The Formation of Minority Governments in Central and Eastern Europe

The Formation of Minority Governments in Central and Eastern Europe Dorothea Keudel-Kaiser European University Viadrina, Frankfurt (Oder), Germany keudel@europa-uni.de First draft please do not quote without the author s permission! Paper to be presented at the 6th ECPR General Conference, Reykjavik, University of Iceland, 25th - 27th Aug. 2011 Section Comparative Political Institutions, Panel Government Formation and Survival

1 Introduction About one third of all governments in Central and Eastern Europe (CEE) are minority governments. It is characteristic that they do not control the majority of seats in parliament. At first glance minority governments seem to be a counterintuitive phenomenon. Why should a party decide to stay in opposition and forgo the benefits of government participation, but support a government that has no majority? Government formation processes in the CEE countries are a diverse, yet so far, widely unexplored area of scientific research. Primarily, research into minority governments, as well as on government formation in general, has been focussed on Western (European) countries. Are the influencing factors that are valid for the formation of minority governments in Western Europe such as institutional facts (e.g. investiture votes or an opposition s influence possibilities), or the policy positions of parties (ideological centrality of the largest party) applicable to the Eastern European context? Or, do the underlying factors differ depending on the different political background? The aim of this paper is to illustrate the combination of conditions that lead to the formation of minority governments in CEE by means of Qualitative Comparative Analysis (QCA). The first part of the article is devoted to theory on government formation: Firstly, a brief overview of government formation in general, and concerning minority governments in particular is presented, culminating with the summary of the state of the art on coalition research regarding CEE. In the second and main part of this paper, hypotheses are derived from the fully developed coalition-research and -theory concerning minority governments in Western Europe, as well as from close examination of the existing hypotheses and the empirical evidence of (minority) governments in CEE. After a short introduction of the cases and definitions (part 4), the performance of the QCA is described and the results are discussed (part 5). In the concluding section, the main results are summarized and further proceedings are presented. 2 Research context 2.A Research on coalition formation Examining the formation of minority governments is part of coalition theory with a special focus on government formation. As coalition theory has primarily focused on Western Europe, the first part of this overview refers to Western Europe, followed by an overview on the state of the art of coalition research pertaining to CEE. Government coalitions are one of the most studied fields in political sciences. 1 Within this field, the formation of governments has received the most attention, as compared to govern- 1 For an overview on coalition research and theory see for example Grofman and van Roozendaal 1997; Diermeier 2006 ; Müller 2004; Strøm et al. 2008. 1

ment duration and practice. 2 Early coalition theory, standing in the tradition of rational choice and cooperative game theory, aimed to predict the composition of coalition governments in multiparty systems, in which no party wins a majority of seats (Gamson 1961; Riker 1962). The size of the coalition and the office-seeking motivation 3 of political parties and their leaders, (conceived as rational, unitary actors), were seen as the determining factors of the coalition building. The authors predicted minimal-winning coalitions as the most probable form of coalition. That is to say, coalitions which have a majority with the minimum number of party and seats. The addition of a surplus member would be irrational, as coalition partners would have to share unnecessarily the advantages of holding office. The size criterion was soon supplemented by the integration of party system characteristics and policy-seeking motivations 4 of the actors: Policy proximity was introduced as an additional constraint on coalition formation, but office seeking was still seen as the basic motivation of the actors (Leiserson 1966). The basic assumption is that parties try to form minimum connected winning coalitions with ideologically proximate partners (Axelrod 1970). However, these theories failed to explain minority and oversized coalitions that are a quite common in Western European countries. Therefore, policy-motivation was placed at the centre stage of coalition bargaining and formation (for an overview see Laver and Schofield 1990:110 et seq.). First, onedimensional accounts of policy bargaining were prevailing (de Swaan 1973). In the tradition of spatial voting models, authors predicted that the party controlling the central place in a policy dimension (left-right-scale), the so-called median legislator, plays a crucial ( dictatorial ) role in the government formation process and will always get into the government. More recent work assumed that more than one policy dimension must be considered at the same time (Laver and Schofield 1990:119). Also the vote-seeking motivation 5 of actors became integrated into coalition theory (Strøm 1990; Müller and Strøm 1999), but as part of the policy- and office-seeking motivation: Votes can be only instrumental goals : Parties only seek votes to obtain either policy influence, the spoils of office, or both. (Müller and Strøm 1999:9). 2 While early coalition theory was concentrated on coalition formation, in the 1980s, attention shifted to the duration and the dissolution of coalitions (for example Dodd 1984; King et al. 1990; Warwick 1992). Today, it is an intensively studied field of research (for an overview see Woldendorp et al. 2000:78; Müller et al. 2008:28). The practice of coalition government is still a widely unexplored area of study, probably due to the difficulty of gathering information about the largely informal processes. But since the 1990s, there is a greatly enhanced interest in how governance occurs once a cabinet coalition has been formed. (Müller et al. 2008; Müller and Strøm 2000:572; see e.g. Kropp 2001). Recently, coalition researchers advocate a unified approach: The authors promote the idea that the different phases of coalition government should be regarded as an interconnected, dynamic process: the life cycle of coalition (Strøm et al. 2008). 3 Office-seeking party: a party that seeks to maximize the control over political office benefits (Müller and Strøm 1999). 4 Policy-seeking party: a party that seeks to maximize its impact on public policy (Müller and Strøm 1999:5). 5 A vote-seeking party: a party that seeks to maximize its electoral support (Müller and Strøm 1999:9). 2

With the neoinstitutionalist turn in political science in the 1980s, institutional variables were integrated into coalition theory. Institutional requirements are seen as constraints that influence the bargaining process and its outcome. These requirements are, on the one hand, rules under which cabinet formation takes place. Many authors stress the importance of investiture rules. 6 Others stress the importance of recognition rules that specify which party will be asked to be the formateur, that is, to form the government (e.g. Austen-Smith and Banks 1988) or analyse the influence of the head of state in the formation process (see Strøm et al. 1994:312). On the other hand, there are rules influencing the work of the cabinet when in office that are said to have an anticipated effect on government formation, for example bicameralism: In political systems with a second chamber, broadly based coalitions are said to be more likely (Müller and Strøm 2000:569; Druckman et al. 2005; Diermeier et al. 2007). Beyond the above-mentioned constraints, the party system is seen as a very important contextual factor in coalition bargaining (see e.g. Laver and Budge 1992; Müller and Strøm 2000). 2.B Coalition theory and the formation of minority governments For a long time, minority governments were neglected by coalition theory. At first glance, the formation of minority coalitions contradicts the office-seeking motivation of parties, as well as the size-principle. Why should a party decide to stay in opposition but support a government that has no majority? Minority governments are a counter-intuitive phenomenon (Strøm 1990:8) and were therefore seen as anomalies and accidents despite of the fact that about one third of coalitions in Western Europe since 1945 are minority governments (Woldendorp et al. 2000). Suggested explanations for the emergence of minority governments ranged from deep crises (e.g. von Beyme 1970:150; Taylor and Herman 1971) and a conflictual political culture (Luebbert 1984), to the extreme instability of the political systems, high fragmentation and deep cleavages (for example Dodd 1976; for an overview: see Strøm 1990:9f). Herman and Pope (1973) drew a more differentiated picture. They showed that the assumption that coalitions that are formed must be winning (that is to say hold at least a majority of seats) has to be questioned (Herman and Pope 1973:191). They demonstrated that the formation of minority governments is not infrequent and specified five reasons that may lead to the formation of minority governments. 7 They particularly stressed the importance of the size of the government: Most minority governments in Western Europe are only a small number 6 For example Strøm 1990; Bergman 1993; Strøm et al. 1994; Martin and Stevenson 2001. 7 1) Coalitions are not the normal accepted form of government; 2) there is an immobiliste situation, e.g. when certain structural features inherent in the party system severely handicap the formation of majority coalition ; 3) the government is a caretaker administration, a non-political government with a limited life-span; 4) a tradition of co-operation between coalition partners collapses; 5) there is a situation in which the party forming the government is only a small number of seats short of a legislative majority. (Herman and Pope 1973:195 et seq.). 3

of seats short of a legislative majority (Herman and Pope 1973:197et seq.). A real turn of perception took place with Strøms pioneering work on minority governments grounded in the rational-choice-tradition. Strøm demonstrates that minority government formation can be understood as the consequence of rational behaviour by the relevant party leaders, evoked by institutional incentives (Strøm 1990:23). He points to the fact that legislative and executive coalitions need not to coincide: Majority status is not necessarily the effective decision point in parliamentary legislature. (38) He poses three main arguments: (1) In the process of government formation, parties are not only motivated by office considerations, but also to a significant extent by opportunities for policy influence. (2) The attainment of cabinet portfolios is not a necessary condition for payoff. Parties in opposition may be able to exert policy influence in the legislative arena. This is especially true for political systems with strong committees. (41) (3) Political leaders take the future into consideration: The more competitive the elections ( electoral decisiveness ), the higher the expected costs of holding office, the more likely the renunciation of office (49). Bergman (1993; Bergman 1995) complements Strøms mainly rational-choice-based assumptions with further institutional factors. He applies a mixture of a neo-institutionalist and a rational-choice-approach. Using the example of Sweden, he takes the importance of government formation rules for the type of government (minority or majority governments) into consideration. He distinguishes two major government formation rules, formulated in positive and in negative terms. He points out that negative parliamentarism where a government only needs to be tolerated, but not formally invested in by parliament leads to a high number of minority governments in countries with proportional systems (Bergman 1993:60; Bergman 1995). Apart from these studies concerning minority governments per se 8, research into the formation of minority governments is included in studies that deal with the general factors that influence the type of government that will form. Variables from six different blocks are analysed: Time and space parameter, structural attributes, preferences, institutions, the bargaining environment and critical events (Mitchell and Nyblade 2008:202). With regard to minority governments, these studies confirm the importance of the mode of the investiture vote and point to other factors increasing the likelihood of the formation of minority governments. These are for example the bargaining duration, opposition influence, or the ideological position of the largest party. 9 8 Besides these two seminal studies, there is a third research project devoted exclusively to the formation of minority governments: Nikolenyi (2010) studies the formation of minority governments in India. 9 See e.g. Mitchell and Nyblade 2008:232; Crombez 1996; Warwick 1998; Reniu 2011. In more detail see section 3. 4

2.C Coalition theory and Central and Eastern Europe More than 20 years after the breakdown of communism, government coalitions in CEE are still a largely unexplored field of research: [T]he new democracies in the CEE countries provide a huge and as yet largely untapped potential to which coalition research ought to be extended. (Bergman and Ersson 2009:4). Obviously, because of the different political, social and economical context, approaches used to examine Western European coalitions cannot easily be transferred to CEE: Coalition behaviour in new democracies is an unexplored area of investigation despite much rich material in post-communist Central and Eastern Europe. Established coalition theory is relevant to such research, but it needs to adapt to special problems found in new regimes undergoing transition and not yet consolidated. (Pridham 2002:75). Only recently have studies considered coalition governments in the Eastern European countries in a general context (Kropp et al. 2002a; Kropp 2008). 10 Some authors assess the factors that lead to the (in)stability of coalition governments. 11 There is hardly any study on the practice of coalition government in CEE (two of the few exceptions: Pridham 2002; Clark and Jurgeleviciute 2008). In contrast to coalition research into Western Europe, very few studies focus on the formation of coalition governments in CEE. In the existing studies, the role of communist successor parties in coalition bargaining and formation is a central topic. Grzymala-Busse demonstrates that the fundamental predictor of coalition formation continues to be the regime divide, the depth and character of the persisting conflict between the successors of the pre-1989 Communist parties and the parties emerging from the Communist-era opposition. (Grzymala-Busse 2001:85; see also Druckman and Roberts 2007). 12 Some studies discuss single factors that might explain the government formation processes, for example presidential influence, or the influence of intra-party conflict (Nikolenyi 2003; Ştefuriuc 2004; Protsyk 2005). In an explorative study building on previous studies of coalition politics in Western Europe, Bergman and Ersson (2009) make a first attempt to elaborate which factors known to be important for coalition formation processes in Western Europe also play a crucial role in coalition formation processes in CEE. 13 Their preliminary result: 10 There are a few articles that give a general overview of the context of coalition formation, stability and praxis in Eastern European countries. For the Baltic States: Tiemann and Jahn 2002; for Hungary and Poland: Dieringer 2002; for Slovakia: Pridham 2002. 11 Nikolenyi 2004; Grotz 2007; Grotz and Weber 2011. These studies are related to other work focusing on the (in)stability of party systems, often referring to Peter Mair s model of party government (Mair 1997), see for example Toole 2000; Müller-Rommel 2005; see also Krupavičius 2001; Bakke and Sitter 2005; Tavits 2008. 12 Druckman and Roberts show that communist successor parties (CSPs) are less likely than other parties to be included in government coalitions and that coalitions that include CSPs are more likely to be oversized (Druckman and Roberts 2007:5) 13 Their dependent variables are coalition formation (government coalition or one-party government) and electoral performance of the cabinet. They use independent variables from five clusters: structure, preferences (policy distances), institutions, bargaining costs and critical events (Bergman and Ersson 2009:11). 5

The inclusion of the CEE countries implicate that some of the institutional effects that can be detected in Western Europe wash out (Bergman and Ersson 2009:18). Regarding CEE, the existence of different types of coalition governments is widely neglected and there are no studies specifically addressing minority governments although, just as in Western Europe, they make up about one third of all governments. 3 Hypotheses on the formation of minority governments In the following, the main hypotheses put forward in coalition research to explain the formation of minority governments are summarized (divided in institutional factors, party system factors and the bargaining environment) and their applicability to the Eastern European context is discussed. 3.A Institutional factors Investiture vote There is broad consensus in coalition research that the type of investiture rule influences the type of government that will form and that minority governments are particularly likely to form if there are only low investiture requirements or no investiture vote at all. 14 In line with this assumption, Bergman (1993; 1995) showed that the type of government formation rule correlates with the outcome of the government formation process. He distinguishes between two major types of government formation rules: positive and negative. Positive means that a party or coalition that wants to come into office needs an explicit level of vote support: It has to win a vote of investiture in the parliament. Negative means the absence of a positive threshold: The new government need[s] only to be tolerated by parliament. (43) The negatively formulated rules are, according to Bergman, associated with a high frequency of minority governments. All CEE countries have, according to Bergman s definition, positively formulated formation rules. But they differ in their precise form: While in six out of the ten CEE countries, a simple majority of votes is sufficient, in Bulgaria, Hungary, Latvia and Slovenia, parties or coalitions need an absolute majority to get into office. 15 That is why it seems to be appropriate to adapt Bergman s typology to the CEE context: In the study at hand, the countries are divided into those where the government needs an absolute majority to get into office and those where a simple majority is sufficient. Previous studies on the CEE countries confirm the appropriateness of adapting this typology: Simple majority requirements correlate with a higher frequency of minority governments (Nikolenyi 2008; Strohmeier 2009: 270 et seq.). 14 Budge and Laver 1992a:12; Strøm 1990:78; Martin and Stevenson 2001:36; Nikolenyi 2008:18; Mitchell and Nyblade 2008:219. 15 See the constitutions of the countries and Ismayr 2010; Döring and Hönnige 2008. 6

Hyp. 1: Investiture vote (simple majority) plays a role in the formation of minority governments. Committees Under the presumption that parties are not only interested in office and vote, but also in policy, the scopes of influence of the opposition attract notice. Strøm (1990) could show for the Western European countries, especially for the Scandinavian countries, that the ability of the opposition parties to exert policy influence make the formation of minority governments more likely (Strøm 1990:40; 71, see also Juul Christiansen and Damgaard 2008:50). Strøm emphasizes the influence via committees as the most important channel of influence. 16 Von Steinsdorff (2011) shows for the Eastern European Countries that the degree of the opposition s influence is relatively equal: She classifies the potential influence of the opposition through the committees in 9 of the 10 CEE countries as high. 17 Because of this uniformity the formation of special types of governments cannot be traced back to the variable influence via committees. Other factors The mode of influence of other institutional factors is contested. This is true for the mode of the constructive motion of no confidence. Some researchers claim that its existence makes the formation of a minority government more likely: The more difficult it is to bring down the government, the more likely the government is to stay in office even if it has no majority and the more likely a minority government is to form (Strohmeier 2009:272 et seq.). Others argue the other way round and state that the existence of a constructive motion of no confidence makes the formation of minority governments less likely (see e.g. Mitchell and Nyblade 2008:219). The empirical evidence in CEE seems to disprove the assumption that there is any link between the mode of the motion of no confidence and the type of government formed. Of the five countries making up more than one minority government, two have a vote of no confidence requiring an absolute majority (Bulgaria and the Czech Republic), one has a constructive vote of no confidence (Poland) and two have a vote of no confidence requiring only a simple majority (Latvia and Romania). Another factor which is ascribed to favour the formation of minority government is a powerful prime minister: When the prime minister and his party have exclusive rights, which improve 16 He names five indicators: (1) The number of standing committees (more than 10 committees stand for high oppositional influence), (2) fixed areas of specialization which (3) correspond to ministerial departments, (4) the number of committee assignments per legislator (if there are restrictions, legislators are more likely to be specialists which enhances the status of the opposition), (5) committee chairs are proportionately distributed (which is to say that the government does not control all chairs). 17 The indicators introduced by Strøm are very similar to those used by Döring (1995), which von Steinsdorff (2011) applies to the Eastern European parliaments. She comes to the somewhat surprising result that committees in CEE legislatives have much more autonomy and influence than committees in the Western European legislatives. 7

their scope of action compared to that of the coalition partners, government participation becomes less attractive the formation of minority government becomes more likely: [T]he more the institutional design rules advantage the party of the prime minister relative to other coalition partners, the less appealing it should be to be a junior partner. (Mitchell and Nyblade 2008:219). There are two reasons why this assumption is not applicable to the CEE countries. First of all, two different datasets on the positions of Prime Ministers inside the cabinet, one based on the constitutions (Woldendorp et al. 2000) and one based on minister interviews (Blondel et al. 2007) come to the same result: There is not much variation in the position of prime ministers in the cabinet in the CEE countries, especially compared to the variety of the powers of prime ministers in Western Europe. Thus, the variable cannot tell us a lot about the different frequency of formation of minority governments in the CEE countries. Anyway, the results of the study by Blondel et al. (2007) indicate that the position of the prime minister does not have any impact on the specific shape of the government: There is in effect no difference among single-party cabinets and coalition cabinets, whether of the minority, minimum winning or surplus majority character, in terms of the proportion of strong ministers [...]. (Blondel et al. 2007:186). So, even if there might be variety in prime ministers power, the exclusion of the variable prime ministers power seems justified against this background. 3.B Party system features Party system characteristics are seen as crucial factors influencing the government formation process. 18 Of special importance are explanations referring to (1) the numerical dimension as well as explanations referring to (2) the ideological dimension. Overview: Party system characteristics influencing the government formation process (A) Position of a single party (1) Numerical dimension (2) Ideological dimension Seat share of the largest party Existence of an ideologically central party ( median party, core party ) (B) All parties Degree of fragmentation Distribution of ideological preferences Seat share of extremist parties (1) Numerical dimension (A) The formation of minority governments is said to be more likely when the seat share of the largest party is large (near to 50%) so that this party can easily form a minority government without being challenged by a powerful opposition (Herman and Pope 1973:197). As 18 Budge and Laver 1992b; Kropp et al. 2002b:26; Müller et al. 2008:21. 8

the seat share variable is usually combined with the ideological dimension, it is subsumed under this point (see below). (B) High degrees of fragmentation, referring to the number and relative strength of parliamentary parties, are supposed to increase information uncertainty in the coalition formation process and are consequently supposed to lead to deviations from the minimum winning status. While some coalition researchers assume that high fragmentation leads to the formation of undersized OR oversized coalitions (Dodd 1976:68; see also Crombez 1996:3), more recent studies could show that high degrees of fragmentation are linked to oversized governments as high fragmentation leads to overall bargaining uncertainty about each other s preferences (Mitchell and Nyblade 2008:216). While low fragmentation renders the formation of minimal winning coalitions more likely, a moderate degree of fragmentation is said to enhance the likelihood of minority governments to form: Nikolenyi (2008) as well as Grotz and Weber (Grotz 2011) show for the CEE countries that a moderate degree of fragmentation, defined as an effective number of parliamentary parties less than five (Sartori 1976) makes the formation of minority government more likely. It is not clear how the five per cent threshold is deduced from Sartori, or why it should be appropriate for the Eastern European context. Also, the empirical evidence is not as straightforward as stated by the authors. As the following table shows, a medium degree of fragmentation is equally linked to minimal winning, minority, and surplus governments. ENPP MW Min SP 3 8 1 2 > 3 < 5 17 8 7 5 7 3 3 Source: Compiled by the author; data: Döring and Manow 2010; Müller-Rommel et al. 2008; Popescu and Hannavy 2001. ENNP: Effective Number of Parliamentary Party (Laakso and Taagepera 1979); MW: Minimal winning; Min: Minority: SP: Surplus Additionally, the hypothesis that high fragmentation is linked to surplus governments seems to be rejected by the empirical evidence. As neither the theoretical nor the empirical evidence is convincing and as the seat share variable is also included in other conditions (polarisation), fragmentation is excluded here. (2) Ideological dimension (A) In the tradition of spatial voting models, coalition researchers point to the strategic advantage of the party being situated in the centre of the decisive policy dimension. The main prediction is that the party controlling the central place in the most important policy dimen- 9

sions, the so-called median legislator or core party, plays a crucial ( dictatorial ) role in the government formation process, will always get into government and can decide to form a government without controlling a majority of seats (Laver and Schofield 1990; Lijphart 1999:95; Warwick 1998:31). But the ideological explanation alone is not sufficient to explain the formation of minority governments. A party might have an ideologically central position, but still not sufficient power in terms of seat share. Therefore, the numerical and the ideological dimension are combined: The basic hypothesis is that the greater the seat share and the greater the ideological centrality of the largest party, the more likely minority government formation will occur. 19 Hyp. 2: The position of the biggest party (biggest party = median party) plays a role in the formation of minority governments. The chances that a minority government can come into office are said to get bigger the larger the ideologically division of the opposition (polarisation). 20 On the one hand, the emergence of minority governments in ideologically divided parliamentary systems can be explained by the fact that ideologically divided opposition parties are not able or willing to form a worthwhile replacement. 21 Alternatively, it can by explained by the value of government: If the parliamentary parties are ideologically divided, potential majority governments would be ideologically heterogeneous, but ideologically diverse governments are expected to be less valuable: The reason that polarisation should encourage the formation of smaller government types is that greater distances among parties augment the importance of policy costs relative to portfolio benefits and thereby make membership in larger governments less attractive. (Warwick 1998:322) Hyp. 3: A high index of polarisation plays a role in the formation of minority governments. There is a second argument with regard to the policy positions of parties: Minority governments are said to be more likely to form the bigger the seat share of extremist parties (Sartori 1976; Laver and Schofield 1990). The basic assumption is that extremist parties are excluded from the government formation thus reducing the number of feasible coalitions (Martin 19 If the largest party is large and centrally located, it has a favorable bargaining position. Since it is large, it is likely to keep control of the government formation process. Moreover, government policy is close to its policy position, because it is centrally located. Therefore, it does not have to offer other parties government portfolios to attract support for a government and it can form a minority government. (Crombez 1996:16, see also Schofield 1993; Laver and Shepsle 1996; Martin and Stevenson 2001: 37; Mitchell and Nyblade 2008: 218). 20 Martin and Stevenson 2001:35, see also Warwick 1998; Mitchell and Nyblade 2008:218. 21 Warwick 1998:31; see also Budge and Laver 1992; Laver and Schofield 1990:79. 10

and Stevenson 2001:37). 22 There are, however, some pitfalls with this assumption: The basic assumption in coalition research that extremist parties are not suitable for coalitions is not completely transferable to the Eastern European context. Extremist right-wing parties which are seen as non-coalitional in the Western (European) countries are not seldom government members in the CEE countries as for example the Slovakian National Party SNS which was not only part of the HZDS regime, but also junior partner of the social-democratic Smer- SD (2006-2010). On the other hand, there are parties which are usually not classified as extremist, as the Czech communist party KSCM (at least since the late 1990s when the party got much more moderate), whose ability to engage in coalitions is restricted. Therefore, in the study at hand, the factor existence of extremist parties is replaced by the factor existence of parties not suitable for coalitions. Hyp. 4: A large seat share of parties not suitable for coalitions plays a role in the formation of minority governments. 3.C Bargaining environment Inconclusive bargaining rounds A basic assumption, also underlying some of the above described hypothesis concerning institutional and party system factors, is that minority governments are more likely to form the weaker the incentive of office seeking. Long formation duration and inconclusive bargaining rounds are considered to indicate a lower value being placed on office holding (Mitchell and Nyblade 2008:229). Indeed, the authors can show for the Western European countries that inconclusive bargaining rounds make the formation of minority governments more likely. However, time to formation, is not significant in explaining minority governments, but in explaining surplus governments. This result corresponds to the empirical evidence that in the CEE countries, on average, formation of a minority government did not take more time than that of minimal winning or surplus coalitions. 23 Hyp. 5: Inconclusive bargaining plays a role in the formation of minority governments. Electoral decisiveness Strøm can show for the Western European countries that the formation of minority governments is more likely when elections are decisive. That is to say, when there are clear governmental options for the voters ( identifiability of government alternatives ), when there is 22 This assumption did not remain unchallenged. Strøm (1990) could show for the Western (European) countries that there is no tendency for minority governments to form in more polarized environments than majority coalitions. (polarisation here: existence of extremist parties) (Strøm 1990:66; see also Bergman 1995). In the CEE countries, however, the seat share of extremist parties is often large and there is strong empirical evidence that the factor is important (Grotz and Weber 2011). 23 The average time to formation for minimal winning governments in CEE (44 days) has even been longer than the average time to formation for minority (42 days) or surplus governments (42 days). 11

significant fluctuation in the distribution of seats between the parties from election to elections ( electoral competitiveness or volatility 24 ), when the government is formed by parties that have gained rather than lost in the elections ( electoral responsiveness ), and when governments are formed in close proximity to general elections ( proximity ) (Strøm 1990:47). When these factors are considered, the electoral consequences of government participation are significant and therefore the value placed on government membership is low: If elections are decisive, there is a trade-off between a party s short term office- and policy-objectives, and its longer-term electoral incentives (Strøm 1990:44). Government incumbency could harm parties future electoral success and parties might decide to stay in opposition and to support a minority government. To what extent are these variables transferable to the Eastern European context? Strøm himself pointed to the fact that the scope of his explanation is limited: It is restricted to political systems with stable party systems: [T]he decision makers should expect to encounter roughly the same competitors and coalition partners. If the future is perceived as uncertain, unpredictable, and uncontrollable, it makes little sense to defer the gratification of holding office. (Strøm 1990:53) There are reasons to believe that in CEE where the party systems are often still in flux and volatility is still comparably high (Shabad and Slomczynski 2004), the future is at least partly perceived as uncertain and unpredictable (see e.g. Niedermayer 2008:365; Bos and Segert 2008). Volatility goes not, as in Strøm s assumption, hand in hand with the identifiability of government alternatives. Additionally, the last variable, proximity, is not applicable to this study because only governments formed after elections are included in the sample. Against this background, it seems appropriate to omit this set of variables. 4. Definitions and cases The term Central and Eastern Europe (CEE) is used here for the new member states of the European Union (accession 2004/2007): Poland, Hungary, the Czech Republic, Slovakia, Estonia, Latvia, Lithuania, Slovenia, Romania and Bulgaria. Minority governments are governments that do not control the majority in the legislature. Minority governments can be single party-governments as well as party coalitions. In contrast to the usually used very broad definition of the formation and duration of governments 25 the project at hand includes only governments formed after elections as there are different logics behind the formation of minority governments after elections, compared to those formed during the legislative period: While the former are built in a regular process and under comparable conditions, the latter 24 Strøm measures volatility as the fluctuation in the distribution of seats between the parties from election to election, following Pedersen (Pedersen 1979; Strøm 1990:73). 25 Usually, the following conditions are regarded sufficient conditions for a change of government: any change of prime minister, any change of parties presented in the government, any general election, any by-election resulting in a change in the government s parliamentary basis from majority to minority status, or vice versa. See for example Woldendorp et al. 2000, Müller-Rommel et al. 2008. 12

are caused by crisis of the precedent government. As the factors influencing the formation of the minority governments are the subject under investigation, the two different types of governments have to be examined separately. The investigation period ranges from the first free elections in the early 1990s up to 2010. Counting only post electoral governments, there are 56 governments in total (as compared to 117 governments using the broad definition): 32 minimal winning governments (MW), 12 minority governments (Min) and 12 surplus governments (SP). Overview: post electoral governments in the CEE countries, first free elections up to 2010 Country MW Min SP Total Bulgaria 3 2 1 6 Czech Republic 3 3 0 6 Estonia 5 0 0 5 Hungary 2 0 4 6 Latvia 2 2 2 6 Lithuania 3 1 1 5 Poland 3 2 1 6 Romania 2 2 1 5 Slovakia 5 0 1 6 Slovenia 4 0 1 5 32 12 12 56 Source: Compiled by the author; data: Döring and Manow 2010; Müller-Rommel et al. 2008; Popescu and Hannavy 2001. 5 Empirical Analysis: Method and operationalization 5.A Method With 56 cases (12 minority governments compared to 44 other governments formed after elections) the number of cases is intermediate. As the number of cases is too small for a statistical analysis and too large to conduct in-depth case studies for each of the cases, Qualitative Comparative Analysis (QCA) (Ragin 1987) that is suitable for intermediate-ndesigns is an appropriate approach. QCA is a case-oriented approach based on Boolean algebra. The particularity of QCA is the concept of multiple conjunctural causation (Rihoux and Lobe 2009:223) which differs from conventional research approaches: Typically, researchers focus on the net effects of causal conditions (Ragin and Sonnett 2005:180), and treat each causal condition as an independent cause of the outcome (187). The strength of the QCA method lies in working out key combinations of conditions leading to the outcome of interest. The QCA approach can be attributed to the European politics tradition as compared to a more quantitative, deductive approach working with a large number of variables and cases, 13

often based on game theory, which aims at testing formal theories and creating formal models. 26 The quantitative, game theoretic approach is prevailing in coalition research, although it is often criticized. 27 Today, coalition researchers argue for a combination of deductive, statistical large-n studies and inductive small-n studies (e.g. Bäck and Dumont 2007; Kropp 2008:531). As the QCA is a medium-n-approach, which allows for a certain degree of generalization, but at the same time a case-sensitive approach devoting attention to every single case (Rihoux and Lobe 2009), the use of QCA in coalition research can contribute to the closure of the above described research gap. Furthermore, it contributes to the modelling of the interactions between the various type of explanatory variables used to understand coalition formation (Mitchell and Nyblade 2008:233), a point so far neglected in coalition research. Here, the crisp set version of QCA is used. To perform the QCA, the variables (in QCAterms: conditions) have to be dichotomized, where 1 stands for: condition is likely to lead to a positive outcome (= the formation of minority governments); 0 for the opposite. 5.B Operationalization of the conditions From the above introduced hypotheses, the following conditions can be outlined as playing a role in the formation of minority governments: the investiture vote (simple majority), the existence of a median party which is at the same time the largest party, a high degree of polarisation, the existence of non-coalitionable parties and inconclusive bargaining rounds. The operationalization of the condition Investiture vote is straightforward as it refers to the constitutions. Formation rules requiring an absolute majority are coded as 0; formation rules requiring a simple majority as 1. The next two conditions, median party and polarisation, refer to the policy positions of the parliamentary parties. There have been a number of attempts to locate parties in the policy space 28 expert surveys (e.g. Benoit and Laver 2006; Hooghe et al. 2010), manifesto data (Comparative Manifesto project, Klingemann et al. 2006), and a new approach combining expert judgements, textual analysis and the parties own judgement (Trechsel and Mair 2009), all of which have their respective pitfalls (for a discussion of the 26 Budge and Laver 1992; de Winter et al. 2002:3; Kropp 2008:615, see also Müller et al. 2008. 27 The main criticism is that the adding of more and more variables leads to an ever-higher complexity of the formal models, while the operationalization of variables tends to be more and more crude. Some of the models are blamed for having lost touch with reality: [R]esources would be more efficiently invested in thick descriptions, rather than in testing ever more complex formal theories with ever more sophisticated statistical techniques a line of research that has not significantly enhanced our understanding of real world formations during the past decade. In other words, coalition researchers should abandon their computer screens and get their hands dirty again. (de Winter 2002:206; see also Pridham 1986:1; de Vries 1999:241). An example for a convincing countermovement : Andeweg et al. (2011) use case studies to overcome these shortcomings and to understand why numerous formation processes and outcomes diverge from theoretical predictions. 28 In coalition research, the location of parties in a common policy space has been introduced by Downs (1957) and has been further elaborated by de Swaan (1973) and Dodd (1976). 14

different approaches see Electoral Studies 26/2007; Marks 2007). As changes in party position over time have to be considered, there is no alternative to using the manifesto data (EMP) as the expert surveys cover only single points of time. The use of the manifesto data in coalition research is contested as policy positions are inferred from salience measures still it is commonly used in coalition research (see e.g. Bergman 1995; Mitchell and Nyblade 2008). Where possible, the data are crosschecked with Expert survey data. The median legislative party is determined by using the left-right scales provided by the EMP data sets in conjunction with seat share data (Bergman et al. 2008:98) 29. If the median party is at the same time the biggest party, it is coded as 1. To measure the polarisation, that is to say the ideological spread of the parliament, the polarisation index as a measure of the standard deviation of the parliamentary party system as introduced by Warwick (1998) resp. Dodd (1976) is used. With regard to Central and Eastern Europe there are no robust criteria in the literature for defining whether a parliamentary party system is highly polarized or not. Therefore, this condition was dichotomized using the median value technique (for this proceeding see Bäck and Dumont 2006:51). Thus, an index of polarisation exceeding 10,8 is coded as 1. As explained above, parties are classified as non-coalitionable if they are excluded from coalition bargaining, given that they are extremist parties, or for other reasons. Grotz and Weber (2011) who look at the influence of a high degree of extremist parties on government formation introduce a threshold of 10% as a high seat share of extremist parties. They do not give a theoretical explanation, but here, this threshold is chosen because it comes near the median value. 30 Thus, a seat share of more than 10% is coded as 1. The measure of the last condition, inconclusive bargaining rounds, is again straightforward: If there is an inconclusive bargaining round, it is coded as 1. The assignment is based on descriptions of the government formation processes (e.g. Electoral studies; European Journal of political research). 5.C Performance of the QCA The first step of a QCA analysis is to create a truth table showing the different combination of conditions (configurations) associated with the outcome (Rihoux and de Meur 2009:44). There are configurations with an [1] outcome (minority government) and configurations with 29 The objection that the use of the left right dimension alone is not sufficient can be disproved in two ways: First of all, the left-right dimension is composed of a multitude of dimensions, also including cultural issues such as traditional morality (Klingemann et al. 2006:5; the authors refer to the leftright positions of parties as given in Laver and Budge 1992). Second, Klingemann et al can show that the Left-Right framework can indeed be used for the CEE countries [W]e can certainly say that the analyses of this chapter do not refute the idea that voting and party competition occur primarily in a one-dimensional Left-Right space. (Klingemann et al. 2006:27). 30 The median value is 11,9%, but there are some outliers. 15

an [0] outcome (majority governments). Besides these two types, there are contradictory configurations [C] that lead for some cases to a 0 outcome, for others to a 1 outcome. 31 There are different program that can be used to perform the QCA. Here, the program Tosmana (Cronqvist 2007) is used. Truth table including ali 56 cases and the five conditions described above (Inv: investiture vote, Med:median party; Pol: polarisation; nonc: non-coalitionable parties; Inc: inconclusive bargaining rounds). v1: Inv v2: Med v3: Pol v4: Nonc v5: Inc O: Outcome id: Fall v1 v2 v3 v4 v5 O id 0 0 0 0 0 C Bul9,Hu42,Hu43,Lat59,Slov127,Slov128 0 1 1 0 0 0 Bul12,/Hu45(Med:1)(Pol:1)/,/Lat63(Med:1)(Pol:1)/,/Lat67(Med:1)(Pol:1)/ 0 1 0 0 0 C Bul14,Bul17,Hu40,Hu41,Lat49,Lat55,Slov124 0 0 1 0 0 0 Bul15,Slov122,/Hu45(Med:0)(Pol:1)/,/Lat63(Med:0)(Pol:1)/,/Lat67(Med:0)(Pol:1)/ 0 0 1 0 1 0 Bul16 1 0 1 1 0 C CR18,Rom95 1 0 0 1 0 C CR19,Rom98 1 1 1 1 1 1 CR21,/CR25(Med:1)(Pol:1)/ 1 1 1 1 0 0 CR22,/CR28(Med:1)(Pol:1)/ 1 1 1 0 0 0 Est29,Est35,Pol88,Rom104,Slok115,/Est39(Med:1)(Pol:1)/,/Lith76(Med:1)(Pol:1)/,/Lith79(Med:1)(Pol:1)/ 1 1 0 0 0 0 Est31,Est37,Lith68,Lith71,Pol94,Rom107,Slok109,Slok114,/Est39(Med:1)(Pol:0)/,/Lith76(Med:1)(Pol:0)/,/Lith79 (Med:1)(Pol:0)/ 0 0 1 1 1 0 Lat51 1 0 0 0 0 C Lith74,Pol83,Pol86,Slok113 1 0 1 0 1 1 Pol80 1 0 0 0 1 1 Pol91 1 1 0 1 0 1 Rom103 1 0 1 0 0 0 Slok117,/Est39(Med:0)(Pol:1)/,/Lith76(Med:0)(Pol:1)/,/Lith79(Med:0)(Pol:1)/ 0 0 1 1 0 0 Slov119,/Hu48(Med:0)(Pol:1)/ 1 1 0 1 1 1 /CR25(Med:1)(Pol:0)/ 1 0 0 1 1 1 /CR25(Med:0)(Pol:0)/ 1 0 1 1 1 1 /CR25(Med:0)(Pol:1)/ 0 1 0 1 0 0 /Hu48(Med:1)(Pol:0)/ 0 1 1 1 0 0 /Hu48(Med:1)(Pol:1)/ 0 0 0 1 0 0 /Hu48(Med:0)(Pol:0)/ 1 1 0 0 1 0 /Slok118(Med:1)(Pol:0)/ 1 1 1 0 1 0 /Slok118(Med:1)(Pol:1)/ It is clear that substantial modifications have to be made: From 26 possible combinations, five configurations comprising 21 out of 56 cases are contradictory configurations. Only five out of the 12 minority governments can be explained. Additionally, there are counterintuitive configurations: configurations showing hardly any a [1] condition, but leading to a [1] outcome or the other way round (see e.g. row 9). 31 Solutions in brackets (see truth table): missing data; the program shows the different possible solutions. 16

Detecting contradictory configurations in the course of a QCA is perfectly normal ; it is an integral part of a csqca analysis and does not mean that the researcher failed (Rihoux and de Meur 2009:48). Addressing them is part of the iterative process of a QCA between theory and cases (Rihoux and Lobe 2009:229). There are two strategies to solve the contradictions: reconsidering the outcome variable or/and reconsidering the conditions. 5.D Re-codification of the outcome and the conditions First, the outcome variable, minority governments [0] as compared to governments commanding a majority [1] (minimal winning or surplus governments), is reconsidered. As shown above (part 2), in coalition theory the formation of minimal winning coalitions are seen as the normal outcome of a government formation process. Indeed, they make up the biggest part of governments in Eastern as well as in Western Europe. Further more, coalitions researchers could show that different factors underlie the formation of the three types of government (minority, minimal winning, surplus) (Mitchell and Nyblade 2008). Therefore, it might be a solution to exclude all surplus governments from the analysis and to compare minority governments only to the expected outcome, minimal winning governments. The result: There are still 4 contradictions including 16 cases (see appendix, table 2). Another source of error could be the time factor: All governments formed from the first free elections up to 2010 are included in the same analysis. But there might be differences due to the different stages of consolidation of the new democratic systems: After the second legislative period (ca. 1998) that coincides with the intensified EU accession process, all the political systems reached a certain level of consolidation. 32 Therefore, the first two regular legislative periods are now analysed separately from the following ones. The new result still includes many contradictions (see appendix, table 3 and 4). Additionally, there are too many conditions compared to the number of cases, leading to a huge number of possible combinations of conditions and a growing number of logical reminders. It might be a compromise to exclude only the first free elections that definitively took place under special conditions not comparable to the later elections. Indeed, the result is a little more satisfying, but there are still two contradictions including eight cases and three counterintuitive rows (see table 5). Maybe the definition of minority governments as governments that do not control the majority of seats in parliament is not precise enough. There is a qualitative difference between minority governments depending on their support in the legislative arena. Minority governments can be formal or substantive : while the former rely on negotiated, long-term binding agreements of support with other parties, the latter have no such (long-term binding) agreements (Bergman 1995:29; Strøm 1990:62). If the minority government has a legislative ma- 32 There are, however, several pitfalls with this quite schematic classification. To mention only one: The degree of consolidation differs between the countries. In Bulgaria and Romania, for example, the cut should be made later than after the two first legislative periods. 17