of migrants. Knigge (1998) finds that rising levels of immigration, not unemployment, are correlated with latent support for extreme right-wing parties in Western Europe. Golder (2003) reports similar results for right-wing populist parties. Other works have studied individual preferences toward immigrants and reach similar conclusions (Billiet and Witte 1995; Sniderman, Hagendoorn, and Prior 2004; Sides and Citrin 2007; Rydgren 2008).

The argument and evidence presented in this article complement the substantial literature connecting attitudes toward immigrants in Western Europe with support for the far right. It also shows that opposition to immigration among more reactionary elements of the population can be translated into a hitherto unexamined political activity, that of terrorist violence.

Terrorism in Western Europe

Those examining patterns of terrorism in Western Europe have offered useful analytical categories. Jacob Ravndal (2015), for instance, offers a typology of right-wing terrorism in Western Europe. He divides right-wing terrorism in Western Europe into two groups: a large number of attacks committed by a small number of well-known groups and a larger number of attacks committed by a larger number of less well-known or unknown groups. A more recent strand of this literature has provided detailed case studies and explanations of jihadist terrorist attacks in Europe and examined the transnational origins of these attacks (Nesser 2006, 2008; Jordan 2012, 2014). Quantitative studies of terrorism in Western Europe have found that the end of the Cold War and the ideology of political leadership affects patterns of terrorism (Brockhoff, Krieger, and Meierrieks 2012), in addition to the effect of economic growth in increasing the opportunity cost of committing terrorist acts (Caruso and Schneider 2011).

Why Migration and Right-Wing Terror Are Linked

Why would terrorism and migration be related in liberal democracies? While early terrorism research focused on psychological motivations, current studies of terrorism in political science frame the choices that terrorist groups make in terms of a strategic interaction with potential supporting audiences, the government, or other terrorist groups.

With few exceptions (Brockhoff, Krieger, and Meierrieks 2012), previous research on the relationship between terrorism and democracy has rarely examined the ideological characteristics of terrorist groups. Existing works have categorized types of attacks into either transnational or domestic (Li 2005),² whether the attack is aimed at a state or civilian target (Carter 2016), or whether the event is a suicide attack (Wade and Reiter 2007). Others examine the effect of domestic terrorism on a state's incentives to launch diversionary wars (Foster 2017) or how the political orientation of the ruling party affects terrorists’ incentives (Crisman-Cox 2018).

Previous works have argued that democracy and terrorism are linked because liberalized borders allow terrorist groups to infiltrate democracies and use the associated political freedoms to organize, collaborate, and plan terrorist attacks (San-Akca 2013). This mechanism is less applicable to Western Europe's particular situation: while democracies generally have more liberal borders than autocracies, borders between countries inside and outside the Schengen Area have hardened over time while borders within the Schengen Area have liberalized. This suggests that the source of migration may be relevant for explaining patterns of terrorist attacks.³

While no direct study adopting the strategic-choice approach has been conducted on the relationship between immigration and terrorism, a few implications can be extracted from existing studies of terrorism and democracy. Chenoweth's survey summarizes the existing explanations for why terrorism and democracy are linked (2013). In addition to exploiting the freedoms that democracies provide, democracies also have higher media reporting and publicize attacks more. Certain democratic domestic institutions, such as the number of veto players, also provide incentives for attacks (Young and Dugan 2011), as does a higher sensitivity to casualties (Pape 2006). In addition, Chenoweth (2010) argues that competition among interest groups provides incentives for more extreme groups to choose terrorism as a method of advancing their interests. It is unclear how migration may increase the level of competition between terrorist groups within a democracy.

Another plausible explanation is that migration is related to terrorism via the economy. As one's economic situation worsens, the opportunity costs of violence decrease, and thus we see a negative relationship between gross domestic product (GDP) per capita and the level of terrorist activity (Humphreys and Weinstein 2008; Caruso and Schneider 2011). Since an inflow of migrants can increase the competition for jobs among similarly skilled native workers, an increase in migration can harm an individual's economic situation, which in turn makes terrorism a more attractive option. Again, this line of argument is agnostic as to the type of terrorism that an increase in migration causes. If immigration is related to terrorism because it makes certain sectors of the economy worse off, then we would expect immigration to be a significant predictor of all types of terrorism, not just of the right-wing strain.

I argue instead that migration and terrorism are linked because immigration increases the grievances among reactionary elements in liberal democracies. Over the past thirty years, the far right has witnessed a resurgence in Western Europe, and this resurgence is largely driven by resentment from native-born Europeans toward a large increase in immigrants from former colonial states and Eastern Europe (Rydgren 2007). Far-right parties have enjoyed varying success in Western Europe, but all have campaigned for restricting immigration. While increasing economic competition explains some of this antipathy, by far the most important predictor of individual support of the far right has been the individual's opinion on immigration (Sniderman, Hagendoorn, and Prior 2004). Quite simply, since far-right terrorist groups do not want immigrants to reside in their country, they commit terrorist attacks in order to drive foreigners out. An increase in migration provides both the opportunity to commit such an attack (by providing more targets, meaning migrants) and a motive (more migrants cause more resentment).

This is not to say that economic conditions are irrelevant to right-wing terrorist activity or that there is not a complex interaction between a right-wing extremist's economic situation, antipathy toward migrants, and potential for committing terrorist attacks. The goal of this article is not to argue that these factors do not contribute to explanations of terrorist activity. Rather, the goal is to argue that the number of migrants is a more important and consistent predictor of right-wing terror than variation in economic conditions. This is consistent with public opinion scholarship, which finds that individual support for the far right is driven not by economic situation but by antipathy toward migrants, and with work on right-wing terror in the United States, which also finds that sociopolitical variables are bigger predictors of right-wing terrorism than economic variables (Piazza 2017).

Right-wing terrorism differs from hate crimes against foreigners in a number of theoretical and empirical ways. One is that not all hate crimes are violent: verbal harassment and trespassing are often included in hate crime statistics but rarely find their way into databases on terrorism because there is no physical violence or explicit threat of violence (Green, McFalls, and Smith 2001). Empirically, Forest et al. (2012) find that, for the United States, there is a weak negative bivariate correlation between right-wing terrorism and right-wing hate crime and that right-wing hate crimes do not predict acts of right-wing terrorism. In addition, there are considerable differences between the perpetrators of hate crimes and terrorism. Hate crimes are often the spontaneous acts of young uneducated males, while terrorist attacks are more likely to be planned, part of a larger political campaign, and committed by organized political groups (LaFree and Dugan 2004).

More importantly, the data analyzed in this article are not limited to right-wing terrorist attacks against immigrants, but include all terrorist attacks committed by right-wing groups. Many of the attacks include actions that bear little resemblance to hate crimes, such as mail bombs sent to proimmigration journalists and assassination attempts against proimmigration politicians. While the intention of any one act cannot be definitively known, I assume that these acts are intended to publicize the agenda of these right-wing groups. Restricting immigration is often one of the most salient aspects of the agenda of the far right in Western Europe. Moreover, I sidestep the issue of how to classify groups such of the Irish Republican Army (IRA) and Euskadi Ta Askatasuna (ETA), which may have an ideological orientation applicable to the left-right scale but nevertheless have much more specific goals, by allowing my classification of left- and right-wing terrorism to not be exhaustive.

The three datasets used in this article—Terrorism in Western Europe Events Data (TWEED), Domestic Terror Victims (DTV), and the Global Terrorism Database (GTD)—only include attacks with a political motive. While it is possible that each dataset includes acts that many would describe as hate crimes and not terrorism, this is unlikely to drive the results. The DTV, for example, only includes fatalities, which means that the most common forms of hate crimes—harassment, beatings, verbal threats, etc.—are by definition excluded. TWEED relies on a news digest, which provides consistency but biases the dataset to only include more notable acts of violence, thereby excluding many hate crimes. The GTD dataset, which is used only for the subnational analysis of Germany, is notably broad with a less restrictive inclusion criteria. The ideological orientation of attacks in the GTD that are used in this article are coded by the author. These events include a small number of nonlethal attacks against individual immigrants. While some may view these as hate crimes, they were often attributed to organized groups or were accompanied by a political message, both of which imply a larger political agenda. Ultimately, however, categorizing edge cases of violent acts as hate crimes or acts of terrorism is difficult, and the main conceptual difference between terrorism and other acts of violence—terrorism is motivated by politics—is one shared by the three datasets examined here. While there is a negative correlation between hate crimes and right-wing terror in the United States, how the correlation between these two variables changes spatially and temporally is a topic for future research (Forest et al. 2012).

The evidence presented below suggests that terrorist groups operating in liberal democracies increase the number of attacks when their core political grievance is exacerbated. Piazza (2017) finds a similar pattern for right-wing terrorism in the United States, where abortion rates and the share of women participating in the labor force are strong predictors of right-wing attacks at the state level. Left-wing, nationalist, and separatists groups do not respond to increasing levels of immigration, partly because this is not an issue they are concerned with. For right-wing groups in Western Europe, immigration from within and outside Europe has remained the most salient issue on their agenda for decades.

The first step in analysis is to see if there is any statistical relationship between migration flows and the total number of terrorist attacks, in accordance with the findings of previous research (Bove and Böhmelt 2016). In order to do so, I test Hypothesis 1. Since immigration is a highly salient issue for right-wing terrorist groups, but is relatively less important for all other types of terrorist groups, then we should see a positive relationship between right-wing terrorism and immigration and no relationship between any other type of terrorism and immigration. This generates Hypothesis 2. If migration increases terrorism primarily by worsening the overall economic situation of terrorists and potential terrorists, and decreases the opportunity costs of committing attacks, then we would expect a positive relationship between migration and all types of terror, including left-wing and non-right-wing terror, which is also reflected in Hypothesis 3.

 

H1:

Migration is positively related to an increase in the total number of terrorist attacks.

 

H2:

Migration is positively related to an increase in right-wing terrorism.

 

H3:

Migration is positively related to an increase in left-wing and non-right-wing terrorism.

What is the theorized mechanism linking right-wing terror and immigration? If immigration encourages right-wing terrorism by increasing grievances and the salience of immigration as a political issue, then it is likely that the presence of migrants at the local level is a significant predictor of the number of terrorist attacks. However, if immigration encourages terror through economic conditions, such as depressing wages and generally increasing labor market competition, then these economic variables would also be significant predictors. To adjudicate between these mechanisms, I test Hypothesis 4 and Hypothesis 5.

 

H4:

The share of foreign population is positively and significantly related to right-wing terrorism at the local level.

 

H5:

Economic conditions are positively and significantly related to right-wing terrorism at the local level.

Research Design, Data, and Estimation

Dependent Variable

The key dependent variable in my study is the number of terrorist attacks that occur in a country during a given year. The data for these attacks comes from the TWEED. TWEED defines terrorism as “a form of violence that uses targets of violence in an indirect way in order to influence third-party audiences” (Engene 2007, 112). The TWEED dataset only includes terrorist attacks that are internal to Western Europe, which includes both international and domestic terrorism. So, for example, an IRA attack in Spain is included in TWEED, but the attacks during the Munich Olympics by the Palestinian group Black September are not included. This is appropriate for my study since I am not concerned with whether lax border control laws increase terrorism (which would have a positive effect on international terrorism), but rather if an increase in the number of migrants that move to and reside in a country increase the level of terrorism within that country.

TWEED also codes terrorist attacks by their ideological orientation. The orientation categories include: right-wing extremist, left-wing extremist, ethnic/regionalist, environmental, anti-militarist, international solidarity, state defense, and other. To test whether immigration is related to an increase in right-wing or left-wing terrorism, I use the labels provided by TWEED to generate the dependent variable. In addition, many of the models that are estimated below use the total number of non-right-wing attacks as a dependent variable. This is simply the difference between the number of all attacks and the number of right-wing attacks.

TWEED is preferred to other datasets that include domestic terrorism (like the GTD) for a number of reasons. The most important reason is that Tweed contains codings indicating the ideology of the actors who committed the attack, which is a key component of my research question. No such coding exists for GTD—indeed, as Ravndal has written, “it is virtually impossible to effectively distinguish right-wing attacks” from other incidents in the GTD dataset (Ravndal 2016, 2).⁴ TWEED is also specific to the countries examined in this study. Since TWEED only covers Western Europe from 1950 to 2004, the coding rules do not have to wrestle with the problem of coding attacks against civilians in the context of a civil or interstate war as terrorism. For example, in the context of a civil war, bombings, explosions, and rocket attacks could all be part of a conventional military campaign. Whether or not these actions are coded as terrorism depends on both the political context and the perceived intention of those launching the attacks. The bombing of a building in a failed state during the middle of a civil war may or may not be terrorism, but a bombing in postwar Paris most likely is. Moreover, all the data from TWEED comes from a single source: Keesing's Record of World Events (Engene 2007). This helps ensure continuity in the reporting of terrorism over time (since the same source is used for attacks that occur in the 1950s and in the 1990s) and space (since Keesing's reports are on all of Western Europe).

As a robustness check, I rerun the main models with different measures of the dependent variable using the RTV dataset and the DTV dataset. This produces the same results, which suggests that the results are not driven by the idiosyncrasies in coding rules or noise in the data.

The dependent variable used in the first batch of statistical models is the number of attacks committed in a given country-year. I use four main measures: all terrorism, right-wing terrorism, left-wing terrorism, and all forms of non-right-wing terrorism.

Independent Variable

The independent variable is the number of immigrants that enter a country during a given year. The data are from the United Nations Population Division, which utilizes statistics reported by receiving countries (United Nations Department of Economic and Social Affairs Population Division 2015). The number of immigrants is composed of the total number of foreigners who enter a country and reside there.⁵ This data covers the period 1980–2013, but there is significant missing data for a number of countries. For France, which is a key country in my study since it experiences variation in the level of terrorism and the amount of migration, I impute the missing data from the data on net migration. This is a reliable proxy for the total immigration, since comparatively few residents of France emigrate. Nevertheless, the results are robust if all observations for France are dropped. All other missing data is assumed to be missing at random, which will only affect the uncertainty of the estimates and not their point estimates.⁶

Since the distribution of migrants per year is right-skewed, I take the log of the total number of migrants. I also lag migration by one year to account for potential sources of endogeneity, since there exists evidence that terrorism can discourage migration (Dreher, Krieger, and Meierrieks 2011).

Controls

By restricting the dataset to similar Western European states that experience varying level of immigration and varying levels of terrorism, I reduce the need to introduce a large number of covariates to control for unobserved heterogeneity (Achen 2005). Omitted variable bias is still a problem, however, so I include three time-varying covariates: the log of GDP per capita, the log of the total population, and the political ideology of the current government. GDP is included because it is known to be a strong predictor of terrorism, and a robust economy may attract new migrants. Scores for political ideology come from the Comparative Political Data Set and are measured in two ways: one is the percentage share of seats held by the right-wing party, and the other is a categorical variable that codes for the ideology of the ruling party (Armingeon et al. 2018). The political climate may affect both the level of terrorism and migration policies (Koch and Cranmer 2007). Population is another variable that is related to both the level of terrorism in a country and may also proxy for out-migration or other demographic changes that affect immigration. Both the log of population and the log of GDP are lagged by one year and come from the Penn World Tables (Feenstra, Inklaar and Timmer 2015).

Since I am also trying to measure the relative weight that economic variables and immigration levels have on terrorism, I also estimate models where the dependent variable is right-wing terrorism and a larger number of economic covariates are included. These include the number of employed persons, the logged total level of trade, the household consumption, and the exchange rate. Data on employment levels comes from Eurostat, while other economic data are sourced from the Penn World Tables (Feenstra, Inklaar and Timmer 2015). Descriptive statistics of the data are in Table 1.

Table 1.

Descriptive statistics

Statistic	N	Mean	St. dev.	Min	Max
Total attacks	271	14.900	62.966	0	799
Left-wing attacks	271	1.443	4.118	0	37
Right-wing attacks	271	12.875	62.436	0	799
Non-right-wing attacks	271	2.026	4.557	0	38
Migration (log)	213	5.041	1.134	2.241	7.409
Right-wing seats	271	27.985	34.302	0.000	100.000
GDP	271	13.657	0.895	11.791	15.070
Population	271	3.464	0.800	2.253	4.407
Total trade	226	12.483	1.116	9.477	14.518
Number employed	255	2.688	0.753	1.526	3.747
Hours worked	255	1,713.525	225.168	1,381.180	2,341.760
Human capital index	255	2.723	0.293	2.011	3.325
Real consumption per household	255	13.224	0.890	11.394	14.565

Estimation

My unit of analysis is the country-year, and since my dependent variable is the number of terrorist attacks, I use a negative binomial model to estimate the effect of the covariates on the number of attacks. The negative binomial model is a gamma mixture of Poisson distributions and is commonly used in much of the terrorism literature, since attacks are measured by positive integers and are very frequently overdispersed.

Results

The main results are displayed in Table 2. For Model 1, the log of migration is positively related to an increase in the total number of terrorist attacks, providing evidence for Hypothesis 1. Model 2 displays a stronger relationship between the log of migration and the dependent variable (number of right-wing attacks), again providing evidence for Hypotheses 2. Interestingly, there is a negative relationship between migration and the number of left-wing attacks. However, the Akaike information criterion for this model is much lower than the other models, which likely is a result of the smaller number of left-wing attacks in Europe overall. Model 4 presents more reliable results and shows that there is no significant relationship between non-right-wing terrorism and migration levels, which does not support Hypothesis 3. Overall, the models in Table 2 provide support for Hypotheses 1 and 2 and no support for Hypothesis 3.

Table 2.

The effect of the level of immigration on terrorist attacks

	Dependent variable:
	All attacks	Right-wing attacks	Left-wing attacks	All non-right-wing attacks
	(1)	(2)	(3)	(4)
Log migration	0.641∗∗∗	0.904∗∗∗	−1.018∗	−0.568
	(0.244)	(0.327)	(0.558)	(0.398)
Right-wing seats	−0.004	−0.010∗∗	0.013∗	−0.003
	(0.003)	(0.004)	(0.007)	(0.005)
GDP	−0.011	−3.473	1.659	3.942
	(2.405)	(2.889)	(6.027)	(4.182)
Population	−17.126∗∗	8.970	−37.000∗∗	−30.757∗∗
	(12.200)	(17.290)	(12.012)	(8.030)
Observations	149	149	149	149
log likelihood	−407.463	−309.662	−152.982	−222.682
θ	2.038∗∗∗ (0.336)	1.795∗∗∗ (0.356)	1.185∗∗∗ (0.387)	1.352∗∗∗ (0.340)
Akaike inf. crit.	882.927	687.325	373.965	513.363

Notes: (1) Statistical significance: p < 0.1; ^∗p < 0.05; ^∗∗p < 0.01. (2) This table shows the results of four negative binomial models with different dependent variables for each type of attack. Each model includes country and year fixed effects. The results show that migration is positively related to an increase in both all types of terrorist attacks and right-wing terrorist attacks. Given that there is no positive relationship for all non-right-wing attacks, the positive result for migration's effect on all terrorist attacks is driven almost entirely by attacks by right-wing groups.

There is no significant relationship between GDP levels and the number of terrorist attacks. This is in accordance with much of the research on terrorism and economic development. Interestingly, the share of right-wing seats is significant for both the number of right-wing attacks and left-wing attacks–albeit in opposite directions. When the right-wing has a strong presence in government, right-wing attacks become less frequent, and vice-versa for the left. This suggests that terrorist groups reduce their number of attacks when the government represents their preferences. In addition, it complicates the finding of (Crisman-Cox 2018) that terrorist groups are more likely to attack when right-wing parties hold power, since this pattern does not hold for left-wing terrorism.

Table 2 shows the utility of disaggregating terrorist attacks by ideology. Model 1 reveals a positive and significant relationship between migration and the number of terrorist attacks.⁷ Separating the dependent variable by ideology shows that this result is driven by the strong positive relationship between right-wing attacks and migration. When the dependent variable excludes right-wing attacks, there is no relationship between migration and terror.

Table 3 shows the results of a number of models wherein the dependent variable is the number of right-wing attacks and economic covariates are included. We may observe a positive relationship between migration and right-wing terror if migration worsens the economic situation of terrorists and incentivizes them to commit terrorist attacks. The purpose of these models is to see if there is an independent effect of migration on right-wing terrorism when economic conditions are controlled for.

Table 3.

The effect of immigration on right-wing attacks, with additional controls

	Dependent variable:
	Right-wing attacks
	(1)	(2)	(3)	(4)	(5)
Log migration	1.073∗∗∗	0.912∗∗∗	0.725∗∗	0.769∗∗	0.715∗
	(0.338)	(0.315)	(0.336)	(0.339)	(0.366)
Right-wing seats	−0.007∗	−0.018∗∗∗	−0.022∗∗∗	−0.020∗∗∗	−0.021∗∗∗
	(0.004)	(0.004)	(0.005)	(0.005)	(0.005)
GDP	0.421	−14.057∗∗∗	−18.056∗∗∗	−29.202∗∗∗	−28.228∗∗∗
	(3.402)	(4.132)	(4.686)	(8.728)	(8.913)
Population	24.080∗	0.317	−21.850	−27.982	−22.626
	(13.585)	(12.300)	(18.604)	(19.003)	(21.501)
Number employed (millions)	−9.792∗∗		11.620	10.600	9.601
	(4.776)		(7.205)	(7.200)	(7.414)
Total trade		6.618∗∗∗	6.453∗∗∗	7.330∗∗∗	7.181∗∗∗
		(1.842)	(1.848)	(1.916)	(1.933)
Household consumption				11.276	10.851
				(7.422)	(7.431)
Exchange rate					0.842
					(1.678)
Observations	149	140	140	140	140
log likelihood	−307.974	−284.373	−283.077	−282.026	−281.908
θ	1.887∗∗∗ (0.379)	2.570∗∗∗ (0.555)	2.608∗∗∗ (0.558)	2.647∗∗∗ (0.570)	2.658∗∗∗ (0.575)
Akaike inf. crit.	685.947	638.747	638.154	638.052	639.817

Notes: (1) Statistical significance: ^∗p < 0.1; ^∗∗p < 0.05; ^∗∗∗p < 0.01. (2) This table shows the results of a number of negative binomial models with various controls for economic variables where the dependent variable is the number of right-wing attacks. Each model includes country and year fixed effects.

Both the effect of migration and the share of right-wing seats in parliament are statistically significant at conventional levels when economic variables are included. Indeed, economic variables matter as well: the coefficient for GDP is negative and significant, and the coefficient for total trade is positive and significant. This suggests that right-wing terror increases when the economy is worse and when a country participates more in foreign trade. Further research is necessary to investigate the microdynamics of this relationship and whether those who commit terrorist attacks are those harmed by trade policies. Relevant for this project, however, is whether migration exerts an independent effect on terrorism when variation in the economic situation is accounted for.

The results in Table 3 show that a one-unit increase in the log of migration is related to a 0.7 increase in the expected log count of the number of right-wing terrorist attacks. Exponentiating the coefficients in a negative binomial model produces incidence rate ratios. The incidence rate ratio for a one-unit increase in the log of migration is around 2.0. According to this model, doubling the number of migrants is related to doubling the rate of right-wing terrorist attacks. This is a practically significant relationship. However, if the rate of right-wing attacks and immigration do not increase substantially, the model predicts a practically insignificant increase. For example, if the number of migrants increases by 10 percent and the rate of terrorist attacks is around ten per year, the model predicts that there will be one more right-wing terror attack per year. What this means for practical purposes is that the impact of large inflows of migrants on right-wing terror, rather than small annual fluctuations, should be the primary concern of policy-makers.

Robustness Checks

Measurement error is a consistent problem in quantitative research on terrorism (Drakos and Gofas 2006). To ensure that these results are not sensitive to the dataset used, I run the main models showing a link between right-wing terror and migration level, but instead of using data from TWEED, I construct the dependent variable from the DTV dataset (Calle and Sánchez-Cuenca 2011). The DTV coding rules differ from TWEED in some important ways: it only includes fatalities, and only if the group or person committing the attack has the same nationality as the location of the attack. A benefit of DTV is that it includes both ideological variables and a variable indicating whether the violence was spontaneous or planned. I only include attacks that were planned and committed by neo-Nazi or right-wing groups. This helps ensure that the type of attacks that are closer to hate crimes—attacks against foreigners by unorganized youths—are not included in the dependent variable. Table A1, reported in the online supplementary appendix, produces similar results as the main findings.

Testing Mechanisms

Refugees

What is the mechanism linking migration to right-wing terror? To answer this question, I adopt two separate tactics. First, I test to see if similar patterns hold when the key independent variable is the number of refugees entering a country. In addition to being interesting in its own right, this serves other purposes. The far right in Europe is opposed not just to immigration but to restricting the flow of refugees from non-European countries. If a similar pattern holds for refugees, then this suggests that an increase in terrorism following an increase in refugee flows is driven partly by dissatisfaction with the types of people entering a country and with the state's inability to maintain the current stock of population. Moreover, a positive relationship between right-wing terror and refugee flows suggests that the mechanism linking the two is unrelated to economics, since refugees are less likely than immigrants to compete with natives for jobs (Foged and Peri 2016).

There are two other important points. One is that refugees, being in a much more compromised position than migrants seeking to permanently resettle, are less likely to choose the country of resettlement strategically (Zavodny 1999). This suggests that there will be less endogeneity between refugee flows and right-wing terror than there is for models where immigration is the independent variable. The other problem that using refugees as the independent variable solves relates to measurement error. For migration data, each state reports the number of immigrants to the United Nations (UN). For refugees, the United Nations High Commissioner for Refugees (UNHCR 2016) collects all of the data for each individual refugee and which country they enter.⁸ This helps smooth over any heterogeneity in the data collection process and also solves the missing data problem for France, which simply does not produce comprehensive data on immigration.

The results are presented in Table 4. The results are broadly similar to models where the key independent variable is the log of migration. Right-wing terrorism increases as a response to refugee flows, and there is no relationship for left-wing or non-right-wing attacks. This suggests that the mechanism linking immigration and right-wing terrorism is the right wing's reaction to migrants coming across borders, rather than any economic effect.

Table 4.

The effect of the number of refugees on terrorist attacks

	Dependent variable:
	All attacks	Right-wing attacks	Left-wing attacks	All non-right-wing attacks
	(1)	(2)	(3)	(4)
Log of refugees	0.076	0.387∗∗	−0.048	−0.082
	(0.145)	(0.197)	(0.279)	(0.192)
Right-wing seats	−0.915	2.704	−3.047	−8.998∗∗∗
	(2.520)	(3.627)	(5.542)	(3.351)
GDP	0.001	−0.003	0.010	−0.003
	(0.003)	(0.005)	(0.007)	(0.004)
Population	−20.228∗∗∗	−12.045	−16.291	−2.540
	(7.279)	(11.151)	(12.052)	(8.589)
Number employed (millions)	6.306∗	3.111	1.075	5.190
	(3.606)	(5.414)	(7.608)	(4.643)
Observations	253	253	253	253
Log likelihood	−566.295	−442.893	−212.890	−305.287
θ	1.133∗∗∗ (0.140)	0.846∗∗∗ (0.120)	0.710∗∗∗ (0.160)	0.992∗∗∗ (0.195)
Akaike inf. crit.	1,220.590	973.786	513.780	698.574

Notes: (1) Statistical significance: ^∗p < 0.1; ^∗∗p < 0.05; ^∗∗∗p < 0.01. (2) This table shows the result of negative binomial models where the key independent variable is the log of refugees entering into a country in a given year. Each model includes country and year fixed effects. The results show that the number of refugees is only a statistically significant predictor for the number of right-wing terrorist attacks.

These findings provide support for right-wing grievances as the causal mechanism linking immigration to right-wing terror. They also contribute to the research on the relationship between refugees and terrorism (Choi and Salehyan 2013; Choi and Piazza 2016). Choi and Salehyan (2013) analyze data on refugees and terrorism from 154 countries for the years 1970–2007. They find that there is a positive relationship between terrorism and refugees, with humanitarian aid being the mechanism linking the two. Greed, rather than grievance, encourages terrorists to exploit the infusion of aid following a refugee crisis via looting material goods and abducting aid workers for ransom. The finding that there is a positive relationship between refugees and right-wing terrorism and no relationship between refugees and other forms of terror suggests that a different mechanism—one related to grievances—may be at play in advanced industrialized democracies where aid workers and lootable resources are less likely to be received after an influx of refugees.

Who Are These Immigrants?

If immigrant and refugee flows increase terror by exacerbating grievances among right-wing groups, then it is likely that the country of origin for these population flows matters. Migration from culturally, religiously, and politically similar groups is less likely to spark a nativist backlash among the far right (Ivarsflaten 2005). This suggests that immigration flows from within European states are less likely to be correlated with an increase in right-wing terror than immigration flows from states outside of Europe. This section estimates a number of models that test whether this empirical pattern exists.

To do this, I use data on bilateral migration flows for all OECD countries (Organisation for Economic Co-operation and Development 2017). This data starts in the year 2000, producing an overlap of five years between the start of this dataset and the end of both terrorism datasets used earlier in the article—TWEED and the DTV.⁹ Estimating the effect of the number of emigrants from every country immigrating to European countries results in more independent variables than there are observations in the dataset. Thus, I create two measures of immigration: the total number of people immigrating to a Western European country from non-European countries and the total number of people immigrating from European countries.¹⁰ I also do the same for refugees and estimate models where the independent variable is the number of incoming refugees from European and non-European countries. (Bilateral data on refugees is more comprehensive and starts in 1980.)

These country-year data are used in a number of quasi-Poisson models where the dependent variable is the count of right-wing terrorist attacks, as measured by TWEED or the DTV.¹¹ The results are presented in Table 5. These models include statistically significant parameter estimates for the effect of both non-European migration and non-European refugees on the number of right-wing terrorist attacks. The number of European refugees and migrants is either not significant or negative and significant. These models provide further evidence that grievances toward an out-group fuel right-wing terror attacks in Western Europe.¹²

Table 5.

Effect of European and non-European immigration on right-wing attacks

	Dependent variable:
	Right-wing attacks	Right-wing attacks	Right-wing attacks	Right-wing attacks	Right-wing attacks		Right-wing attacks
	(TWEED)	(DTV)	(TWEED)	(DTV)	(TWEED)		(DTV)
	(1)	(2)	(3)	(4)	(5)	(6)	(7)	(8)
Non-European refugees	0.322∗	0.364∗∗	0.157∗	0.122
	(0.173)	(0.170)	(0.087)	(0.088)
European refugees	0.070	0.020	−0.091∗∗	−0.021
	(0.049)	(0.058)	(0.042)	(0.044)
European migrants					−1.041∗∗∗	−0.926∗∗∗	−0.141	−0.481
					(0.275)	(0.338)	(0.410)	(0.371)
Non-European migrants					0.936∗∗	0.935∗	0.370	0.823∗∗
					(0.347)	(0.467)	(0.553)	(0.372)
Right-wing seats	0.002	−0.012∗	0.009∗∗	−0.009	0.006	0.003	−0.032	−0.012
	(0.004)	(0.007)	(0.004)	(0.006)	(0.009)	(0.012)	(0.028)	(0.015)
GDP	0.247	7.693∗∗∗	10.172∗∗∗	0.088	−8.060∗∗	−4.647	−2.810	−17.440∗∗
	(1.482)	(2.209)	(1.734)	(2.120)	(3.897)	(5.945)	(4.083)	(8.146)
Population	−34.320∗∗∗	16.861	1.476	−2.375	10.998∗∗	8.482∗	3.507	−0.644
	(12.318)	(12.619)	(2.298)	(2.306)	(5.330)	(4.959)	(5.138)	(5.075)
Time trend	0.040	−0.374∗∗∗	−0.166∗∗∗	−0.042	−0.161	−0.109	−0.078	−0.154
	(0.074)	(0.065)	(0.047)	(0.045)	(0.209)	(0.240)	(0.229)	(0.179)
Total trade			−3.199∗∗∗	−0.532		−0.739		1.273
			(0.725)	(0.837)		(1.882)		(2.021)
Number employed			−8.280∗∗∗	3.843		−1.882		20.488∗∗
			(2.908)	(2.565)		(7.307)		(8.450)
Hours worked			−0.009∗∗∗	−0.001		−0.004		0.008∗
			(0.002)	(0.002)		(0.003)		(0.004)
Country fixed effects?	Yes	Yes	No	No	No	No	No	No
Observations	271	271	219	219	48	48	48	48

Notes: (1) Statistical significance: ^∗p < 0.1; ^∗∗p < 0.05; ^∗∗∗p < 0.01. (2) This table shows the results of quasi-Poisson models where total migration is divided into European and non-European migration. The same division is done for refugees. These results show that migrants and refugees from non-European states are significantly correlated with right-wing terror, while migrants and refugees originating from European states are not correlated with right-wing terror.

Subnational Probe: Germany

I also conduct a subnational analysis on Germany with the purpose of probing whether economic or demographic variables link immigration to an increase in right-wing terror. I choose Germany for two reasons: first, the data on the percentage of foreign population, economic variables, and especially anti-immigrant violence are available for a number of years at the subnational (state) level. Secondly, there is substantial variation in both the foreign population and in the level of right-wing violence. By focusing only on Germany, I can test the effect of local interactions between the right-wing and immigrant population on the increase in terrorism.

To construct the data for this analysis, I cannot rely on TWEED since the location of each terrorist attack is not always included. To compile data on the number of right-wing terrorist attacks, I rely on the oft-used GTD (LaFree and Dugan 2007). The GTD is the most comprehensive database on terrorist attacks and relies on newspaper reports to collect information on terrorist incidents. The GTD has a relatively broad definition of terrorism, with the ability to restrict the list of attacks using stricter criteria. I require that two criteria be met: the attack “must be aimed at attaining a political, economic, religious, social goal” and there must be “evidence of an intention to coerce, intimidate, or convey some other message to a larger audience (or audiences) than the immediate victims.” For Germany during the period 2008–2015, this results in a total of eighty-three terrorist incidents.

Since the GTD does not include the ideological orientation of the attacker, I manually code each attack as right-wing or not. I code an attack as right-wing if the attack was carried out by a neo-Nazi or right-wing extremist group. In addition, I code an attack as right-wing if it is directed at refugee-related targets. The majority of these attacks are targeted at refugee housing or shelters, while others are directed at proimmigration politicians. Other attacks include setting fire to the car of refugee initiative supporters and the throwing of a Molotov cocktail at the Parliament building in Berlin, which was later claimed by a right-wing group. The vast majority of attacks during this time period—sixty out of eighty-three—are coded as right wing according to these rules.

There are sixteen federal states (Länder) in Germany. The GTD includes information on the location of each attack. I match the location and number of attacks in each state to construct a dataset where the state-year is the unit of analysis. The key independent variable is the percent share of foreign population. Ideally more fine-grained data on the location and number of refugees and immigrants would be available, but the share of foreign population serves as a reliable proxy. All models include state fixed effects to control for unobservable heterogeneity across states, while some models include year fixed effects. The time varying economic covariates include the log of employed persons, the log of total imports, and the tax revenue of the state. All economic data come from the federal statistics of Germany. Maps showing the distribution of attacks and the share of foreign population are presented in Figure 1.

Figure 1.

Right-Wing Attacks and Foreign Population in Germany Notes: These maps show the spatial distribution of the total number of right-wing attacks that occurred in Germany from 2008 to 2015 and the average share of the foreign population during the same time period for each German state.

Since the count data for terrorist attacks used in these models exhibit a variance that is relatively close to their mean, I estimate a quasi-Poisson model. The results are shown in Table 6. An increase in the share of foreign population within a state is positively and significantly related to an increase in the number of right-wing terrorist attacks. Only imports are a significant economic predictor, and the sign is negative. These results lend support for Hypothesis 4 and do not support Hypothesis 5. Note that, for Model 5, the coefficient for foreign population is not statistically significant at conventional levels, although it remains positive. This should not be a major cause for concern, because there are only ninety-one state-year observations and the model includes both time and year fixed effects with four other covariates, which reduces the power of the significance test.

Table 6.

Effect of foreign population and employment levels on attacks by the right-wing in Germany

	Dependent variable:
	Right-wing attacks (GTD)
	(1)	(2)	(3)	(4)	(5)
Foreign Population (Percent)	10.653*	3.723**	5.900**	12.914*	3.369
	(5.513)	(1.733)	(2.651)	(6.747)	(2.490)
Number of Employed (Log)	−0.007	−0.003	−0.008	−0.010	−0.001
	(0.030)	(0.010)	(0.011)	(0.023)	(0.012)
Imports (Log)		−11.241***		−2.625	−11.673***
		(3.193)		(8.747)	(3.910)
Tax Revenues (Log)			−1.994	−1.415	0.284
			(1.221)	(5.286)	(1.444)
Year Fixed Effects?	No	Yes	Yes	No	Yes
Observations	91	91	91	91	91

Notes: (1) Statistical significance: ^∗p < 0.1; ^∗∗p < 0.05; ^∗∗∗p < 0.01. (2) This table shows the result of quasi-Poisson models where the dependent variable is the number of right-wing terrorist attacks in German states. All models include state fixed effects. The dependent variable is the number of terrorist attacks, and the years range from 2008 to 2015.

Conclusion

This article has demonstrated that in Western Europe immigration is positively related to the level of terrorism in a country, but the relationship only holds for right-wing terrorism and non-European immigration. There is no statistically significant relationship between levels of migration and other forms of terrorism, such as terrorism committed by left-wing, nationalist, or separatist groups. The analysis also demonstrates the utility of disaggregating terrorist activity by the ideological orientation of the group committing the attacks. Further research could explore whether other predictors or causes of terrorism exert similar heterogeneous effects.

I argue that the mechanism linking immigration and terror is likely related to grievances about the existence of migrants and the changing social order rather than economic competition between natives and immigrants. I support this claim two ways. I first present evidence that refugees—a group who are generally unlikely to compete with natives for jobs—also predict right-wing terror. I then conduct a subnational analysis of Germany, showing that a higher percentage of foreign-born population is a significant predictor of right-wing terror, while employment level is not. Moreover, migration is positively and significantly related to right-wing terror, even when controlling for a host of economic variables. In addition, I show that the level of non-European migration is a significant predictor of right-wing terror, while European migration is not.

Why do political groups who are not excluded from state power and are not traditionally defined as an at-risk minority commit terrorist attacks? Many existing theories about the strategic use of terrorism have little to say about what explains variation in the number of attacks by right-wing terrorist groups. Moreover, a key variable for explaining any behavior of the extreme right in Western Europe is that, in many countries, they enjoy the benefits of electoral success and share an ideological affinity with far-right and anti-immigrant political parties (Ignazi 2003). Further research is needed to explore the effects of this type of terror and its interaction with partisan democratic politics.

The results presented here have important policy implications. For governments in host countries, the most obvious implication is that restricting immigration flows may decrease terrorism in the host country, but only because of the domestic reaction to the immigrants. This suggests that, if immigration levels (especially immigration from culturally dissimilar countries) are high, governments should devote more resources to combating homegrown extremism. The subnational results from Germany suggests that the anti-immigrant reaction is a local-level phenomenon. Therefore, host countries should be aware that right-wing terrorist attacks are more likely to occur at locations like asylum shelters and less likely to occur in more prominent urban landmarks. Both human rights organizations and incoming immigrants should be aware that the potential threat of a backlash is real, but the source of this backlash is a small subset of right-wing terrorists. Perhaps most importantly, the public should recognize that, although it is possible for the levels of terrorism and migration in a country to be positively linked, one should be skeptical of arguments that this results from any terrorist activity committed by the migrants themselves.

Acknowledgements

Thanks to Page Fortna, Allison Carnegie, Tonya Putnam, Laura Resnick Samotin, and Justin Canfil for their helpful comments.

Footnotes

1

Most of these works take a similar empirical approach of using a large-time, series-cross-sectional dataset on a wide variety of countries. Many also use the same Global Terrorism Database (GTD) to construct their dependent variable (LaFree and Dugan 2007). The GTD is a rich resource on terrorist attacks but many (if not most) of the attacks listed in the GTD are what we would normally consider acts of political violence. The GTD, for example, includes both routine acts of piracy and violence occurring during the Rwandan civil war (Calle and Sánchez-Cuenca 2011).

2

See Chenoweth (2010)’s critique of this division.

3

Later in the article I examine the relationship between immigration source (inside or outside Europe) and the number of terrorist attacks.

4

Regardless, results wherein the dependent variable is the number of terrorist attacks as measured by the GTD are presented in online supplementary appendix Table A4. None of these models produce significant results. Since the GTD does not provide codings of the ideological orientation of the attacker, only the total number of attacks is used.

5

Since the data come from the countries themselves, and is not compiled by the UN, there exists a variation in the methods and definition of an immigrant for each country. Since many of the models estimated use relative levels of migration and country fixed effects, this does not pose a serious problem for inference.

6

The main models presented below are robust to using multiple imputation methods to account for the missing data.

7

This qualifies Bove and Böhmelt's (2016) finding that overall migration is related to a decrease in terror.

8

The UNHCR 2017 defines a refugee as the following: “individuals recognised under the 1951 Convention relating to the Status of Refugees; its 1967 Protocol; the 1969 OAU Convention Governing the Specific Aspects of Refugee Problems in Africa; those recognised in accordance with the UNHCR Statute; individuals granted complementary forms of protection; or those enjoying temporary protection.”

9

Reliable annual data on migration with source country information for the pre-2000 time period does not exist. The migration dataset used by Bove and Böhmelt (2016) spans the years 1960–2000. Bove and Böhmelt (2016) address this problem by linearly interpolating the data on migration for the years between censuses. This results in a dataset where the vast majority of observations are interpolated.

10

The official list of European countries is from the UN statistics division (UNSD 2018).

11

Convergence issues occurred when estimating negative binomial models, so quasi-Poisson models are preferred in this case.

12

I also categorize immigration into Europe according to whether the source country is a Muslim-majority country. I find significant positive effects for non-Muslim immigration and significant negative effects for Muslim immigration. Results are presented in online supplementary appendix Table A5. However, the coefficient estimates, while they have small standard errors, are infinitesimal and not practically significant. This is probably an artifact of the high correlation between immigration from Muslim-majority countries and non-Muslim-majority countries (0.95). Regardless, it seems that the Muslim/non-Muslim divide is not meaningful for explaining immigration's effect on right-wing terror. Further research can investigate whether this effect is constant across states and times and whether it matters at the local level.

Descriptive Statistics

• 1.1 Model results for DTV data

• 1.2 Model results for RTV data

• 1.3 Model results using anti-refugee data in Germany

• 1.4 Model results using all attacks using the GTD

• 1.5 Model results for migration from Muslim- and non-Muslim-majority countries