Losers together? Grand coalitions in the EU member states

Abstract Over the last two decades, the formation of grand coalitions has grown in the European Union (EU), even in countries with no previous political experience with them. Alongside a significant rise in both new and radical parties, grand coalitions signal the increasing fragmentation of contemporary European politics. We, therefore, investigate the electoral performance of both mainstream and new parties entering and leaving grand coalitions. We find that mainstream parties do not appear to enter grand coalitions after negative election results. They are, however, punished in the following elections, albeit not as heavily as previous findings have shown. This post-grand coalition electoral penalty is true for both major and minor grand coalition members. These findings contribute to the literature on party competition and provide insights into the choices mainstream parties' have been making in response to recent and rapid changes in the electoral landscape of the EU.


Introduction
In recent decades, grand coalitions (GCs) in which mainstream parties ally with previous rivals or even new challengers, are becoming more common in European politics. When election results produce no clear winners or no coherent majorities can be formed, GCs become an option in order to avoid political paralysis or a second election. Some political scientists consider GCs 'last resorts' in desperate times (Campbell, 2015). Safire's Political Dictionary defines these government's majority arrangements as 'unholy alliances' and the parties involved as 'strange bedfellows ' (2008: 132).
Such a reputation suggests that, up until recently, GCs were considered exceptional, second-order choices to achieve working arrangements for 'special times' such as when a country was at war or suffering a deep economic recession. The exceptional nature of GC's highlights their potentially greatest challenge to democratic norms, namely, how GCs confound the basic logic of voting. If rival parties from opposing sides of the ideological spectrum join forces in a GC, what was the point of voting in the first place? However, from the rational choice approach embedded in the government formation literature, GCs can also be seen as normal as any other combination as they are minimal winning coalitions, fulfil the bargaining proposition, andall in allare simply the best option for office-seeking political actors (Martin and Stevenson, 2010).
The number of GCs in the European Union (EU) has been increasing, with no difference whether countries are using proportional representation or majority election systems. Since 2004, there have been 17 GCs in 11 European countries. While some countries have become quite familiar with the arrangement: Austria (three times), Belgium (two times), Germany (three times), others have had only brief experiences. In the literature, the participation of mainstream parties in GCs is seen as the beginning of a spiral toward electoral irrelevance (Strøm, 1990;Hix, 2009;Jacoby, 2017), particularly for socialists and left-wing movements (Garton Ash, 2018). As one recent example, the German SPD entered as a minor partner in a GC with Angela Merkel's CDU twice (in 2005 and 2013) and its share of the vote subsequently declined. After the 2017 federal election, the SPD joined its third grosse Koalition as the only viable alternative that did not involve the populist right-wing AFD. At the same time, GCs that governed Greece from 2012 to 2015 led to the collapse of the centre-left party PASOK. Here, we examine the empirical basis for expectations about the choice to enter GCs and the electoral costs for doing so. We focus on GCs formed after general elections in the EU member states since 2004 (or since the country's first accession to the EU), rather than those arising from government crises or during parliamentary terms. 2004 provides an accommodating start date as it coincides with the 'big bang enlargement' in the EU. To examine the effect of the choice to enter a GC, we consider the change in electoral performance for parties entering GCs and how these parties performed in the following elections. Building on previous work, we also include new parties in the analysis to identify any correlation between their performance and the formation of GCs. The empirical findings here demonstrate that while mainstream parties do not necessarily enter GCs because of disappointing election results, GC member parties both major and minor partnersare punished in the following elections, although not heavily. This research fills two gaps in the literature dominated by the German experience. One, we expand the comparative study of GCs in Europe and identifying the electoral gains or losses of political parties entering and exiting GCs. In doing so, we unpack a crucial and understudied aspect of the functioning of parliamentary democracies. Two, the findings here are salient to broader and timely discussions of growing party system fragmentation and polarization in Europe as well as the mounting set of challenges to long-term legitimacy of the EU. The article is divided into five parts. Following this introduction section, we unpack the review of the current literature on GCs. The third section summarizes the current rise of GCs in the EU member states. In the fourth section, we examine the electoral performance of mainstream parties previous to and following their participation in GCs, in addition offering an overview of major and minor partners in GC and the role of emerging new parties. Finally, we draw the findings together and identify areas in which this study has contributed and directions of future research.
Coalitions, electoral volatility, and new parties in the EU member states A GC is a strategic alliance in which the two largest parties, usually from the right and the left, unite in a coalition government (Strøm, 1990). In most cases, they are a coalition arranged between the largest centre-left parties (e.g. Socialists) and the largest centre-right parties (e.g. Christian Democrats or Liberals, in some cases) in exceptional times. The study of coalitions has provided a great deal about how coalitions form and under which conditions and what causes them to fall apart (Laver and Budge, 1992;de Swann, 1973;Ansolabehere et al., 2005). Recent cross-national comparative research has investigated the determinants of coalitions (Albala, 2017;Mölder, 2017) as well as at the local level (Debus and Gross, 2016;Gross, 2018). GCs can be considered a unique subset of coalition governments in developed democracies and have been a growing object of scholarly research. Yet, the number of cases has remained relatively low for several years and most research is built on single-case studies (Helms, 2006;Proksch and Slapin, 2006).
In the past, GCs have been considered as a regular feature of 'consociational democracy' (Lijphart, 1969); and in continental Europe, a consequence of the 'fragmentation of political culture' (Lijphart, 1969: 207). Dahrendorf argued that GCs were a 'cartel of elites ' (1967: 11), somehow anticipating the famous 'cartel party' literature in which Western Europe parties have adapted themselves to declining levels of participation, using the resources of the state in a collusive manner in order to maintain its position within the political system (Katz and Mair, 1995).
While it is important to recall that government formation in many countries does not refer to pre-election arrangements and that many parties insist on fighting only for themselves without making any coalition pledges; in general, normal government coalitions should resemble preelection arrangements or, at a minimum, cohere ideologically. In this sense, GCs can be considered as a sign of resistance or frustration by mainstream parties. That is an inevitable arrangement which is made to provide the country with a government, due to the increasing fragmentation and the difficulty to integrate new parties and populist parties in the traditional dynamics of government formation. 'While grand coalitions can be formed for many reasons, the rise of populist parties are challenging conventional party systems, exhausting, and discrediting many classic coalition formulas, and shrinking the room for manoeuvre of large parties (along with their vote share)' (Jacoby, 2017: 331).
Despite the infrequent formation of GCs, there can be several reasons for normally opposing parties to form a government together. One is an extraordinary national political or economic crisis or a war. In this case, the need for national unity goes beyond standard ideological contrasts, as happened in the United Kingdom during the two World Wars. More recently, another possibility has become the most popular: that no other coalition is stable or is considered numerically feasible (Strøm and Nyblade, 2009). In this way, the chosen arrangement for mainstream parties of both the left and the rightis often to keep new parties or populist parties out of government (Mral et al., 2013). This might be considered a cordon sanitaire, a term that was introduced in the late 1980s in Belgium. Other well-known examples existed in Italy towards the neo-fascist Movimento Sociale Italiano (MSI) and in France against the Front National (FN). Here the criticism might be that most of the literature conceives a cordon sanitaire as the systematic exclusion of a party from any kind of cooperation on all levels (local, regional, and national). Hence, something that has to be measured over long periods of time. This is different from not being willing to form a government coalition with a certain party at one specific moment in time. However, the logic behind it is the same: the necessity to exclude a challenger that is considered unworthy to be part of any possible coalition (Stefuriuc, 2013).
Jacoby described three logics that stand behind large parties' choices for GCs (2017: 331): 'renovation', which sees large parties entering GCs together primarily to pursue policy goals unreachable in smaller coalitions, especially those goals requiring constitutional supermajorities. The second logic is 'clientelism' and sees GCs as a chance for large parties to channel benefits to their constituencies and to enjoy benefits themselves (similar to cartel party politics). The third logic is 'sterilization', and it sees GCs as a strategy for large parties to limit the room for manoeuvre of (often new) parties unacceptable to one or both (i.e. a cordon sanitaire).
In recent years there has also been an increase in coalitions which include minor parties that are traditionally distant from the mainstream one. An example comes from the 2010 LibDem-Conservative coalition in the United Kingdom. Following the general election resulting in a hung parliament with the Conservatives as the largest party, the Liberal Democrats and Conservatives agreed to enter into a formal coalition. In the UK, the ideal GC would obviously be a Conservative-Labour onethe two long-term mainstream rival parties. That is why we do not consider here these types of coalitions, which are also quite frequent in Central and Eastern Europe (CEECs), especially in Latvia, the Czech Republic and Slovakia (Cabada et al., 2014: 138-139;Foy, 2016). In fact, for many young CEECs democracies, where it is difficult to highlight 'long-term mainstream rival parties' (Kreuzer and Pettai, 2003) and personalistic parties are the norm, these are very frequent arrangements. They are also another sign of the growing fragmentation of politics, in which mainstream parties are losing votes and are incapable or unable to form congruent coalitions (Gherghina, 2015).
In the literature on government formation, the starting point is usually identified in Downs's (1957) original work on electoral competition, in which parties are 'teams of men' seeking to maximize their electoral scores to achieve government further originating the theories of competitive party behaviour which distinguish between vote-seeking, office-seeking, and policy-seeking models of party behaviour. Strøm (1990: 570) criticized these models and, drawing on New Institutionalism, argued that a 'more general behavioral theory of competitive political parties requires an understanding of the interrelations and trade-offs between different objectives […] pure vote seekers, office seekers, or policy seekers are unlikely to exist'. In dealing with GCs as specific kinds of government formation, what do we expect about the behaviour of parties? Why should they enter GCs that sometimes prove to be electorally costly?
As one response, Laver and Budge (1992: 7) make the argument that 'each election and each process of government formation has to be analyzed separately, as if parties had no history and no future'. Wagner andMeyer (2014: 1023), having in mind pre-electoral environments, argued that 'parties are primarily driven by a vote-seeking, and ultimately an office-seeking, incentive. Parties will pursue catch-all strategies and aim to represent as large a portion of the population as possible to increase their bargaining power and chances to enter government'. However, the literature suggests that these pre-electoral incentives may become a set of new incentives in post-electoral bargaining. In following Ecker and Meyer (2020), the basic assumption is that parties are genuinely policy-seeking actors that bargain over government formation, in order to get some of their proposed policies enacted. However, when a post-electoral coalition game is needed, the officeseeking approach also becomes crucial. Because other motives here apply: the wish to keep other parties out of government, the desire to stay in power, and not to go to new elections (Anria and Cyr, 2017). Or as Bäck et al. (2011) pragmatically illustrated: 'ministerial portfolios are the most obvious payoffs for parties entering a governing coalition in parliamentary democracies'.
Earlier studies have highlighted several alternative potential causes for GC formation: previous experience with GC, leadership change/continuity, pre-electoral alliances, or political scandals (Helms, 2006). Here, we do not theorize about why parties choose to enter into GCsother than to provide a comparative assessment of whether negative electoral performance is one culpritbut rather focus on their trajectories and electoral performances that follow. That is, we focus on the electoral balance of parties which have been GCs partners in the elections which brought to the formation of GCs and the ones after and the results of new parties in the same elections.
These recent trends in aggregate electoral volatility, the evidence of the mobilization and success of new parties and the decline of mainstream parties push back against Lipset and Rokkan's theories of the 'freezing of party systems' (Mair, 1998). Mair's work on the competition structure of party systems (open vs. closed) is a key framework in assessing the success of new partiesboth electorally and their chances to join coalitions (Mair, 1998: 199-200): '…the structure of competition in any given party system (…) is centred on the assumption that there exists a stable structure of competition. Structures of competition can be seen to be either closed (and predictable) or open (and unpredictable), depending on the patterns of alternation in government, the degree of innovation or persistence in processes of government formation, and the range of parties gaining access to government (…) there is the possibility of situations in which electoral change is the consequence rather than the cause of party system change…' Given this continuation of trends in declining turnout and increasing electoral volatility, with the enduring period of political turbulence, voting behaviour has become increasingly unstable (Drummond, 2006). And, in the last decade (2008-18), major democracies in Europe have witnessed radical shifts in electoral results, often resulting in a defeat of the ruling coalition (Bellucci et al., 2012).
The rise of GCs is concurrent with many European countries experiencing a prolonged period of electoral turmoil. Voters have punished incumbents and the major reason has been considered a negative perception of the state of the economy (Kriesi, 2014). Especially since the Great Recession of 2007-08, many mainstream parties have been severely punished by voters where the economic conditions were getting worse or not improving quickly enough. According to Kriesi's data, in Western Europe, political parties which were part of governments' majorities lost an average of 3.6% in the following elections, whilst the ones that also hold the Prime Minister lost 5.1% on average (2014, 34). Theoretically, this discussion leads us to the first hypothesis that we will test below: Hypothesis 1: Mainstream parties entered GCs after negative election results.
In recent years, political science research has increasingly directed its attention towards coalitions as an integral part of the decision-making calculus of voters and particularly the retrospective voting approach of examining the government performance (Fisher and Hobolt, 2010;Plescia and Kritzinger, 2017). Meyer and Strobl (2016) studied how voters perceive policy positions of coalition governments and found out that voters consider not only the programmatic offer of parties but also coalition formation processes and coalition bargaining. Angelova et al. (2016), in their study on German coalitions, suggested that, while both coalition partners are held responsible, the larger coalition party with the prime minister carries the largest responsibility burden, but possibly receives also the largest rewards for positive performance evaluations.
Bergman and Ersson (2009) focused on cabinet parties' fate in the elections following government participation. They collected 546 observations and they measured an averaged negative performance (mean: −4.60%, std. dev.: 9.189). Narud and Valen (2008) also analyzed electoral cabinet performance based on their 1945-99 data and they found similar results: the average electoral performance was negative and worse in the 1990s than in previous decades. However, also different outcomes have been reported: Plescia (2017), with an Italian case-study, found no consistent empirical evidence that coalition parties collectively suffer from perceived negative performance. Most of these studies focused on coalition payoffs. Yet, what about the electoral consequences of being part of GCs? Existing research does not say much about it, given the limited number of cases and their recent soaring. Does it electorally cost more being in GC than being in a normal coalition government? In our empirical analysis, in the next section, we will draw a comparison for parties entering GCs and parties not entering GCs.
Aichholzer and Willmann (2014), focusing on Austrian GCs, proved that the two largest parties in GCs (SPOE and OEVP) have gradually lost a significant share of their voter basis over recent decades. Scarrow (2012) showed that, in the elections immediately following the 2005-09 German GC, turnout decreased, volatility rose, party fragmentation increased, and protest voting became more vibrant. On the same line, Banaszak and Doerschler (2012) reported that in elections after GCs, voters tend to move away from the coalitional parties towards opposition parties, and this movement is more likely among the more radical voters of the parties. The electoral price that parties pay for joining a GC government is generally rather high. Thus, we expect to find that: Hypothesis 2a: Mainstream parties that entered GCs lose votes in the following election.
Research on the 2013 Italian general elections showed that the electoral volatility was unprecedented, making that election the most volatile Italian post-Second World War election. The two main parties that in a GC supported the inter-elections Monti technocratic government (PD as a major partner and PDL as a minor partner) lost together almost 10 million votes from the previous election (3.5 million and 6.5 million, respectively). Moreover, 'approximately one out of four valid votes went to a new party that claimed to be outside the traditional left-right ideological divisions, namely the Movimento 5 Stelle (M5S)' (Vegetti et al., 2014). Banaszak and Doerschler (2012), in studying German GCs and contradicting their original hypotheses, found that minor parties are more likely to lose core constituents in a GC because their voters will likely be less willing to make compromises. Because coalition partners are more ideologically distant in a GC, the resulting policy outcomes represent more policy compromise than in 'standard' coalitions. Having in mind the German SPD, the two authors also argued that where viable more ideologically similar parties exist, core constituents of GCs minor partners participating in a GC may shift their support as a result of coalition compromises. Building upon these authors' findings: Hypothesis 2b: The GCs' minor partners suffer the most in the following elections.
Obviously, the electoral losses of GC parties are not only causally connected to being in government but also because of the general de-alignment of voters and the increased volatility in voting (Klüver and Bäck, 2019). We are aware that government parties lose votes most of the times. However, research has demonstrated that is particularly the case in coalition governments and for the junior coalition partners (Spoon and Klüver, 2019).
The next section offers a descriptive look of the GCs formed in the EU in the period 2004-18, followed by an empirical analysis of the 17 GCs over the same period in the EU.
Analyzing the GCs in the EU member states  Since the year 2004, and, for many CEECs considering their accession to the EU, there have been 17 GCs in the EU member states. Yet, while GCs have been extensively studied at the EU Parliament level (Hix, 2009) and as single-case studies, with a particular emphasis on the German case (von Wahl, 2008), there is less comparative work. Here, we aim to fill this gap.
GC's 'traditional' definition features two components: ideology and size. Concerning the size criterion: a GC needs to muster a parliamentary majority (minority governments not included). What about the ideological orientation of its members? The traditional definition would require that parties should be 'usually from the centre-right and the centre-left'. This is a vague criterion, and, moreover, it relates much more to (formerly) dominant mainstream parties on the left and the rightmostly Christian Democratic and Socialist/Social-Democratic parties. Nowadays, in many countries, the two largest parties are very different from those that dominated party systems up until recently. The current polarization has brought mainstream parties to be perceived as ideologically-closer as ever. And GCs like several ones in Austria or Germany were minimum or minimal winning coalitions, so the formation of a GC can be explained by standard coalition formation theories. Recent studies proved that especially the German GCs are ideologically closer than any other potential coalitions that had a majority in parliament (cf. Bräuninger et al., 2019). Moreover, as another example: would a coalition between the two major Irish parties FF and FG, which have similar positions on a left-right scale, not be a GC? Certainly not.
This motivated us to update the original label of GC and focus on coalitions which include the two largest mainstream parties in a party system (which were not in a pre-electoral agreement), rather than trying to forcing a theoretical concept from the German-speaking literature clearly tied to the specific cases in Austria and Germany to other institutional settings. What are 'preelectoral coalitions'? They are quite unusual in Western Europe if considered as pre-electoral pacts (Golder, 2006) but very likely in some settings when considered as so-called Koalitionsaussagen (Debus and Müller, 2013). Maintaining this reference, however, it allows us to exclude government coalitions formed in countries with proportional electoral systems in which parties formally run on their own (because of the electoral system) but then have already pre-arranged potential government coalitions with other individual parties (as happened in Italy in several times with pure proportional systems).
We do not consider 'simple' incoherent coalitions, such as was the case of the Lib-Con coalition in the UK in 2010 and the 2013 'unholy' alliance in the Czech Republic which included CSSD, ANO and KDU-CSL. Two large and 'unlikely' coalitions that, however, cannot be called GCs, because they did not include the Labour Party and OSD (respectively), that is, the two largest opponents of the major partners of the GC. We also do not include 'large' bipartisan coalitions which might have supported 'technocratic governments' and instead analyze GCs which are formed in post-election settings only and not in inter-election ones (although the latter are somewhat rarer). This approach, however, does not prevent us from including coalitions that also have more than one minor party involved, as was the case of Finland (six parties involved in 2011, with both largest parties included in the GC) and Belgium in 2010 (six parties forming the GC, but with the largest N-VA excluded from the coalitionin the ideal case of cordone sanitaire).
We define mainstream parties as parties with a significant political history and government experience, which have an aggregative nature and a large parliamentary seat share. They are clearly defined in the literature as parties that already participated in government at the national level (Sartori, 2005;Shamir and Rahat, 2017: 22). However, according to Ignazi (2021: 1) 'the expression mainstream parties has never been precisely defined in the literature. It usually refers to established and "relevant" (in the Sartorian sense) parties, irrespective of their political alignment and location on the political spectrum'. De Vries and Hobolt (2020: 250) considered as mainstream 'the parties that occupy an overall advantageous position in the system'. Meguid argued (2005) that the best way to understand the meaning of mainstream is in juxtaposition with its opposite: minor, extremist, anti-system, and/or 'niche' parties.
Interestingly, also part of the existing literature on challenger parties has largely implied its definitions focusing on the programmatic reaction to mainstream parties (de Lange, 2012). Moreover, they are often referred to as challenger/anti-establishment/outsider parties almost as synonyms. Abedi (2004) defined an anti-establishment party as an actor that simultaneously: 'challenges the status quo in terms of major policy issues and political system issues […] perceives itself as a challenger to the parties that make up the political establishment […] asserts that there exists a fundamental divide between the political establishment and the people' (Abedi, 2004: 11). Akkerman (2016: 277), criticized this approach arguing that when these parties leave behind 'their lone opposition and increasingly cooperate with other parties', they usually do so while maintaining their radical positions and without 'moderat[ing] their anti-establishment ideology'. Wolinetz (2018: 285-286) reasoned around the label of 'outsider' parties and he identified three areas within a party system: 'the core'' consisting of insider parties that govern or oppose and rotate in and out of office; an 'intermediate zone' consisting of parties that could govern but don't do so often; and 'outsiders'actors that 'represent and never govern'. He specified that both actors that 'never govern' as well as those that 'rarely' govern qualify equally as outsiders. Finally, Hobolt and Tilley (2016: 974) argued that challengers simply are parties that have not previously held political office. This simple and straightforward definition has recently been adopted by the literature (Schulte-Cloos, 2018; Van Spanje and de Graaf, 2018; Bakker et al., 2020). In a recent effort, De Vries and Hobolt (2020) illustrated the nature and the behaviour of challenger parties, observing how, like disruptive entrepreneurs, these parties offer new policies and defy the dominance of mainstream ones. According to them, because they are not burdened by government experience, challenger parties adapt more quickly to shifting voter tastes and harness voter disenchantment. Consistently with the literature, in this article, we consider challenger parties the ones which have never participated are in national government (Hobolt and Tilley, 2016). Table 1 summarizes the occurrence of GCs in Europe per year of election. There are countries that have a significant tradition of consensual-style politics, such as Austria, Luxembourg, and Netherlands, while other European nations have become very familiar with GCs only recently (Germany, above all). Since 2004, 11 countries experienced a GC, for a total of 17 GCs which have been formed in the same period. In the previous 22 years , in the same countries, only five GCs were formed. In this analysis, we could have selected all European countries since 1990 or as soon as they became democracies, however, we have decided to select these cases as, starting with the year 2004 (the largest EU enlargement so far), they provide a comprehensive representation of contemporary European politics.
It appears that GCs have become a less exceptional choice for mainstream parties to hold onto power and to exclude populist movements and new parties from going to the government. Not being in the position to govern on their own or with the traditional allies for mere arithmetical questions, the major partners of future GCs prefer to enter a coalition with traditional mainstream long-term rival parties, instead of including 'awkward' partners in the government coalition. Long-term rivals may be ideologically distant, but they are preferred as coalition partners to political organizations which have no government experience and are often characterized by anti-establishment sentiments and harsh rhetoric against mainstream parties and their constituents. In some cases, however, apart from the mainstream parties' natural mistrust towards these 'new challengers', it is the challengers themselves who are against any form of post-election alliance or government negotiation with other parties. Together, this provides an alternative explanation for the mainstream parties' intentions for entering a GC; namely, their desire to exclude unwanted extreme parties rather than flagging electoral performance.
This was the case, for instance, in the 2005 German Federal elections when both Angela Merkel (CDU) and Gerhard Schröder (SPD) claimed victory, although it soon became clear that neither could form a majority government within the existing coalitions. The first grosse Koalition since 1969 became a viable option when the left-wing coalition of the SPD, Greens, and Left Party was not possible because of the strong hostility between Schröder and Oskar Lafontaine, a former SPD chairman turned leader of the Left Party. Lafontaine also rejected any possibility of participating in a coalition with either of the two main parties (Clemens and Saalfeld, 2013). A similar case is Italy in 2013: the centre-left coalition gained an absolute majority of seats in the Lower House but was in need of partners in the Senate. After a brief negotiation, it became clear that the new-born M5S was unwilling to enter any coalition with the 'traditional' centre-left, so the centre-left had no other choice than to reach an agreement with the traditional rivals of the centre-right, the once enemy Silvio Berlusconi's centre-right. Also, the 2017-18 German post-election negotiation talks were mainly conducted bearing in mind the necessity to exclude the radical right-wing AFD from any negotiation talks and when the 'Jamaica' coalition failed to prove successful, the by-now classic grosse Koalition again became feasible. The 2011 Finnish government formation is another example of a GC which was formed out of arithmetic reason and with the intention of excluding some 'challengers' (in this case True Finns, which gained 19.1%, +15 points compared to the previous vote). After the election results, Jyrki Katainen, leader of NCP, the largest party in Parliament said that the result of the election supported a coalition consisting of the three largest parties (National Coalition, Social Democratic Party, and the populist True Finns). However, a few weeks later, the True Finns announced that they would withdraw from the government formation negotiations due to disagreement around the bailout issue. Party leader Timo Soini said he did not want to compromise the party's core principles. After Soini's statement, Jyrki Katainen invited another five parties (including the SDP and the Christian Democrats) to negotiate and a GC led by the NCP was finally formed.

Election results of government parties in GC and the rise of new parties
The aim of this research is also to examine the motivations and effects of parties' participation in GCs in the EU member states in the 14 years since 2004. We take an empirical approach to assess if voters rewarded or punished GCs' members in the next elections. The methodological approach taken in this section is straightforward. We measure the change in parties' electoral performance for the elections which were followed by a GC and the change in parties' electoral performance for the elections which ended a legislature characterized by a GC government.
The total number of cases is 17 GCs. Despite this small number of observations, the results are consistent and clear (see Table 2): based on data drawn from the Organization for Economic Co-operation and Development and NSD data, GCs are born after general elections which saw parties losing an average of 1.25% in vote share. At the same time, major partners in the GCs gained a consistent 1.79% in vote share in these elections, while minor partners suffered an average loss of −3.49%. The figure for mainstream major parties is quite surprising as it goes against long-standing theory about the motivations for mainstream parties to consider and eventually join a GC (Laver and Budge 1992;Strøm 1990). However, it does correspond to the qualitative cases above that provided evidence that mainstream parties concern about new and extreme challengers may partly explain their decisions to enter into GCs. Again, we refer to major partners as those parties entering the GC with the most votes and hold the Prime Minister. 2 Again, Hypothesis 1 tells us to expect mainstream parties entered GCs after negative election results. Using data from Comparative Political Dataset (CPDS, Armingeon et al., 2018), we can compare the previous election performances for mainstream parties that joined GCs and those that did not join GCs. Previous election performances are different for mainstream parties (−0.55, std. dev. = 1.18; N = 22) and mainstream parties that are GC members (−1.45, std. dev. = 0.99; N = 44), a difference that is not statistically significantly (t-test: 0.555; P < 0.581; df = 64). Simply, mainstream parties which join GCs do not appear to have statistically different performances than other mainstream parties that did not join GCs. Poor election performance does not appear to be a key motivating factor in mainstream parties' decisions to enter into GCs (i.e. no support for H 1 ).
Some cases are very evident and representative: in 2008 in Austria, the OVP entered a GC as a minor partner after losing 8.4 points, in Bulgaria, in 2005, Prince Simeon's NDSV joined a GC as a minor partner after suffering −22.9% in the national elections. The CDU/CSU is leading the 2018 grosse Koalition after scoring a terrible −8.6% in the 2017 German Federal elections. Obviously, also notable exceptions apply: in 2012 the Dutch PvdA joined a GC led by the liberals of the VVD after gaining +5.2% and in June in the same year, the ND formed a GC after winning the snap elections with a tremendous +10.8%. The already-mentioned GC which was formed in Finland in 2011 was literally a sum of losses, with all six coalition members having lost votes compared to the previous election.
For Hypotheses 2a and 2b, we can see in Table 2 how voters react to GCs: on average, GC members lose an average of 3.84% in the following elections, with a consistent decrease for major partners, −5.54%, and a more moderate decrease for minor partners who, on average, lose 2.79% in the elections after being part of a GC. Some findings are more astonishing than others: in 2010 the Dutch CDA lost 12.9 points and in Greece in June 2015 the centre-left PASOK fell by 7.6 points. As usual, the major partners similarly achieve impressive results: in the 2017 Austrian elections, the GC's leading OVP scored an enviable +7.5% and in the 2010 Slovakian Parliamentary elections the leading Direction-SD came out with another success: +5.7%. In general, minor partners seem to be the biggest losers in GC formations, they enter on a loss and are punished by voters in the next elections. We note that although some previous studies have pointed out the importance of differentiating GCs that did not last an entire term in office (Golder, 2006: 195). Considering our research corpus, there are five cases: Austria (2006), Belgium (2007Belgium ( , 2010, Greece, and Italy. However, because we strictly link our analysis to the electoral event, the swift termination of the GC does not play a role here. Again, continuing to use data from the CPDS, we can build toward a more comprehensive test of Hypothesis 2a: mainstream parties which entered GCs lose votes in the next elections. The first is a test of the differences of parties' performances in the election following the formation of GC. In the first one, we test all parties (mainstream and non-mainstream). Those that were members of a GC did statistically worse (on average, 3.66% less votes) than those that were not members (1.22 more votes; a statistically significant difference; t-test: 5.36, P < 0.001, df = 108). Limiting the test to mainstream parties, the difference remains statistically significant. Non-GC members got, on average, 1.70 more votes to GC members 3.67 less votes (t-test: 4.07, P < 0.001, df = 62). Thus, in support of H2a, both non-mainstream and mainstream parties that entered GCs were punished in the next elections more than those that did not enter into GCs.
Another useful comparison is between parties entering into GC's and parties in the same countries over the same period that entered into 'non-grand' or 'normal' coalitions. Using the CPDS, we find that members of a governing coalition go into these types of 'normal' coalition with a mean change in vote share of 1. 66 (std. dev. = 4.18, N = 49) and suffer a −3.46 decline in vote share (std. dev. = 6.79; N = 48) in the following election. 3 The former result is at odds The difference in following election performance between normal coalition members and GC members is not statistically significant (t-score = 0.21; P(diff≠0) = 0.83; df = 93). Again, this is in congruence with Kriesi's data in Western Europe (2014) and others in the literature on the electoral costs of governing (Green and Jennings, 2017).
with what we have seen for mainstream parties going into GCs. However, the latter result suggests that perhaps GCs are not uniquely detrimental to electoral success and that given the relative similarity across grand and 'normal' coalitions' lack of electoral success in the following elections, perhaps we have oversubscribed the electoral cost to GCs that belongs broadly on incumbents in general. We see this potentially provocative finding as a springboard for future research. However, it requires a more thorough analysis that we can provide in the current submission whichagain is focused on the specific characteristics and results of GC membership.
Finally, in this research, we also aim to check the new parties' scores in the elections which led to a GC and in the ones which followed. The idea is to show how the new parties' increases in votes would generally compress the political space and help mainstream parties to 'manoeuvre' and form government majorities (Hypothesis 2b). Interestingly, in the 17 elections which gave rise to government talks which resulted in GCs, new parties are on the rise, but not as significant as expected. It is true that some impressive peaks were reached, such as in the 2013 Italian general elections (M5S and Scelta Civica summed 33.9% of the vote) and in the 2012 Romanian legislative election (14.7% for Dan Diaconescu's People's Party). However, no new parties entered the German Parliament in the elections which were followed by a GC and the same happened for Greece, Luxembourg, and Slovakia. We should bear in mind, however, that with regards GCs, we have adopted a 'narrow' definition for new parties, considering as new only the political organizations that newly contest national elections and achieve national representation (Rose and Mackie, 1988;Kitschelt, 1989;Krouwel and Lucardie, 2008). We prefer this restrictive definition because it gives us the possibility to isolate and identify pure new parties, meaning political organizations with no previous experience in national parliaments and government negotiations. In doing so, for instance, we do not consider True Finns as a new party in the 2011 Finnish parliamentary election or the M5S as a new party in the 2018 Italian general elections, because while they were still 'young and unpredictable' they are not new, having already contested and gained representation at a national level in past competitions.
New parties occupied, on average, 4.4% in the elections which led to the formation of a GC (12 parties in total) and 2.6% in the ones following a legislature characterized by a GC government (7 parties). While the numbers are small, they show how new parties impact the pre-GC elections was almost double that of the elections post-GC. And, for instance, they were clearly a reason for the formation of the GC in Italy in 2013.
Turning again to the CPDS, we can empirically evaluate Hypothesis 2b in which GC minor partners suffer the most electorally. In the case of minor GC parties, comparing all parties, minor GC members had 2.78 less votes (std. dev. = 1.15, N = 26) and minor non-GC members had 0.19 less votes (std. dev. = 0.54, N = 82; t-test; 2.24; P < 0.05; df = 108). For mainstream parties, minor GC members were the same as previous while minor mainstream non-GC members had 1.52 less votes (std. dev. = 5.44, N = 36; t-test; 0.87, P < 0.39; df = 62). Thus, comparing all parties, minor GC members did worse. Although, limiting the sample to mainstream parties, they did not do statistically worse.
We take a final step by using each party's electoral performance in the following election as the dependent variable in a straightforward Ordinary least squares (OLS) regression with standard errors clustered by country. We regress a series of dummy variables the codes each party's GC membership (or not) and whether they were a major, minor, or very minor party. We include all of the national elections of EU member states which produces a GC. With parties as the unit of analysis, we can examine the impact of GC membership and the type of party in the membership on their subsequent electoral performance. Given the current literature, we control for GC party control of the finance minister, GDP rate change, change in unemployment, and change in youth unemployment. We estimated models with various operationalizations of economic performance, unemployment, youth unemployment and all worked well. All alternatives are available from the authors.
In Table 3 we can see two sets of two models. Models 1 and 2 show that being a GC member (in Model 1) is statistically related to lower electoral performance and this is true for both major and minor GC members (Model 2). In the second set of models (Models 3 and 4), we limit the examination to mainstream parties. In this subset, we find substantively the same relationships. These results lend support to both Hypothesis 2amainstream parties that entered GCs lost votes in the next elections and Hypothesis 2b (GCs minor partners suffering the most electorally).

Conclusion
We have investigated the electoral success for parties joining so-called GCs compared to the electoral success for other and especially 'new' parties. GCs have long been considered a last resort for mainstream parties. Unable to govern alone or with traditional allies, parties can choose to form a GC with long-term mainstream rivals. However, in the past two decades, several European countries have more willingly entered into GCs which have historically been limited to periods of war or national emergency.
We have confronted this timely topic in European party politics by seeking to explain what election results brought political parties to form GCs and which results in the same parties obtained in the general elections that happened after a GC's experience. The literature has long supported the notion that mainstream parties enter into GCs as a saving move for poor electoral results in previous elections. Newer work has suggested another motive; namely, that mainstream parties enter into GCs with the aim to exclude new parties and populist movements from government majorities and the complexity of forming coalitions with minor partners with whom it is difficult to negotiate. Here, using both qualitative and quantitative evidence, we show that GC parties do suffer electorally in subsequent elections although the entrance of mainstream parties to GCs may be unrelated to their electoral losses in previous elections. That is, while political parties, in general, form GCs when they have lost votes, for mainstream parties this is not necessarily the motivating factor. Poor performances in previous election performances by mainstream parties do not appear to be the central reason for becoming GC members. This corresponds to newer theorizing about the impetus to participate in GCs as a means to exclude new and more extreme challenger parties. Notes: (t statistics in parentheses) *P < 0.05, **P < 0.01, ***P < 0.001. Bold values are just the most relevant to our analysis. Source: Comparative political dataset (2018).
Assessing the direct effect of new parties on GCs' formation is more complicated. While it is true that in some cases (Bulgaria in 2005 and Italy in 2013), the rise of new parties had an effect on the birth of GCs, the number of actual events is too small to discern the validity of such a claim. The electoral presence of new parties has certainly restricted mainstream parties in their government formation optionsand forced them to form not only GCs but also other 'unlikely' coalitions (such as the 2018 M5S-Lega government majority in Italy) and minority governments.
At the same time, and congruent with the literature, being a GC member is statistically related to lower electoral performance in following elections. And this appears to be true for both major and minor GC members. Thus, as theorized, GCs do not profit from mainstream parties that have been part of them. It seems that voters dislike GCs and 'punish' the parties that joined them. While GCs may be perceived as a potential tool of resistance, both mainstream and minor parties are penalized by voters in the next round of elections.
This study also opens room for potential future research. For instance, in investigating whether the distribution of cabinet portfolios in coalitions might matters. If a party in a GC is not able to win control over ministerial posts that are salient for it (or its voters), then one could expect that this party is likely to suffer in the polls and will lose votes in the next election. Other studies could be conducted upon the performance of populist parties, which have been certainly on the rise in this same period.
Taken together, the empirical evidence confirms some of the literature. While electoral losses do not seem to be a key factor in driving mainstream parties into GC, in the elections which follow GCs, both major and minor GC member parties suffer decreases in electoral success. What is by now certain, is that GCs might help mainstream parties to hold onto power for additional years, but they decrease their chances of success in the next elections. It is likely that while GCs are increasingly considered as a feasible coalition arrangement by many European policymakers, voters still equate them to 'unholy alliances between strange bedfellows.