Party Politics or (Supra-)National Interest? External Relations Votes in the European Parliament

Parliamentary votes on foreign and security policy have often been demon- strations of patriotism and unity. This resonates with the notion that external relations are exempted from party politics, or that politics stops at the water’s edge. Its supranational character makes the European Parliament (EP) a particularly interesting laboratory for subjecting this thesis to empirical scrutiny. Analyzing roll-call votes from 1979 to 2014, this article shows that group cohesion and coalition patterns are no different in external relations votes than in other issue areas. Members of the EP (MEPs) do not rally around a European Union ﬂag, nor do MEPs vote as national blocs in votes on foreign and security policy, trade, and development aid. Based on statistical analyses and interviews with parliamentary civil servants, it concludes that the EP stands out by having party politics dominate all business, including external relations.


Introduction
Parliamentary votes on external relations have often been demonstrations of unity. An analysis of votes in the US Congress shows that bipartisanship was more frequent when members of Congress voted on external relations as compared to domestic policy issues (Kupchan and Trubowitz 2007, 11). 1 This finding resonates with the notion that external relations are exempted from party politics, or, as Senator Arthur Vandenberg famously put it, that politics stops at the water's edge. Vandenberg's statement is primarily normative, rather than descriptive: whereas political parties may legitimately disagree over domestic politics, they should understand that toward the outside world, there is a common interest that transcends and trumps party politics. According to Kenneth Schulz (2001), a strong signal of resolve is sent to the outside world if the opposition supports the government's foreign policy, whereas a lack of such support will be interpreted as weakness. It is therefore in the community's interest to have maximum support for the government across political parties.
In this article, we further explore the notion that external relations are exempt from party politics by examining external relations roll-call votes in the European Parliament (EP). Although the European Union (EU) is not a nation-state, the "politics stops at the water's edge" logic applies because-with few exemptions from the political extremes-politicians from all member states and political parties share an interest in the EU's negotiation success vis-à-vis third parties. As argued by Schulz (2001), the chances for successful negotiations with third parties increase when government policies are supported by a broad coalition in parliament, whether or not it possesses any formal decision-making competencies.
Its supranational character makes the EP a particularly interesting laboratory. In the next section, after briefly outlining the role of the EP in the external relations of the EU, we will show that, on the one hand, supranational party politics is very much encouraged by the historical development and rules of procedure of the EP. On the other hand, the members of the EP (MEPs) are (re-)nominated by their national parties that control candidate selection, thus facilitating a stronger influence for national interests. Despite a rich body of work on voting behavior in the Parliament, scholars have not systematically investigated how MEPs vote on external relations issues. Our analysis is thus driven by three main questions: Is the level of agreement about external relations-among MEPs of the same political group, MEPs of the same national delegation, and MEPs of the parliament as a whole-different from that on other issues? Are external relations contested along a left/right axis, as shown in coalition patterns? Are there differences across time periods and different categories of external relations votes?
In the theoretical framework, we discuss whether foreign and security policy indeed forms a "special case" in the specific context of the EP while differentiating between various external relations issues, and we put forward a number of hypotheses that guide our empirical analysis. Based on all plenary roll-call votes from 1979 and 2014 and supplementary interviews, the empirical section is divided into three parts. The first compares cohesion-of the parliament as a whole, of party groups, and of national delegations-from 1979 to 2014. The second part focuses on votes from EP 6 (2004) and EP 7 (2009-two legislative terms featuring a much higher number of external relations votes-and uncovers coalition dynamics between the party groups. The final empirical part zooms in on voting behavior in specific external relations issues in EP 6 and EP 7.
We show that, in contrast to the votes in the US Congress, mentioned earlier, in the EP external relations are not exempted from party politics. MEPs do not rally around an EU flag when they vote on external relations. Nor do MEPs from the same countries reach higher levels of unity when it comes to foreign affairs. Furthermore, coalition patterns in external relations are very similar to the ones in other areas, pointing to the dominance of the left-right dimension. Our findings are also remarkably stable, with hardly any consistent differences over time or between policy areas or geographical regions. This underlines the need to engage in systematic analysis of larger voting data sets, as case studies of highly contested policy processes can exaggerate the level of parliamentary disagreement found within individual policy areas. At the same time we did uncover some variation, with especially votes on arms trade and on North America and Turkey divisive and thus requiring further investigation. Hence, we conclude that the EU's supranational parliament stands out by having party politics dominate all business, including external relations.

Mind the Context: The Increasing Role of the European Parliament
Over the years, the EP changed from a purely consultative body with members seconded from national parliaments to a directly elected EP with significant legislative, control, and budgetary powers. The Parliament shapes EU laws and the EU's budget, it is involved in the appointment of the Commission and can force it to resign. In addition, MEPs have proven remarkably inventive in pushing for more powers between intergovernmental conferences, adopting practices that have over time become the established course of action (Ripoll Servent 2018).
These developments apply also to external relations, an issue area long beyond the reach of MEPs. In line with research on voting in the US Congress (see footnote one above), we use "external relations" as an umbrella term covering all relations or policies the EU has with third countries or international organizations from trade and development aid to the development of common foreign and security policy/common security and defence policy (CFSP/CSDP). The Parliament enjoys after the Lisbon Treaty substantial legislative rights in international trade and international agreements. Laws defining the framework for common commercial policy, such as on antidumping, are adopted by co-decision procedure, and the consent of the EP is required for international agreements concluded in areas where the ordinary legislative procedure applies. Perhaps more importantly, as in other policy fields where the Parliament has gained powers, stronger legislative rights have contributed to better information rights, with the Commission-that negotiates on behalf of EU-reporting actively to the EP on trade and international agreements, both before and during the negotiations. MEPs have also exploited their new powers, notably when striking down the SWIFT Agreement in February 2010 and the Anti-Counterfeiting Trade Agreement in June 2012. Considering that the EU is the largest trading bloc, through its legislative and veto powers in international trade and other agreements, including accepting new member states to the Union, the Parliament can also wield at least indirect influence in CFSP/CSDP (Ripoll Servent 2014; Rosén 2015b).
Foreign and security policy remains decidedly more intergovernmental, with a weaker role for the supranational institutions. Essentially the Parliament debates and adopts resolutions on CFSP/CSDP and can engage in a dialogue with the EU's foreign minister (High Representative of the Union for Foreign Affairs and Security Policy who is also a Vice President of the Commission) and the other EU institutions over various CFSP/CSDP matters. However, while the EP brings much-needed transparency to CFSP/CSDP through its debating function, the main instruments of the Parliament are nonetheless more indirect in kind. Through its legislative powers in trade and development policy and its powers to strike down international agreements, the EP can exert pressure on third countries and international organizations, for example regarding human rights where the Parliament has shown notable activism (including its annual Sakharov Prize for Freedom of Thought). Another avenue of influence is money, as the EP must approve the annual CFSP budget (Smith, Huff, and Edwards 2012;Rosén 2015b). The EP also has a large number of delegations to international or regional parliamentary assemblies (such as the ACP-EU Joint Parliamentary Assembly and EuroLat) and individual third countries, through which MEPs can exercise "parliamentary diplomacy" and networking. The Parliament has thus emerged as a significant actor in EU's external relations, with its votes decisive in international trade and agreements, development policy, and accession of new member states, but in CFSP/CSDP MEPs still remain primarily as underdogs vis-à-vis the Council and national governments (e.g., Rosén 2015aRosén , 2015bStavridis and Irrera 2015;Raube and Rosén 2018).

Voting in the Parliament: Overall Patterns
The EP party system has throughout the history of the directly elected Parliament been effectively dominated by the center-right European People's Party (EPP) and the center-left Party of the European Socialists (PES; the official group name has been Socialists & Democrats (S&D) after the 2009 elections), with the liberal group (currently Renew Europe, previously ALDE) also present in the chamber since the 1950s. The groups of the greens (including the regionalists), the radical left, the conservatives, and the eurosceptics have also become "institutionalized" in the chamber since the first direct elections, although particularly the conservative and the Eurosceptical groups have been much less stable in terms of the composition and names of the groups.
Much of the research on EP party groups is based on roll-call voting data. 2 Following the "first generation" of studies, which examined limited numbers of votes (e.g., Attinà 1990;Raunio 1997), subsequent research has been far more ambitious and methodologically sophisticated. This research has produced two main findings (see Roland 2005, 2007;Hix and Høyland 2013;Bressanelli 2014;Bowler and McElroy 2015). 3 First, the party groups achieve relatively high levels of cohesion, often even above 90 percent and generally higher than national delegations. One reason for individual MEPs and national parties to vote with their group most of the time is policy influence. Cohesive action is essential for the achievement of a group's objectives, and cooperative behavior within groups helps individual MEPs to pursue their own goals. As the Parliament has acquired more legislative powers, party groups thus have a stronger incentive and need to act cohesively. Moreover, given the enormous number of amendments and final resolutions voted upon in plenary sessions, the voting cues provided by groups and particularly by group members in the responsible EP committee are an essential source of guidance for MEPs (Ringe 2010). Less institutionalized or ideologically more fragile groups, such as the radical left, often do not even try to form unitary group positions. The fact that party groups achieve higher levels of cohesion than national delegations is a strong indication that the EP is a truly supranational institution. By implication, political conflict cannot be captured by the "international relations model" (Steenbergen and Marks 2004, 5) according to which positions result from national interests. Enlargements have not really changed group cohesion levels; the two main groups continue to vote together around two-thirds or 70 percent of the time, and representatives from new member states do not defect any more from their group than average MEPs (Hix and Noury 2009;Lindstädt, Slapin, and Wielen 2012).
Studies based on European Election Study survey data and on expert surveys have produced largely similar results with respect to the levels of group cohesion (Thomassen, Noury, and Voeten 2004;Benoit 2007, 2012;Schmitt and Thomassen 2009). Research has also uncovered how various individual-level factors, such as ideological distance from the group position, proximity of elections, or the career plans of MEPs, influence party cohesion and coalition formation. Most importantly from our point of view, it indicates that when MEPs receive conflicting voting instructions from national parties and their EP groups, they are more likely to side with their national party, particularly in parties in which the leadership has more or better opportunities to punish and reward its MEPs (such as 2 There is also debate over the validity of the roll-call data. Because recorded votes represent only a sample of the totality of votes in the Parliament, the representativeness of the sample is of crucial concern (e.g., Carrubba et al. 2006;Trumm 2015;Cicchi 2016;Thierse 2016). However, a possible bias does not affect our main aim in this article, namely a comparison between external relations and other votes, because there is no reason to expect that the roll-call votes in foreign policy would be any less representative of the overall votes than in any other issue area.
3 Voting behavior in the EP has received detailed attention from scholars since the late 1990s. Here, we only discuss the main findings of this literature and those studies that feature external relations votes.
Downloaded from https://academic.oup.com/fpa/article/16/4/547/5823728 by Tampere University Library user on 20 October 2020 through more centralized candidate selection or closed lists): "Despite the fact that the parliamentary principals in the EP control important benefits-such as committee assignments and speaking time-it is the principals that control candidate selection (the national parties) who ultimately determine how MEPs behave. When the national parties in the same parliamentary group decide to vote together, the EP parties look highly cohesive. But when these parties take opposing policy positions, the cohesion of the EP parties break down" (Hix 2002, 696). This implies that when national interests are at stake, party group unity is likely to suffer. Second, the left-right dimension is the main cleavage in the Parliament, with the anti/pro-integration dimension constituting the secondary axis of competition, particularly since the start of the euro crisis (Otjes and van derVeer 2016;Blumenau and Lauderdale 2018). This means that any two party groups are more likely to vote the same way, the closer they are to another on the socioeconomic left-right dimension. For many issues, there is an absolute majority requirement (50 percent plus one MEP) that facilitates cooperation between the two main groups, EPP and PES, which between them controlled around two-thirds of the seats until the 2014 elections. Cooperation between EPP and S&D is also influenced by inter-institutional considerations, because only moderate amendments are likely to be accepted by the Council and the Commission (Kreppel 2002). In fact, the clear majority votes are passed with large super majorities, constituting thus "hurrah votes" (Bowler and McElroy 2015).

External Relations: A Special Case or Business Like Usual?
External relations are often identified as a policy area where strong collective interests overshadow party-political differences. Inter-institutional considerations provide additional incentives for overcoming partisan divisions. As argued earlier, the EP has throughout its history been the "underdog," fighting to be granted more powers. In line with the "politics stops at the water's edge" idiom whereby partypolitical squabbles are set aside in favor of a common interest, this suggests the need to act responsibly and not to endanger EU level unity that often requires lengthy bargaining in the Council. The Parliament has also consistently favored deeper integration, including in external relations, and thus it should be supportive of EU level positions. These considerations result in the following hypothesis:

H1a: Cohesion in the whole Parliament is higher in external relations votes than in other votes.
The main rationale for group formation in any legislature is that it helps likeminded legislators to achieve their policy goals, and previous research shows that the EP groups are ideologically quite cohesive. Yet whether such like-mindedness extends to external relations is not self-evident. As argued in the introductory article of this Special Issue (Raunio and Wagner 2020), especially in security and defence policy the notion of a core "national interest" that forms the "raison d'être" of the state in international politics is still widely shared. This national interest might stem from geography (such as proximity to an unfriendly neighbor), past historical events (such as experience of the Second World War), or military alliances (such as NATO). In these cases, there probably is broad partisan and societal consensus behind the national interest, and the development of CFSP/CSDP certainly has been influenced by such strongly held national preferences. Attinà (1990) showed that in the 1984-89 EP the cohesion of party groups was higher than average in international issues. Examining votes on the Gulf War, on relations with Arab countries, and on Yugoslavia, Raunio (1997, 110-111) found group cohesion to be similar when compared with his overall sample of votes. Moving to studies utilizing larger data sets, Hix, Noury, and Roland (2007, 127) compared group unity across issue areas (economic, environment, social, external, agriculture, institutional, internal EP) between 1979 and2004, concluding that in general, the substantive subject of the vote does not seem to affect the level of cohesion of the parties. Nevertheless, parties are generally less cohesive on external relations issues (such as trade, aid, security and defence policies) and internal parliamentary issues (such as the organisation of the parliament's timetable). This is not surprising. Some external relations issues probably split MEPs on national lines as well as along party lines.
On the other hand, EU member states have also key interests to defend in internal market legislation, with votes in the Parliament having strong distributive effects between and inside individual countries. These are also the core areas of the EP's law-making powers unlike in external relations where non-legislative votes still dominate, with party groups thus having a stronger need to build unitary group positions. EU's external relations, particularly CFSP/CSDP, are discussed and decided primarily in the Council and the European Council between national executives, not in the Parliament. Hence, there are good grounds for arguing any interests, be they national, party-political, or constituency-related, are in fact less "intense" in external relations votes. Another counter-argument to foreign affairs constituting a "special case" is institutional in character. As the research outlined earlier showed, party groups and committees have developed their own "ways of doing things" over the decades, with new MEPs quickly socialized into existing parliamentary norms. This line of reasoning suggests few if any differences between external relations and policy sectors.
These contradictory findings and arguments ask for a more systematic inquiry, which is guided by the following hypothesis:

H1b: The cohesion of national delegations is higher in external relations than in other votes, with party group cohesion in external relations thus lower than in other votes.
In line with our argument about national interests being more prominent in external relations, once the EP has won powers we should see increasing division in the chamber as the votes simply matter more than before, both within groups and in the chamber as a whole. Furthermore, while the clear majority of MEPs have always been pro-integrationist, ideological diversity inside the Parliament has increased due to enlargements and the politicization of European integration (Hooghe and Marks 2012), with various Eurosceptical voices also becoming gradually more numerous in the institution. This diversity should impact on all votes, including on external relations, with particularly more Eurosceptical representatives opposing EU level measures. Hence, we formulate two hypotheses about longitudinal developments: H2a: Cohesion in the whole chamber in external relations votes has decreased over time.

H2b: Party group cohesion in external relations votes has decreased over time.
With a view to coalition patterns, Raunio (1997, 112-117) found that votes on the Gulf War showed strong disagreements between the center-right and the leftist groups, while votes on Yugoslavia produced no stable coalitions among the party groups. Both sets of votes were highly contested, with a low level of unity in the chamber. Such differences were also established by Viola (2000) in her detailed analysis of the stances of EP party groups on the Gulf War and on Yugoslavian conflicts, with party groups disagreeing especially about use of force. However, these studies were based on a limited number of votes. Hix and Høyland (2013, 179-180) show that coalition patterns in external relations largely follow the overall coalition dynamics in the chamber, although trade votes displayed somewhat more competition between the EPP and PES. Examining votes on international trade from 2009 to 2013, Van den Putte, de Ville, and Orbie (2015, 60-62) find that coalition patterns follow a left-right division.
As outlined in the introductory article of this Special Issue (Raunio and Wagner 2020), these findings are in line with studies of the US Congress and select other legislatures that found the left-right dimension to explain voting patterns on trade issues, with party preferences driven by constituency interests. Center-right parties are more supportive of free trade than leftist parties that are more willing to use protectionist measures to safeguard social, consumer, industrial, or environmental interests. Leftist MPs are more likely to support foreign aid and funding of international financial institutions, while populist and radical right parties are particularly critical of development aid (e.g., Verdier 1994;Thérien and Noël 2000;Hiscox 2002;Milner and Judkins 2004;Broz and Hawes 2006;Broz 2011;Milner and Tingley 2015). More recently in Europe much of the criticism toward the transatlantic trade and investment partnership (TTIP) has come from the political left, with the center-right more in favor of free trade with the United States (Jančić 2017). These votes on trade are in the EP primarily about legislation which means that national parties need to take the views of their constituents more seriously while party groups need often to build absolute majorities and with group leaders having stronger incentives to ensure that all MEPs follow the group line.
While coalitions on trade and development can be expected to follow the leftright dimension, it is less clear whether how party groups vote in other types of external relations questions can be deducted from their positions on either the leftright dimension or any other cleavage. The bulk of the existing research has focused on the specific question of the use of force, showing quite consistently that "hawks" are more often found among center-right MPs and "doves" on the left (see the articles by Haesebrouck and Mello 2020;Raunio and Wagner 2020 in this Special Issue). In Europe parties on the right, with the exception of the radical right, are more supportive of military missions than leftist parties (Wagner et al. 2018), and center-right governments are also more likely to enter into military conflicts (e.g., Palmer, London, and Regan 2004;Clare 2010;Mello 2014). These considerations result in the following hypothesis: H3a: External relations votes are structured along a left-right dimension. Voting likeness therefore is, ceteris paribus, higher between party groups that occupy neighbouring positions on the left-right-scale than between other party groups.
However, our findings may be influenced by lumping votes on a broad range of issues into a single "external relations" category and thereby glossing over important differences across issues. For example, the "rally around the flag-effect" that partly explains the hurrah!-votes in the US Congress may be absent in votes on development aid or the condemnation of human rights violations. In order to further test the robustness of the hypotheses outlined earlier, we therefore disaggregate external relations votes.
Building on securitization theory of the Copenhagen school (Waever 1995;Buzan, Waever, and de Wilde 1998), we assume that party politics will be the less pronounced the closer an issue sits at the hard core of security and defense. When it comes to existential threats and the survival of a community, it is more likely that appeals to overcome party-political divisions succeed. By contrast, trade issues have mostly distributive consequences that make it more likely that different political parties will express the interests of winners and losers, respectively. Especially following the Lisbon Treaty, this is also an area where the Parliament enjoys legislative powers through co-decision procedure and the right of veto in international agreements.

Data
The analysis is based on all roll-call votes taken in the EP between the first (1979) and eighth (2014) direct EP elections (N = 28,349), as provided by Simon Hix, Abdul Noury, and Gerard Roland (Hix, Noury, and Roland 2007). The number of roll-call votes per electoral term has increased consistently, with 886 votes in EP 1  to 6961 in EP 7 (2009)(2010)(2011)(2012)(2013)(2014). 21,031 votes were coded by issue area and 2750 (13 percent) of these dealt with external relations, with a significant increase over the decades. This growing number of external relations votes is obviously in part explained by legislative empowerment, with the Parliament simply having to process trade laws and international agreements. Yet it also probably reflects both the gradual development of CFSP/CSDP and the politicization of external relations.
Unfortunately, the coding scheme differs between the legislative terms. From EP 1 to EP 3 the category "external/trade" lumps together votes on foreign policy, trade and development while for EP 4 the category was titled "external trade/aid." For EP 5, we have a more fine-grained measure that distinguishes between votes on "accession," "defence," "development," "enlargement," "external/trade," "foreign affairs," and "trade." For EP 6, votes were identified by both policy area and the responsible committee. The relevant committees are the Committee on Foreign Affairs (AFET) and its two subcommittees (on Security and Defence and on Human Rights), Committee on International Trade (INTA), and Committee on Development (DEVE). For EP 7, no information on committees is available. Instead, policy areas are identified that include "development," "foreign and security policy," and "international trade"; the same categories are also used for EP 6.
For EP 6 and EP 7, the two legislative terms with by far the highest number of external relations votes, we disaggregated votes thematically. First, we divided the votes into three broad categories-international trade, development, and foreign and security policy. This classification follows committee jurisdictions in EP 6 and policy area coding in EP 7. In order to arrive at a more nuanced picture of voting behavior, we also categorized the votes in EP 6 and EP 7 into more specific issue areas and geographical regions. In terms of issue areas, we distinguished between democracy, rule of law and human rights (347 votes); enlargement and European Neighborhood Policy (266); CFSP/CSDP (254); arms and arms control (77); conflict (271); and agreements/treaties (53). With a view to regions, we distinguished between Europe (269); Middle East and North Africa (MENA) (147); Russia, including the Northern Dimension (85); USA and Canada (60); China (40); Asia (168); Sub-Saharan Africa (83); Latin America (64); Turkey (109); and Global (220). 4 Furthermore, we carried out in May-September 2017 five interviews with civil servants that have considerable experience of working for the relevant EP committees (AFET, INTA, DEVE) and party groups. Many of the interviewees had also worked for other EP committees, and were thus able to compare different sectoral committees. The purpose of these interviews was to provide first-hand information about decision-making culture in the committees and the party groups, but given their limited number, information derived from the interviews should be seen as supplementary to our main voting data set.
To examine the cohesion of a group, either the parliament as a whole, a party group or a national delegation, we use the agreement index (AI) as developed by Hix, Noury, and Roland (2007, 91-95). It draws on the often-used index of Rice (1928) but differs from it by taking into account abstentions as MEPs have three voting options (Yes, No, Abstain). 5 The AI ranges from 0 to 1. When all party group members vote the same way, the score is 1. If the party group is completely divided, with a third of the members voting "yes," a third voting "no" and a third abstaining, the score is 0.
For each of the seven EPs, we calculated the average AI for external relations votes and for votes on other issues for every political group, every national delegation and for the parliament as a whole. We ran t-tests to determine whether the differences are statistically significant. We report all calculations and the results from the t-tests in the online appendix. Figure 1 visualizes the cohesion of the whole parliament when voting on external relations and on other issues. While the box plots are tall and the whiskers stretch across a broad range of values, the medians for external relations and other votes are close to each other. The t-tests confirm that there is only little support for our first hypothesis according to which cohesion in the EP as a whole would be higher in external relations votes. As reported in detail in the online appendix, the average AI for external relations votes is not consistently higher than the average AI for votes on other issues (in EP 3, it is lower). Where it is higher, the difference is not statistically significant in half of the cases, and where it is statistically significant, the difference is small (the average for all seven EPs is 0.021). With a view to overall cohesion, the EP clearly differs from the US Congress (Kupchan and Trubowitz 2007) and many national parliaments (Wagner et al. 2018). Hypothesis H2a that cohesion is decreasing is confirmed: the trendline for external relations votes shows a small decrease in the average AI. 6 t-tests for each national delegation and each political group find weak support for our hypothesis H1b: for almost 80 percent of the national delegations, external relations votes do not have a significantly higher cohesion than other votes. Of the 54 political groups that operated in the seven EPs, only the Rainbow group in the 3rd EP and UEN group in the 5th EP had significantly lower levels of cohesion for external relations votes. For ca. 37 percent of all political groups, the opposite applied, i.e., external relations votes had significantly higher levels of cohesion than other votes. However, the average difference in this group remains small (0.046).

Cohesion of the Parliament as a Whole, Party Groups, and National Delegations
Previous research has consistently shown party groups to be considerably more cohesive than national delegations-that is, all MEPs elected from a particular member state. This finding applies also to external relations votes. Higher AIs are primarily reserved for small countries with a small number of MEPs such as Malta, Luxembourg, Slovenia, and Estonia, but also for Hungary and Romania. The UK, the Czech Republic, and the Netherlands were the least united national delegations in EP 6 and EP 7. Otherwise no clear pattern emerges either over time or between various issues-foreign and security policy, trade, and development. Moreover, the AIs are rather similar with the overall cohesion scores, thus providing further evidence that foreign policy is indeed "business like any other" in the Parliament.
Turning to party groups, the findings are again by and large in line with earlier research. The main party groups, notably EPP, PES, and ALDE, are highly cohesive, with voting cohesion in external relations on average just below or above 90 percent. The same applies to the Greens with the exception of the 2004-2009 term and to the conservative ECR (previously ED). Many smaller and "less institutionalized" groups, particularly the Eurosceptical EFD, display consistently lower levels of cohesion. As with national delegations, no clear picture emerges either over time or between the three main issue areas. The cohesion scores in external relations votes are also roughly comparable with overall AIs. The differences in EP 6 and EP 7 are small but in the same direction, with groups generally more cohesive in external relations votes.
In order to compare the cohesion of party groups and national delegations systematically over time, we also calculated relative cohesion scores that take into account the level of conflict in the whole Parliament (figure 2). The relative AI is computed per legislative term simply by taking the AI of the party group or national delegation and then dividing it with the AI of the whole chamber, with the result then divided further with two so that the relative AI also ranges from 0 to 1 (Hix, Noury, and Roland 2007, 92-95). Figure 2 visualizes that party groups are on average considerably more cohesive than national delegations. Furthermore, the figure shows that for both groups and delegations, differences in cohesion for external relations votes, on the one hand, and for votes on other issues, on the other hand, are small (the political groups in the 5th EP being an exemption). In the majority of cases, cohesion for votes on other issues is slightly higher than for external relations votes. We find only weak support for H2b about party group cohesion in external relations decreasing over time. Although the overall trend is indeed toward less cohesion, the effect is very small (the trendline has a slope of −0.012).
Examining EP 5, EP 6, and EP 7, we find no consistent pattern between trade, development, and other foreign policy votes. In the 1999-2004 electoral term development produced more division in the chamber, whereas in the subsequent electoral terms, trade was most and development was least contested. An exploration of cohesion scores thus offers consistent support about the "normalization" of foreign affairs, with national interests no more disruptive inside the EP party groups or in the whole chamber and with the votes just as divisive as votes in other issue areas.

Coalition Patterns in External Relations
The second part of our empirical section focuses on coalitions. Given the much smaller number of external relations votes until the 2004 elections, we limit our analysis to EP 6 and EP 7. We calculate the voting similarity between all possible pairs of party groups in a straightforward manner (see tables 1 and 2): two party groups were deemed to have voted the same way if the plurality of members in each group voted the same way (Yes, No, or Abstain) (Hix and Høyland 2013, 179). For each pair of party groups, we code whether they occupy ideologically neighboring positions on the left/right scale (1 if yes; 0 otherwise). We ran t-tests to determine whether the differences in average voting likeness between ideological neighbors and other pairs are statistically significant. 7 Theoretically the main difference should exist between distributive votes (trade and development) and other types of external relations questions, with the former set of votes expected to find distances between party groups following their locations on the left-right dimension.
In both EP 6 and EP 7 and across all external relations votes as well as development, foreign and security or international trade votes, ideologically neighboring party groups have a significantly higher voting likeness than non-neighboring groups. Turning first to EP 6, the highest voting likeness indices are between EPP and UEN (82 percent), ALDE and EPP (79 percent), ALDE and PES (74 percent), and ALDE and UEN (73 percent): The grand coalition between EPP and PES was formed in 69 percent of external relations votes. Largely the same situation prevailed in the 2009-2014 Parliament. Here, the highest voting likeness was between ALDE and EPP (83 percent) and between ALDE and PES (82 percent). External relations votes thus produce the same pattern as in all votes, with a broad coalition of the three centrist party groups (EPP, PES, ALDE) in the clear majority of the votes (Bowler and McElroy 2015).
There is hardly any systematic issue variation in either EP 6 or EP 7 and, overall and allowing for some exceptions, coalition patterns reflect parties' placement on 7 We follow standard practice and treat the following party groups as ideological neighbours: European United  the left-right dimension. As with cohesion values, we find no real systematic differences in coalition dynamics between external relations votes and overall votes in the chamber. The analysis thus produces mixed findings: on the one hand, our H3a is clearly supported by the data as voting likeness reflects party positions on the left-right dimension, but the same pattern applies to all votes as well. Regarding trade, our interviewees also underlined the importance of the north-south divide, with MEPs from northern member states more in favor of free trade and southern representatives more supportive of protectionist measures.

Variation Across Issues and Regions
As regards variation across issues, party groups exhibit little variation in general. However, interestingly, enlargement votes produced divisions inside EUL/NGL, ALDE, EPP, IND/DEM-EFD, and UEN. In votes on CFSP/CSDP, the three centrist groups (EPP, PES, ALDE) vote often together, although again the differences between issue areas are small. The outlier category is nonetheless clearly "arms," where at the level of the whole chamber the AI was 0.57. We had a closer look at this category of votes, and arms exports proved a particularly contested topic among MEPs, although group cohesion in arms exports votes was not lower than in other types of votes on arms. As for national delegations, their cohesion was often lowest in votes on arms. The most common coalitions in votes on arms were formed between EPP and UEN and PES and ALDE while this category saw less cooperation between EPP and PES than any other issue. This confirms that the left-right dimension structures votes on arms. However, overall we find very little support for H3b as cohesion in the whole chamber was not higher in votes on security than in other external relations questions. Figure 3 shows variation across foreign policy issues and regions for the entire parliament.
The whole Parliament was clearly least cohesive in votes on Turkey and USA/Canada, while votes on Sub-Saharan Africa and Asia saw least contestation in the chamber. As tables I and II in the online appendix show, EPP and ALDE were particularly divided over Turkey. Previous research, drawing on a much more limited number of votes, has also shown that Turkey divides opinions among MEPs, primarily along the left-right dimension (Braghiroli 2013). The two large groups, PES and EPP, disagreed most in votes on USA/Canada. Otherwise the differences between regions are relatively small, both in terms of the cohesion of groups and national delegations and coalition patterns.
To summarize, the closer inspection of votes in EP 6 and EP 7 did reveal some interesting variation in terms of both cohesion scores and coalition dynamics between issue areas and geographical regions. Yet, the big picture is one of consistency across issues and regions. The differences are often in the details and the contestation between or inside party groups may not always be about the actual issue or country. This came up also in essentially all our interviews, with the interviewees stating that excluding certain topical and highly contested issues, it can be very difficult to predict which topics prove controversial. For example, seemingly unimportant votes on trade with a distant developing country might trigger controversy because of the legal precedent set by a particular clause in the treaty. The same applies to votes on the human rights situation, with MEPs perhaps disagreeing about whether to support a stronger role for the EU/EP in safeguarding LGBT rights or in advancing corporate social responsibility. National delegations can also have different priorities explained by colonial history (e.g., Portuguese and Spanish MEPs interested in Latin America), economic factors (e.g., trade with Pakistan being important for member states with textile industries), or geographical proximity (e.g., the salience of Russia for several member states sharing a border with it). Furthermore, individual MEPs can be highly sensitive toward particular topics, such as Western Sahara, and this may explain the length of debates and the high number of votes.

Concluding Discussion
This article has analyzed external relations votes in the EP. Whether we examine party group cohesion or coalitions, the voting patterns hardly differ from overall votes. National interests appear no more prominent when external relations are on the plenary agenda, with party groups reaching similar levels of cohesion as in internal market votes. Nor did we find any real change across time or between various external relations questions. Our results show that foreign policy does not constitute a "special case" in the EP. The EP thus differs from the US Congress and other parliaments where votes on external relations are often less politicized than domestic politics. The EP is used to building large majorities behind its resolutions-vis-àvis the Commission, the Council, and the outside world-and also external relations thus find broad agreement between the main party groups, with opposition coming often from the Eurosceptical right and the radical left groups. Hence, our findings are in line with the article by Cicchi, Garzia, and Trechsel (2020) in this Special Issue, as they show that if a political party supports the EU, it will also support the further development of EU's foreign and security policy, and vice versa.
Our decision to focus on the "big picture" was intentional. For example, as TTIP and the EU-Canada Comprehensive Economic and Trade Agreement show, there is ample proof that international trade has become a much more contested issue area in recent years (Laursen and Roederer-Rynning 2017). Yet our analysis shows that international trade as a whole is an issue area like any other. The same obviously holds for votes on Russia or the development of CFSP/CSDP, topics that may produce heated debates but overall are no more divisive in the Parliament than other topics. While we thus strongly encourage further research on individual foreign policy questions in the EP or other legislatures, our findings should also be taken as a warning against drawing too strong conclusions from case studies of individual policy processes. Such case studies are always tied to specific issues or time points, and while often insightful and informative, they necessarily lack both a longitudinal and comparative perspective.
An obvious question is why national interests do not surface? Returning to the puzzle outlined at the start of this article, we believe our results are explained by three factors. First, the EP remains to some extent sidelined in CFSP/CSDP, and hence its votes are less consequential than in some national legislatures, most notably the US Congress. This applies especially to the use of force and the EU's main security policy strategies that are primarily decided between member states in the Council and the European Council. Second, external relations cover a lot of ground, from condemnations of human rights violations in distant countries to important trade deals and military conflicts in neighboring areas. As a result, while individual questions can produce controversy, most external relations votes are probably not that salient for most MEPs, a situation that is found also in other policy sectors like agriculture, environment, or even internal market legislation. As our exploration of more specific issue areas and geographical regions suggested, there is a need to dig deeper into voting patterns in questions such as arms trade, Turkey and the enlargement process, or relations with USA. Only through such indepth investigations, combining voting behavior with plenary speeches and perhaps interviews, can we truly understand the positions of MEPs in various foreign affairs questions (see Braghiroli 2013Braghiroli , 2015. After all, our study was limited to studying voting cohesion and coalitions, excluding thus the content of the issues voted upon in the EP plenary. Third, and most importantly, the normalcy of external relations votes suggest that EP party groups manage to build unitary positions across all issue areas. While exact causal mechanisms are difficult to establish, our results thus emphasize the explanatory value of established institutional codes of conduct. EP committees and especially the more stable party groups have over the decades developed their own decision-making mechanisms and behavioral norms, and research has shown how new MEPs are socialized into such existing parliamentary practices. This point came strongly across in our interviews, with the interviewees underlining the similarities between committees and party groups. To be sure, there are differences between committees, depending for example on committee chairs, but clearly the committees responsible for external relations questions, including AFET, follow largely similar routines as other sectoral committees. For example, if MEPs or national parties intend to vote against the group position, they should inform their group beforehand-exactly as happens in other policy areas as well. The transparency of EP committees may account at least for some of the observed similarities, as foreign and security policy questions are not debated behind closed doors as often is the case in domestic legislatures. Nonetheless, "committee cultures" in the EP remain under-researched (Settembri and Neuhold 2009), and future research should thus explore further decision-making in external relations, also inside party groups. Finally, our results about the "normality" of external relations will hopefully lead scholars to pay more attention to party politics, as until now research on the foreign policy activities of the EP have treated the institution as a unitary actor, thus neglecting the role of competing ideologies behind the positions of the Parliament.