Military participation in the anti-Daesh coalition

The outcome to be explained is the military participation of the EU − 28 in the anti-Daesh coalition, also known as operation “Inherent Resolve”. The multilateral coalition was formed under President Obama in September 2014 (WH, 2014),^{Footnote 11} after a request from the Iraqi government. Military operations against Daesh were initially limited to Iraqi territory and subsequently expanded to Syria (in 2015–16). Combat intensity reached peaks in the summer of 2017, when the UK-based NGO “Airwars.org” reported civilian deaths due to airstrikes by the US-led coalition of up to 2,600 in Iraq and 1,300 in Syria.^{Footnote 12} Table 2 summarizes military contributions across the EU−28 (including the UK before Brexit).^{Footnote 13}

Table 2. Military contributions to the anti-Daesh coalition

Sources: Own compilation based on Drennan (Reference Drennan2014), McInnis (Reference McInnis2016), Saideman (Reference Saideman2016), information from ministries of defense, and the coalition website: http://theglobalcoalition.org (accessed October 8, 2020).

Essentially, there are three groups of countries. The first group includes those that were actively involved in the airstrikes, either by bombing targets themselves (UK, France, Belgium, Denmark, and the Netherlands) or through the provision of air support or reconnaissance operations (Germany, Italy, and Poland). Notably, contributions in this category were made at different points in time. Whereas the UK and France conducted airstrikes from September 2014 onward, Belgium, Denmark, and the Netherlands followed a month later. Italy deployed reconnaissance aircraft in November 2014, while Germany followed in similar functions in December 2015, and Poland in June 2016. The second group of countries contains those who limited their contribution to training local forces, mostly at the Kurdish Training Coordination Center (KTCC) in Erbil, but also in other locations. While all of these were small-scale training missions, some countries deployed substantial numbers (between 30 and 500 soldiers, including Spain, Hungary, Finland, Sweden, Romania, and Portugal), whereas others send only a handful of officers for the instruction of local forces (the Baltic countries and Slovenia). Finally, a third group of countries made no material contributions beyond logistics (overflight and basing rights in Greece and Cyprus) or expressions of political support (Austria, Bulgaria, Croatia, Luxembourg, and Slovakia). Two countries did not officially endorse the anti-Daesh coalition (Ireland and Malta).

Fuzzy values reflect the primary criterion of whether countries were involved in the airstrikes, with slightly lower values assigned to those who restricted their contribution to air support. While training contributions often also entailed substantial military personnel, these were assigned lower scores due to the different qualitative nature of the involvement. From the perspective of democratic politics, key questions concern the risks associated with a military deployment and the legality and legitimacy of the mandate. Hence, the fuzzy values reflect the major difference between contributing to the airstrikes (which often requires parliamentary approval) and other kinds of contributions, which are often outside the scope of parliamentary veto rights and receive substantially less attention in parliament and among the public.^{Footnote 14}

Explanatory conditions: data and calibration

The set-theoretic analysis contains five explanatory conditions: external threat (T), alliance value (A), right executive (R), parliamentary veto rights (V), and public threat perception (P). Table 3 lists the EU − 28 by relevant cabinet or government leader and provides raw data and calibrated fuzzy values for the included conditions. This section summarizes the fuzzy-set calibration of the included conditions and outcome, which is the transformation of qualitative and/or quantitative raw data into set-theoretic membership scores used for QCA (Ragin, Reference Ragin2008; Schneider and Wagemann, Reference Schneider and Wagemann2012; Mello, Reference Mello2021). Additional documentation is given in the online appendix.

Table 3. EU governments and explanatory conditions

External Threat indicates whether citizens of the respective state have gone to Syria or Iraq to join Daesh as foreign fighters. Data on foreign fighters stems from the Soufan Center and Global Strategy Network (TSC, 2017). Given the nature of the phenomenon, the numbers cited remain estimates, albeit most of these have been confirmed in later investigations (see also UNODC, 2019). For September 2015, a total of 30,000 foreign fighters from over 100 countries were estimated to be in Syria. About 5,000 of these came from EU countries. The empirical pattern is relatively clear-cut: 12 EU states had citizens who joined Daesh as foreign fighters, many of these countries also experienced deaths from jihadi-motivated terrorist attacks on their own soil.^{Footnote 15} The fuzzy-set condition external threat was calibrated so that all countries with foreign fighters receive fuzzy-set values above 0.50, whereas those without received a value of 0. Countries with 200 or more foreign fighters were considered ‘fully inside’ the set external threat.

Alliance Value reflects whether a country has specific incentives to contribute militarily, rooted in its alliance relationship. The measure rests on an aggregate index that takes into account: (1) whether the country has a formal alliance membership in NATO; (2) is considered dependent on the alliance for its security; and (3) whether it is generally regarded as following an Atlanticist foreign policy orientation. The first indicator simply distinguishes between NATO and non-NATO members. The second indicator, alliance dependence reflects the security concerns of countries like the Baltic states – Estonia, Latvia and Lithuania – as well as Poland. Due to their conflictive neighborhood with Russia, these four EU members have the greatest interest in securing a strong partnership with the USA and therefore seek to maintain a status as loyal NATO members. Finally, the third indicator takes into account countries with a strong Atlanticist foreign policy orientation such as the UK, Denmark, and the Netherlands, as apparent from the UK’s traditional ‘special relationship’ with the USA or the Danish ‘opt-out’ from the EU’s Common Foreign and Security Policy (CFSP). My estimate of Atlanticist foreign policy orientation follows prior work in its classification (Asmus et al., Reference Asmus, Everts and Isernia2004; Biehl et al., Reference Biehl, Giegerich and Jonas2013; GMF, 2014). To be considered inside the set ‘alliance value’, a country needs to effectively show two out of three indicators. Hence, based on the aggregate alliance index, countries with a score of 1.25 and above are considered rather inside the set alliance value, whereas scores of 2 and above indicate full membership and less than 0.75 full non-membership. This means that all countries that score on two or more dimensions of the index are considered fully inside the set alliance value and those that score on a single dimension are considered rather outside the set. Table 3 displays the individual scores and the resulting fuzzy values for this condition.

Right Executive refers to governments’ political positions on a left-right scale. The estimate draws on partisanship data provided by the ParlGov database (Döring and Manow, Reference Döring and Manow2018), which in turn originates from the Chapel Hill Expert Survey (CHES, Bakker et al., Reference Bakker, De Vries, Edwards, Hooghe, Jolly, Marks, Polk, Rovny, Steenbergen and Vachudová2015). Since many of the EU − 28 have coalition governments, parties’ individual left-right scores and their parliamentary seat share was used to calculate a weighted left-right score for the overall government. The general left-right variable in the CHES data runs from 0 for extreme left to 10 for extreme right. Using the direct method of calibration, this data was turned into fuzzy values where a score of 5 is a natural cross-over, resulting in fuzzy values of 0.50, and 6 and 4 were used as upper and lower boundaries (fully inside and fully outside the set ‘right executive’, respectively).

Parliamentary Veto Rights reflects a legislature’s formal right to veto military deployments. The key distinction is whether or not a parliament enjoys the right to debate and decide upon military missions before the armed forces are dispatched. If such a formal veto right exists, the respective country is coded as 1, if not it is assigned a score of 0. My estimate primarily draws upon the ParlCon data set (Wagner et al., Reference Wagner, Peters and Glahn2010) and subsequent updates (Wagner, Reference Wagner2018), as well as data from the PAKS project (Dieterich et al., Reference Dieterich, Hummel and Marschall2015).^{Footnote 16} Of the EU − 28, a total of 18 countries have parliamentary veto rights that apply to ad hoc coalitions such as the anti-Daesh coalition. Notably, some countries adopt different procedures depending on whether an operation takes place within the institutional context of NATO, CFSP, or as part of an ad hoc coalition of countries.^{Footnote 17} Several countries have also amended their constitutions in recent years, such as Spain, France, and Italy.^{Footnote 18}

Finally, High Threat Perception reflects European citizens’ attitudes towards terrorism and the foreign fighters phenomenon and whether people believed that terrorism and foreign fighters would pose increased security challenges for the EU in the coming years. My estimate draws on data from a Special Eurobarometer survey on European citizens’ attitudes towards security (EC, 2015). The survey was requested by the European Commission and conducted in March 2015, when the terrorist attacks on Charlie Hebdo in Paris, which occurred on January 7, 2015, were still fresh in people’s minds (EC, 2015). The survey collected responses from 28,083 Europeans across all EU member states, which makes it a suitable choice for this study.^{Footnote 19} On average among the EU − 28, 68% of the respondents said that the security challenge from terrorism and foreign fighters was likely to increase, with responses from individual countries ranging from 42% in Latvia and 44% in Lithuania to 79% in Germany and 80% in the Netherlands. Accordingly, countries were coded as being ‘fully inside’ the set high threat perception if 70% or more of the respondents saw an increased security challenge from terrorism and foreign fighters. The 0.50 cross-over was set at 62.5%, which effectively separates the group of countries with average to high responses (65% and higher) from those with below-average responses (60% and less). Scores below 50% were considered ‘fully outside’ the set high threat perception. Table 3 lists the raw data and calibrated fuzzy values for the included conditions. Table A1 and Figure A1 in the appendix provide additional documentation.

Set-theoretic analysis

The first stage in the analytical procedure of QCA is the testing for necessary conditions. Table 4 shows that none of the five included conditions passes the conventional threshold of 0.90 consistency (Schneider and Wagemann, Reference Schneider and Wagemann2012), and thus none can be considered necessary for the outcome of military participation in the anti-Daesh coalition. This also holds for the absence (negation) of each of the included conditions.

Table 4. Testing for necessary conditions

The second stage is the truth table analysis, which aims at identifying individual conditions or combinations of conditions that consistently lead towards the outcome. Table 5 displays the truth table for the outcome military participation in the anti-Daesh coalition (MP) and the explanatory conditions, external threat (T), alliance value (A), right executive (R), parliamentary veto rights (V), and public threat perception (P). With five conditions, this explanatory model entails 25 = 32 rows of logically possible combinations of conditions. As indicated in the table, rows 22 to 32 are omitted, because these do not contain any empirical cases. These are logical remainder rows, which can be incorporated in the minimization to derive QCA solution terms.

Table 5. Truth table for military participation in the anti-Daesh coalition

Notes: T = external threat; A = alliance value; R = right executive; V = parliamentary veto rights; P = public support; MP = Military Participation, bold cases hold membership >0.50 in the outcome.

Empty rows are omitted for presentational purposes (see online appendix for a table with logical remainders).

As measures of fit, Table 5 shows ‘consistency’, which reflects the extent to which a row (combination of conditions) is sufficient for the outcome MP. The second indicator is ‘PRI’, which refers to the ‘proportional reduction in inconsistency’. This measure can help to identify ambiguous subset relationships, which can be the case if PRI is considerably lower than consistency (see Schneider and Wagemann, Reference Schneider and Wagemann2012, 242). To minimize the truth table and to derive solution terms, a consistency cut-off point is set by the researcher. This indicates which rows are considered to be consistent enough to be included in the ensuing Boolean minimization procedure. Here, an effective consistency threshold of 0.87 is used, higher than the recommended minimum of 0.75 (Ragin, Reference Ragin2008, 46). This means that the top seven rows are included in the minimization procedure. Notably, this entails all cases that show the outcome (printed in bold in Table 5) and there are no contradictory rows with cases that show the outcome and those that do not as part of the same configuration, which meets an important criterion for the truth table analysis (cf. Rihoux and De Meur, Reference Rihoux, De Meur, Rihoux and Ragin2009).

The third stage of the set-theoretic analysis with QCA is the minimization of the truth table. Based on a Boolean minimization algorithm, the software (the ‘QCA’ package in R, Duşa, Reference Duşa2019) can derive three solution terms, which differ in how they treat logical remainders, resulting in more or less parsimonious or complex solutions. Table 1 displays the intermediate solution term for the outcome military participation in the anti-Daesh different combinations of conditions, which are listed in the left-hand rows. I follow established notation, where black circles (‘⦻’) refer to the presence of a condition and crossed-out circles (‘•’) indicate a condition’s absence. Below each path are listed the measures of fit, including coverage scores. These reflect how much of the empirical data is explained by each path. Moreover, the table lists which countries/cases are covered by which path and which of these are solely accounted for by an individual path (bold font). The lower end of the table details the total solution consistency and coverage, and the number of models derived (a single model). The intermediate solution incorporated four directional expectations, which guide the inclusion of logical remainder rows, namely the presence of an external threat (T), alliance value (A), public threat perception (P), and the absence of parliamentary veto rights (∼V) which were expected to contribute towards the outcome. Because party political incentives can cut both ways, as discussed above, no specific expectations were formulated for this condition. The complete truth table with an indication of which logical remainders were used for the intermediate solution as simplifying assumptions is documented in the online appendix (Table A2).

What do the results tell us about contributions to the anti-Daesh coalition? There are four notable findings. First, it is apparent that all solution paths contain either an external threat (Paths 1 & 2), or alliance value (Path 3), or a combination of these two conditions (Path 1). This underscores the relevance of international-level incentives for coalition contributions – as most of those countries that contributed militarily had foreign fighters in Iraq and Syria (many of these countries also experienced jihadi terrorism on their own soil), and many valued their alliance membership in NATO and their relationship with the USA. Moreover, the analysis documents the utility of a configurational perspective that integrates international- and domestic-level factors (Bennett et al., Reference Bennett, Lepgold and Unger1997; Haesebrouck, Reference Haesebrouck2018; Massie, Reference Massie2019), because neither external threat nor alliance value are sufficient on their own to bring about the outcome. Throughout the three paths, the two conditions combine with other push factors (Path 1 and Path 2) and with the absence of a constraint against military involvement (Path 2 and Path 3).

Second, the results also show that many of those countries that made meaningful military contributions do not have legislatures with an ex ante veto right over military deployments (parliamentary war powers). Precisely, it is the absence of this institutional constraint that characterizes Path 2 and Path 3 (either in combination with an external threat and public threat perception or together with alliance value, a right executive, and the absence of public threat perception, respectively). This finding broadly resonates with studies that have emphasized the relevance of parliamentary veto power on military deployments (Peters and Wagner, Reference Peters and Wagner2011; Ruys et al., Reference Ruys, Ferro and Haesebrouck2019), but which have also highlighted that there is an interaction between institutional rules, political preferences, and the context of military missions (Wagner, Reference Wagner2018, 131; Mello, Reference Mello2019, 49). Of those that participated militarily, only the legislatures in Germany and Denmark have formal veto rights over military deployments (hence these two countries are uniquely covered by Path 1 instead of Path 2), and Germany limited its involvement to air support, arguably due to political and constitutional considerations, if the parliamentary debates in the Bundestag following upon the terrorist attacks against Charlie Hebdo and the Bataclan are taken as an indicator. Indeed, the German government’s policy on the anti-Daesh coalition seems to reflect the country’s traditional reluctance when it comes to military involvement and warfighting (Brummer and Oppermann, Reference Brummer and Oppermann2016). This contrasts with Denmark, where involvement in US-led military coalitions enjoys broad political support and has become part of the country’s strategic culture (Jakobsen and Rynning, Reference Jakobsen and Rynning2019). These political preferences also help to explain why the institutional rules of formal parliamentary involvement on military deployments did not stop Denmark from engaging in airstrikes against Daesh in Syria and Iraq.

Third, the analysis shows that rightist partisanship was associated with military involvement, which resonates with the direction expected by most of the literature on party politics and security policy (Hofmann and Martill, Reference Hofmann and Martill2021; Wagner et al., Reference Wagner, Herranz-Surrallés, Kaarbo and Ostermann2017). Yet, this finding must be qualified because a right executive features in only one of the three solution paths and this combination is solely populated by Poland (Path 3). Moreover, when we examine Table 5, we also see that three countries that were involved in the air strikes did not have right executives (France, Denmark, and Italy). Based on the discussion in the theory section, this is not too surprising, because the anti-Daesh coalition was neither a classic left-of-center military mission, like a humanitarian military intervention (cf. Rathbun, Reference Rathbun2004), nor a purely right-of-center strategic use of force, but a mixture of both types of missions, where clear-cut partisan patterns are less prone to materialize, even though a recent study on EU member states’ involvement in military operations since the late 1990s found that left governments, by-and-large ‘were more inclined to participate’ in these military missions (Haesebrouck and Mello, Reference Haesebrouck and Mello2020, 579).

Finally, public threat perception is the only condition that appears in two qualitative states in the solution term (presence and absence of the respective condition). However, apart from the case of Poland, public threat perception was present among all of the cases that showed the outcome (Belgium, Denmark, France, Germany, Italy, Netherlands, and UK). As such, this pattern resonates with the expectations derived in the theory section. While Poland was considered to be rather outside the set public threat perception, this case contained three conditions that were directed towards military involvement and which thus arguably outweighed the lower level of threat perception.

Drawing on public opinion data from a Special Eurobarometer survey on public perceptions of terrorism and the foreign fighters phenomenon (EC, 2015), this paper has made a first step towards filling a gap that previous studies identified, but which had not yet been addressed with comparative data (Haesebrouck, Reference Haesebrouck2018; Saideman, Reference Saideman2016). That said, public opinion on foreign and security policy, and especially on military missions, is notoriously difficult to measure. This is one reason why many comparative studies have not taken it into account, even though it is often acknowledged as an important factor to consider. To further explore this factor, future work could focus on the development of public opinion throughout the anti-Daesh coalition, among a smaller subset of EU member states, where more fine-grained data might be available. Such an approach was taken, for instance, in a temporal comparison of public support for military involvement during the Afghanistan missions of Canada and Germany throughout the ISAF mission (Lagassé and Mello, Reference Lagassé and Mello2018).