Terrorism and affinity of nations

Terrorism is the premeditated use or threat to use violence by individuals or subnational group against noncombatants to obtain a political objective through the intimidation of an audience beyond that of the immediate victims (Enders and Sandler 2012; Hoffman 2017). The essential ingredients in this definition are the political motive, the nonstate perpetrator, and the intended general audience. Assassinations, armed attacks, bombings, or hostage taking that are intended to coerce political change or concessions from a targeted country are terrorist incidents. In contrast, violence solely for extortion purposes constitutes a criminal activity. Thus, kidnappings or bombings meant to secure ransoms, but not intended to further a political aim, are not terrorist acts. State-sponsored, but not state-conducted, terrorism is permitted by this definition.⁴ Terrorists seek a large audience in order to create a general atmosphere of anxiety, such that the public feels sufficiently at risk that it applies pressure on officeholders or rulers to concede to the political demands of the terrorists. Our definition agrees with those that underlie ITERATE and GTD from which we draw our terrorism data for the empirical tests.

An important distinction for the current study is between transnational and domestic terrorism. Based on the nationalities of its victims, targets, supporters, or perpetrators, transnational terrorism involves two or more countries (Enders et al. 2011). If, for example, a letter bomb is mailed from Brussels to Paris to induce political change there, then the letter bombing is a transnational terrorist incident. Politically motivated skyjackings that commence in one country and terminate in another constitute transnational terrorism. That also is true of a skyjacking in which the victims and/or the perpetrators are citizens from more than a single country. The January 2002 kidnapping and subsequent beheading of Daniel Pearl, a Wall Street Journal reporter, in Karachi, Pakistan, by al-Qaida-affiliated terrorists, is classified as a transnational terrorist incident.

Domestic terrorism is homegrown: the victims and perpetrators are citizens of the venue country (the scene of the attack). If, for example, a flight solely with nationals aboard is hijacked to another domestic destination for political purposes, then the flight diversion is a domestic terrorist incident. Politically inspired bombings carried out by nationals that impact other citizens of the venue country only are domestic terrorist attacks. Most terrorist acts during a struggle for independence are domestic terrorist incidents. Domestic terrorist attacks far outnumber transnational terrorist events (Enders et al. 2011). As we argue at a later point, political affinity may inspire nationals to engage in terrorist attacks at home to protest their country’s foreign policy alignment with the United States or its key allies. Even though the inspiration has a foreign flavor, the act remains domestic when it does not involve foreign nationalities in terms of victims, perpetrators, or the venue.

Next, we turn to alternative measures of political affinity, employed in the literature to gauge a country’s foreign policy alignment with, say, the United States. The US political proximity index, used by Dreher and Gassebner (2008, p. 27) and many others, is “the share of roll call votes cast in line with the United States in the UN General Assembly.” That measure varies between 1 and 0, where 1 indicates that a country’s UN General Assembly votes precisely match those of the United States on all issues put to a vote, and 0 indicates that a country’s votes depart from those of the United States on all issues put to a vote. An alternative affinity measure is the S score, put forward by Signorino and Ritter (1999). The score represents a cardinal measure of revealed foreign policy preferences based on the votes cast by a country relative to a comparison country, such as the United States. Consider the S score between country a and the United States, which is computed as one minus the absolute difference between the United States’ and country a’s vote scores, summed over all included votes and divided by the number of votes (Bailey et al. 2017, p. 433). The value of a country’s vote is 1 for a yea, 2 for abstain, and 3 for a nay. If the United States and country a vote the same on all issues, then S equals 1; if the two countries display maximal disagreement, then S equals −1. A potential drawback of both measures of political proximity or affinity is that they implicitly assume a simple correspondence between the two countries’ foreign policy preferences and how often they vote relative to one another. For the UN voting data in this paper, the S score and the UN voting share measures are perfectly correlated (see Table A2 in the Online Appendix), so that we can use either. Given what is favored in the literature, we rely on the S score as our first affinity measure.

Our other affinity index is the ideal-points measure, which uses the absolute distance between a country’s ideal (or most favored) policy position and that of the United States, some other reference country, or the average ideal policy position for a set of countries. The ideal-points affinity measure is somewhat related to the median-voter model, wherein agents’ most preferred policy positions are identified along a single dimension. For this third affinity measure, countries, whose ideal points are closer together, are viewed as displaying greater political affinity. Bailey et al. (2017) apply a spatial model to estimate single-dimensional ideal points for a state. To ensure more valid intertemporal comparisons, they match UN resolutions that possess identical positions over time on a US-held liberal world-order scale and use that information for their estimation. Their model ensures that the ideal-points estimates—in contrast to the S score—are not sensitive to agenda shifts that may change the types of issues put to a vote. The model also mitigates the impact of votes that create division among countries for reasons unrelated to the states’ foreign policy positions in relation to the Western liberal order. As a result, the Bailey et al. measure better distinguishes agenda changes from true preference changes, thereby identifying more precisely foreign policy orientation between two countries.

Over the years, a large literature has developed for which a measure of political affinity is used as either a dependent or independent variable regarding a host of political-economy questions (Bailey et al. 2017). In terms of an independent variable, Gartzke (1998) examines whether foreign policy affinity influences the likelihood of conflict. Bapat (2011) investigates if US political affinity, US-provided military aid, and other controls affect the duration of terrorists resident in aid-recipient countries. Political affinity with the United States is shown by Bapat to increase the longevity of resident terrorist groups despite US-supported counterterrorism assistance to the host country. That result follows because the United States is more tolerant of aid-recipient countries’ poor counterterrorism performances owing to their support for US-held foreign policy positions in the United Nations. The greater longevity of the terrorist groups stems from a moral-hazard problem, whereby the aid recipient has no incentive to annihilate the resident group because doing so will end its US-provided aid. In a different exercise, Dreher and Jensen (2007) show that International Monetary Fund (IMF) conditionality is relaxed for recipient countries that display enhanced political affinity with the United States in terms of UN General Assembly votes. The underlying reason is that such political affinity provides a direct benefit to the United States that motivates it to soften loan conditionally. Using political proximity as a dependent variable, Dreher and Jensen (2013) indicate that regime change can alter foreign policy directions in terms of countries’ affinity with the United States. The above studies are merely representative of more than 70 studies using political proximity as an explanatory or response variable.

Very few articles investigate political affinity as a potential cause of terrorism. Dreher and Gassebner (2008) establish a somewhat weak relationship (at the 0.10 level) between political affinity and the number of transnational terrorist incidents for 116 countries during 1975–2001, a sample that includes hardly any post-9/11 terrorist attacks. Their measure of affinity corresponds to the share of a country’s UN General Assembly votes that agree with those of the United States. That share measure cannot gauge changes in preferences over time as is possible with the ideal-points measure employed here. For the Dreher and Gassebner (2008) study, political affinity displays a much stronger relationship in affecting the average and median number of deaths from transnational terrorist attacks, sustained by countries with political affinity with the United States. In an extreme-bounds analysis (EBA) of the determinants of terrorism, Gassebner and Luechinger (2011) indicate that US affinity is a robust determinant of transnational terrorism in venue and victim countries, but not in the perpetrators’ countries. A few terrorism articles focus on other determinants of transnational terrorism, such as political instability (Campos and Gassebner 2013), government decentralization (Dreher and Fischer 2010), and foreign aid (Bapat 2011). Those articles consider only political affinity with the United States, based on the simplest measure of political proximity—a country’s voting correspondence to the United States on UN General Assembly votes.

Countries face vastly different risks of their citizens being the victims of transnational terrorist attacks. Since the start of 1968, US interests have been targeted disproportionately. Based on ITERATE data, the solid curve in Fig. 1 indicates the annual percentage of transnational terrorist attacks directed against US citizens during 1968–2014. For 1968–1990, the average percentage of US-victim attacks was 36.3%; for 1991–2014, this average percentage of US-victims attacks fell to 30.5%. Clearly, US victims attracted the lion’s share of transnational terrorist incidents, most of which occurred abroad. Next, we find and display the annual percentage of transnational terrorist attacks against citizens from the United States, France, Germany, and the United Kingdom.⁵ Those observations are represented by the dashed plot in Fig. 1. During 1968–2014, the annual percentage of transnational terrorist incidents directed at those four countries’ citizens averaged 49.6%. Both series’ annual percentages have declined since 2003.

Taken together, the four countries attract about half of all transnational terrorist attacks because of terrorists’ grievances and media-generated attention. Savun and Phillips (2009) hypothesize that foreign policy positions can create grievances that result in transnational terrorism. Surely, the four countries considered here are active in the foreign policy arena and, as such, can attract terrorist attacks. Three of the four countries are permanent members of the UN Security Council, which can dispatch troops to address civil and interstate conflicts. UN peacekeeping troop deployments also may create grievances that motivate terrorism.

Rich industrial countries with active foreign policy agendas often take measures to harden their borders against foreign terrorists. To the extent possible, these targeted countries also deploy defenses to protect targets at home. In response, terrorists often choose to attack these countries’ citizens abroad, where they are more vulnerable. Political affinity offers terrorist groups a means of lashing out at targeted countries’ values by hitting softer venue countries, espousing or following similar foreign policy positions. Terrorist attacks based on political proximity represent “attack transference” to more vulnerable countries. Such transference makes the most of terrorists’ limited resources and may have far-reaching political influence by affecting the foreign policy positions of many intimidated countries. As a consequence, affinity-based terrorist attacks may have larger effects than terrorist attacks directed at grievance-generating countries, thus making displaced attacks desirable and cost-effective from the viewpoint of the terrorists. Even a few terrorist attacks to protest political proximity can intimidate many countries into not supporting the policy positions of the United States and its three major allies. As an added bonus, political-affinity attacks are less apt to result in retaliation if the (weak) venue country cannot project its power abroad to the terrorist group’s home base or its supporters. Given the crucial foreign policy roles assumed by the United States and its three European partners, we state the following hypothesis:

A second hypothesis follows because affinity-motivated attacks often seek out US citizens. US presence in terms of US residents and US visitors likely is more visible in countries that have a strong political affinity with the United States, which raises the number of US targets for transnational terrorist groups. This hypothesis is stated in terms of transnational terrorist attacks, because incidents involving US citizens abroad necessarily are transnational in nature. Thus, we have a second hypothesis:

Domestic terrorism may morph into transnational terrorism as resident terrorist groups seek wider exposure for their attacks and grievances; thus, Palestinian terrorists (e.g., the Popular Front for the Liberation of Palestine) engaged in high-profile skyjackings to gain worldwide attention starting in 1968 (Hoffman 2017). In fact, Enders et al. (2011) demonstrate that domestic terrorism Granger-causes transnational terrorism, but the latter does not Granger-cause the former. Blomberg et al. (2011) show that most terrorist groups employ both types of terrorism, with a much larger share of domestic terrorist attacks, which are simpler logistically and less risky than transnational terrorist operations. Domestic terrorism is less risky because borders do not have to be crossed and foreign powers are not motivated to retaliate since their interests are not hit. Thus, if political affinity can result in transnational terrorism, then affinity can also generate domestic terrorism as terrorist groups diversify their attack portfolios and seek less risky attack modes for gaining attention.

Political affinity may motivate a country’s terrorist groups to resort to domestic terrorism to protest a home country’s adoption of Western liberal values. That was the case for some domestic terrorist attacks in Greece, Spain, Germany, Italy, and elsewhere when leftist terrorists constituted the dominant terrorist influence during 1968–1989 (Gaibulloev and Sandler 2019). Those groups objected to Western-style capitalism, their country’s membership in NATO, and other US-associated policies. Many of the resident terrorist groups’ attacks targeted home-country institutions, power figures, and firms. Even subsequent domestic attacks by religious fundamentalist terrorist groups in the 1990s and beyond were at times the result of backlashes against the home country’s relationships with key Western powers and their political posture. In many instances, Western powers gave aid with the implicit condition that the aid-recipient countries support the donor’s votes in the UN General Assembly (see Bapat 2011; Dreher et al. 2008).⁶ As a consequence, resident terrorist groups lashed out at their home countries owing to their Western-based political affinity, thereby giving rise to domestic terrorist attacks.

Piazza (this issue) shows that public opinion can support domestic and transnational terrorist attacks when democratic governance is adopted by countries in the Middle East. Democracy is viewed by some as representing Western liberal values that are un-Islamic (Piazza this issue). Countries introducing democratic rule often possess a political affinity with the United States and its three major allies. Terrorist backlashes to democracy thus may foster a relationship between domestic terrorism and political affinity when attacks are staged at home against domestic targets. Thus, we have a third hypothesis:

We construct a panel dataset that includes 159 countries over the 1966–2014 period, giving us a much longer dataset than the pioneering study by Dreher and Gassebner (2008). The full sample is divided into ten periods: 1966–1969, 1970–1974, 1975–1979, 1980–1984, 1985–1989, 1990–1994, 1995–1999, 2000–2004, 2005–2009, and 2010–2014. Grouping years into longer periods generates more within-country variation, which allows for more accurate fixed-effects estimates of slowly changing variables, such as political affinity. The descriptive statistics are presented in Online Appendix Table A1.

Our dependent variable is the total count of terrorist attacks in a country in a given (five-year) period. A terrorist attack is ascribed to the country, where the attack takes place. We use ITERATE transnational terrorist attacks (Mickolus et al. 2016), GTD transnational terrorist events, and GTD domestic terrorist incidents (START 2016) as alternative dependent variables. The GTD data are separated into domestic and transnational terrorist attacks following the decomposition method devised by Enders et al. (2011). In robustness analyses, we also examine the 5-year counts of ITERATE transnational incidents with casualties (injuries and deaths), GTD transnational incidents with casualties, and GTD domestic incidents with casualties. For 1966–1969, the averaged transnational terrorist attack count is based on ITERATE’s reported attacks in 1968 and 1969. The averaged domestic terrorist attack count starts in the 1970–1974 period, given GTD’s coverage only after 1969.

Two measures of political proximity or affinity between states are employed in the regressions. Our first proximity measure is constructed using data on the absolute difference between estimated ideal points of paired countries (Bailey et al. 2017). We create four dissimilarity indices by separately calculating the five-year averaged values of the absolute distance in ideal points between each sample country and France, Germany, the United Kingdom, and the United States, respectively. For example, France’s ideal-points distance index represents the difference between a sample country’s ideal policy position and that of France, with larger values indicative of greater dissimilarity between the two countries’ preferences over foreign policy. The pairwise correlations between these four variables range from 0.84 to 0.995. The strong correlations are anticipated given that the sampled countries are like-minded as they adhere to the Western liberal order. Our index of political proximity, Ideal Points Distance, is the average of those four variables. In our robustness analysis, we also examine separately the distances of the United States’ ideal points (Ideal Points Distance US) and Europe’s ideal points (Ideal Points Distance EUR). The latter is constructed by calculating the average of the dissimilarity indices of France, Germany, and the United Kingdom. The correlation between those two variables is around 0.88—see Online Appendix Table A2. The second measure of political affinity between a sample country and France, Germany, the United Kingdom, and the United States, which we label Affinity, is computed using the S score data from Bailey et al. (2017) and Voeten (2013); those measures follow our steps in constructing Ideal Points Distance.

A number of control variables are added from the literature on the determinants of terrorist attacks (e.g., Gaibulloev et al. 2017; Gassebner and Luechinger 2011). We rescale annual Polity 2 scores from the Polity IV Project’s dataset (Marshall et al. 2016), which range between − 10 (complete autocracy) and + 10 (full democracy), so that the rescaled normalized index varies between 0 and 1. Our measure of regime type, Polity2, is the averaged value of the normalized index for a country in a given five-period period. Following Gaibulloev et al. (2017), who find a non-linear impact of regime type on terrorism, we use both Polity2 and its squared value. Studies suggest that long-lasting regimes endure less terrorism (e.g., Eyerman 1998). Therefore, we obtain information on regime durability in years from the Polity IV Project’s database and compute its five-year averaged value for a given country (Durable). Political instability may lead to more terrorism. Furthermore, terrorism might be applied as a strategy in civil conflicts. The five-year averaged incidence of civil conflicts for each country, Civil War, is calculated using the major episodes of political violence dataset (Marshall 2017). The Civil War variable’s magnitude ranges between 0 and 7, with larger values indicative of a more severe conflict. To control for ethnic discrimination, we compute the average share of a country’s population that faces ethnically motivated political discrimination during a given period, which we label Discriminated POP. The information required for this variable is taken from the Ethnic Power Relations dataset (Wimmer et al. 2009, updated through 2010). Ethnic discrimination is expected to induce terrorism as the marginalized population turns to political violence to express grievances.

To account for a country’s economic conditions and population size, we use the natural logarithm of the averaged value of its GDP per capita (in constant 2010 US dollars) and the natural logarithm of the averaged size of its total population during a given period, which are labeled log(GDP/POP) and log(POP), respectively. The data for those right-hand side variables are taken from the World Development Indicators database (World Bank 2018). In the terrorism literature, the findings on the impact of income per capita on terrorism are mixed (e.g., Abadie 2006; Enders et al. 2016; Li 2005). Population and terrorism are found to be positively correlated because of higher exposure to attacks, larger pool of potential terrorist recruits, and greater challenges in preventing attacks (e.g., Eyerman 1998; Gaibulloev 2015). Globalization may either reduce terrorism by addressing terrorism risks associated with a country’s underdevelopment, or facilitate terrorism by lowering the costs of terrorists’ movements and illegal activities (Li and Schaub 2004). Moreover, the number of potential international targets in a country—e.g., tourists, foreign companies, and embassies—increases with globalization. The degree of a country’s integration into the global economy (e.g., trade volume, absence of trade barriers, and FDI flows) and its integration into the international community (e.g., number of hosted embassies, memberships in international organizations, and number of signatories to treaties) are represented by the country’s five-year averaged economic globalization index (Econ. Globalization) and its five-year averaged political globalization index (Polit. Globalization), respectively. The data are taken from Dreher (2006), updated in Dreher et al. (2008). Finally, past studies have shown that foreign aid is negatively associated with transnational terrorism (e.g., Azam and Thelen 2008, 2010; Young and Findley 2011). We compute each country’s averaged foreign aid per capita during a given period using information in the World Development Indicators (World Bank 2018) on population, net official development assistance, and official aid received (in constant 2013 US dollars).

A negative binomial model is applied to estimate our panel regressions. The conditional mean of the number of terrorist events, \( T_{it} \), in country \( i \) at time period \( t \) iswhere \( PP_{it} \) is a measure of political proximity, \( {\mathbf{X}}_{{{\mathbf{it}}}} \) is a vector of other explanatory variables, \( \mu_{i} \) denotes country fixed effects to control for country-specific unobserved heterogeneities, and \( \eta_{t} \) represents time effects to account for time-specific common shocks.

Table 1 presents the baseline negative binomial regressions that include a measure of political proximity, country dummies, and time dummies. Two alternative measures of political proximity—Ideal Points Distance (Columns 2, 4) and Affinity (Columns 3, 5)—and two measures of transnational terrorism—ITERATE (Columns 2, 3) and GTD Transnational (Columns 4, 5)—are entered. Both measures of political proximity are statistically significant across all models. The estimates of Ideal Points Distance and Affinity are negative and positive, respectively; their opposite coefficient signs are expected because closer political proximity corresponds to lower values of Ideal Points Distance and higher values of Affinity. Based on incidence-rate ratios, a one-standard-deviation increase in Ideal Points Distance (0.63) is associated with about an 18% (a 23%) reduction in the rate of ITERATE (GTD) transnational terrorist events. Thus, a country that has foreign policy preferences similar to the four major Western powers is likely to have more transnational terrorism, which supports Hypothesis 1.

We now estimate our full regression models, which expand the baseline models to include key and standard determinants of terrorism—see Table 2. Our main conclusion of positive correlation between transnational terrorism and political proximity holds. The Affinity variable is based on yea and nay votes, but our finding holds if we use S scores that incorporate abstentions (available upon request). The effects of other statistically significant variables concur with the literature. The estimates of Polity2 and its squared term are positive and negative, respectively, indicating an inverted U-shaped relationship between regime type and terrorism (Gaibulloev et al. 2017). The impact of regime durability is negative, whereas the effects of civil war and population are positive (see our earlier discussion, which anticipated such findings).

We perform a number of robustness analyses using a political proximity measure based on the absolute distance between countries’ ideal points. The results generally are similar to those reported above if we use S scores (Affinity); the Ideal Points Distance measure is less noisy and more robust, consistent with Bailey et al. (2017). First, we replace ITERATE and GTD transnational incidents with terrorist incidents with casualties. Attacks with casualties are more likely to be reported consistently by the media across all countries. This alternative terrorism variable ameliorates undercounting problems, particularly in countries where the international media is less represented. Second, European and North American countries have strong political affinity with the major Western powers. Many European countries experienced transnational terrorist attacks. Thus, we examine whether our results hold if European and North American countries are removed from the sample.

Third, we include a number of additional control variables in our regression models. Interventionist foreign policies or being allied with the United States may make a country a target of transnational terrorist attacks (Savun and Phillips 2009), because those alignments may either challenge the interests of terrorist groups or create grievances that translate into terrorist attacks. Three international affairs variables are constructed: We compute the five-year averaged count of a country’s foreign interventions (Intervention) using the information from the International Military Interventions dataset (Pearson and Bauman 1993), which has been updated through 2005 by Kisangani and Pickering (2008). We also create a dummy variable that equals one if a country has experienced an international crisis during a given five-year time period (International Crisis), and another dummy variable that equals one if a country was allied with the United States during a given five-year time period (U.S. Alliance). The data on international crises come from the International Crisis Behavior (ICB) dataset (Wilkenfeld et al. 2010), which records countries facing one or more international crises over a three-year period. The data on alliances are taken from Gibler (2009); these data are extended through 2012. Finally, we replace our economic globalization measure with five-year averaged trade as a percentage of GDP (Trade), a common indicator of trade openness, and we enter the logarithm of the five-year averaged military expenditure per capita (constant 2010 US dollars) [log(Mil.Exp/POP)], a standard control for a country’s military capacity. Those variables are constructed using information on trade, military expenditures (as percentages of GDP), and GDP per capita (in constant 2010 US dollars) from the World Development Indicators (World Bank 2018). The 2010–2014 period is dropped from this robustness analysis owing to missing observations on international interventions and crises.

Fourth, we lag the political proximity measure by one year to ameliorate potential simultaneity concerns. Fifth, to ascertain that the correlation between political proximity and terrorism is not driven solely by the affinity with the United States, we separately enter the ideal-points difference with the United States and the ideal-points difference with Europe (France, Germany, and the United Kingdom). We cannot enter both measures in the same regression because of a very high correlation (0.88) between them.

Tables 3 and 4 report the findings for the political proximity variables; the complete results are given in the Online Appendix. Our main result—a negative correlation between dissimilarity in states’ preferences and transnational terrorism—holds across all of the robustness regressions, thus supporting Hypothesis 1. Hence, political proximity is associated with greater transnational terrorism and the relationship holds using either affinity with the United States or affinity with the three European countries. Moreover, to account for a significant shift in the counterterrorism policies of the United States and its allies following 9/11, we interact Ideal Points Distance with a dummy variable for the post-2000 period. The results suggest that the impact of political affinity on terrorism is smaller after 2000. However, the effect is not statistically significant for GTD transnational terrorist events (available upon request).

Next, we examine the statistical association between domestic terrorism and political proximity. Table 5 presents the regression outputs for two measures of domestic terrorism: the number of domestic terrorist incidents (upper panel) and the number of domestic terrorist events with casualties (lower panel). The model specifications are similar to those of Tables 2, 3 and 4 with the exception of economic globalization being replaced with the trade variable in all models owing to convergence issues. In particular, for each dependent variable, Model 1 uses the main specification (see Table 2), Model 2 removes European and North American countries, Model 3 enters additional explanatory variables, which are foreign interventions, international crises, alliance with the United States, and military expenditures (see Table 4), Model 4 replaces the ideal-points measure with its one-year lagged value, and Model 5 uses the Affinity variable as a measure of political alignment. Ideal Points Distance is used in all other models. Furthermore, similar to a transnational-terrorism specification in Table 4, we enter separately the ideal-points difference with the United States and the ideal-points difference with Europe in Models 6 and 7. The results for the political affinity variables are presented in Table 5, with the full results given in the Online Appendix. Consistent with Hypothesis 3, we find a robust positive relationship between political proximity and domestic terrorism, which is represented by the negative coefficient on Ideal Points Distance and the positive coefficient on Affinity. This result is statistically significant across all specifications reported in Table 5.

Finally, we examine the effect of political proximity with the United States, measured by the ideal-points difference from the United States, on incidents involving non-US victims (Models 1, 2) and incidents involving US victims (Models 3, 4)—see Table 6. The two dependent variables are constructed by disaggregating ITERATE attacks and computing five-year counts of attacks against US victims and those against all other countries’ victims. Using ITERATE, we cannot fully identify attacks against German, French, and British victims, in contrast to attacks against US victims. The results in Table 6 show that a country with closer political proximity to the United States experiences more attacks against US and non-US targets. Those findings hold after removing North American and European countries from the sample (Models 3 and 4). The positive relationship between the political affinity with the United States and terrorist attacks against the US citizens abroad is consistent with Hypothesis 2. Political ties to the United States likely increase the number of potential US victims in a given country (e.g., businessmen, aid workers, and tourists), which, combined with a weaker security protection than in the United States, may attract international terrorists who aim to harm US citizens. Again, the findings for political proximity are shown in Table 6 and the complete results are reported in the Online Appendix.⁷

For two alternative measures of political proximity or affinity with the United States and its three primary Western allies (Britain, France, and Germany), countries are shown to be in greater peril from transnational and domestic terrorist attacks when their foreign policy positions revealed in UN General Assembly votes are more aligned with those four countries. The ideal-point measure of proximity outperforms the standard “S” affinity measure in the reported regressions. The inclusion of US primary allies in the political proximity variable is a novel feature of our analysis. The careful examination of political proximity and domestic terrorism also is a unique feature of this study. Our findings are robust to attacks with casualties, dropping North America and Europe, lagging the ideal-point measure, and adding a host of further controls. Important determinants of terrorism—population, regime durability, regime type, and civil wars—are robust controls with significant coefficients in the anticipated direction. When the proximity measure is disaggregated, countries with closer political proximity to the United States or to its three key allies attract more terrorism, so that our results are not just driven by political proximity to the United States.

A key takeaway is that being aligned with the policy positions of the United States, its principal allies, or both invites a potential cost in terms of transnational and domestic terrorist attacks. The domestic terrorism risk is novel to this study, but stems from citizens resorting to violence at home to demonstrate their displeasure with their country’s foreign policy alignments with key Western policy preferences. Such discontentment may be enhanced if the United States or its prime European allies provide foreign assistance, because citizens view their government’s Western-oriented policy positions as a betrayal of their own values. That perception may be exacerbated if the aid-recipient country then is more supportive of the aid-giving countries’ UN positions as a way of staying on good terms in order to keep the aid coming (see, e.g., Bapat 2011). As a consequence, countries must prepare for such attacks as their leaders’ foreign policy stances become more Western-oriented. Thus, counterterrorism efforts may need to be strengthened when voting patterns in the United Nations become more closely aligned with Western powers. The finding that political proximity to the United States also increases terrorist attacks on US victims abroad councils that the US government take steps to alert its citizens living in or visiting countries that are more closely aligned with US foreign policy positions to the greater risk. That is another important takeaway. Potential-target countries likewise must take measures to protect their US visitors and their own citizens from affinity-related terrorist attacks. Terrorism risks can work through circuitous routes!