Abstract
Studies of government participation in established democracies demonstrate that ideological factors significantly influence whether or not a party gets into government. Thus far, research on government participation in Central and Eastern Europe (CEE) indicates that ideological considerations have been insignificant. This is unexpected given what we know from the literature on parties and party systems in the region. Party systems have become more stable and parties themselves rapidly developed identifiable policy platforms. I argue that one of the reasons for this disparity between the government participation literature in new and established democracies is the failure to understand the ideological space across the CEE region. Although party competition can be conceptualized as a one-dimensional space in each country, the policies that underpin notions of ‘left’ and ‘right’ vary. Imposing a definition of the ideological space that was developed for West European countries ignores the ideological context of CEE. This article finds that when the left–right space is defined in a way that is meaningful to CEE countries, ideological factors are highly significant indicators of government membership across the region. Specifically, ideological proximity to the formateur and proximity to the median significantly increase a party’s probability of participating in government.
Introduction
Who gets into government and why? These are the key questions for scholars of coalition formation. The obvious answer is the ‘party that won the election’, but while it is true that the largest party in parliament tends to govern, in most democracies it rarely governs alone. It is for this reason that we must consider the factors that determine which particular parties get to participate in government as part of a coalition.
Over the years the coalition formation. literature has been one of the most vibrant and extensive in political science. Many scholars have offered explanations of why certain coalitions form while others fall by the wayside, ranging from government or party size (Axelrod, 1970; Riker 1962; Von Neumann and Morgenstern, 1953) to institutional considerations such as the role of the informateur (Austen-Smith and Banks, 1988; Baron, 1993) or investiture requirements (Strøm et al., 1994). The role of party ideology has also been scrutinized. One of the earliest variations on size principle approaches to coalition formation suggested that parties forming the government should be connected along an ideological scale (De Swaan, 1973). More recent empirical analyses have tended to include measures of distance from the ideological median or ideological compactness in their models of coalition formation (Martin and Stevenson, 2001; Matilla and Raunio, 2004; Warwick, 1996).
However, the vast majority of this research has been conducted using data for established democracies in Western Europe. Such countries have undergone decades of party system development, which has facilitated the growth of political parties with structured programmes and easily identifiable ideological profiles. Much less work has been done on coalition formation in the newer democracies of Central and Eastern Europe (CEE). Despite some perceptions to the contrary, the evidence shows that CEE party systems are generally stable – or stabilizing – and that party policy platforms have become more identifiable. We would therefore expect that party competition for government takes place on a similar basis to that in Western Europe, where ideological factors are important indicators of coalition membership. However, the evidence thus far shows that ideology is unimportant in determining which parties get into government in CEE. If parties and party systems have developed in a way indicating that ideology should be important in influencing party competition for government, why does the evidence indicate the contrary?
In this article, I argue that, contrary to previous research, ideological competition is one of the most influential determinants of coalition membership in CEE countries. Previous research has indicated otherwise, largely due to a failure to take into account the specific ideological context of CEE countries. Both ideological distance from the formateur and median significantly account for participation in government in CEE countries. In contrast to research on coalition membership in more established democracies (Glasgow et al., 2011; Martin and Stevenson, 2010; Matilla and Raunio, 2004), parties at the ideological median in CEE gain no significant advantage in seeking access to government.
This article uses an established expert survey dataset (Benoit and Laver, 2006) that covers government formations between 2000 and 2004, but I also make use of a new dataset of expert judgments of party policy positions that covers a longer time period (1990–2004). These new data are used to show that ideology is central to party membership of coalition governments in CEE. The data are also used to re-emphasize Benoit and Laver’s (2007) and Tavits and Letki’s (2009) contention that ideological competition in CEE needs to be operationalized correctly and that traditional meanings of ‘left’ and ‘right’ need to be reconsidered when analysing some countries in the region.
Ideology and coalition formation
Party ideology has long been considered to be one of the most influential factors in the coalition formation process; for example, early theorists such as Axelrod (1970) and De Swaan (1973) suggested that parties forming a coalition should be connected along the ideological scale. Over time, the role of ideology has been conceptualized in a number of different ways, such as the central player (Van Roozendaal, 1992, 1993), core party (Laver and Schofield, 1990) or strong party (Laver and Shepsle, 1996) hypotheses. Although these studies contend that party size is also relevant in gaining access to government, they all assert that the position of parties in either the ideological or policy space is crucial to participating in a coalition as well.
Many scholars have drawn on these theoretical outlines of the role of party ideology to specify large-scale empirical studies of coalition formation in parliamentary democracies. Franklin and Mackie (1984) produced one of the first forays into this particular approach to coalition formation and found that ideological considerations added significantly to their estimates of government-formation in West European democracies. Since Franklin and Mackie, ideology has been specified in different ways in empirical studies. The most notable recent contribution to the literature on government formation by Martin and Stevenson found that ideological divisions reduced the likelihood that a particular coalition will form (Martin and Stevenson, 2001: 41). Warwick (1996) and Matilla and Raunio (2004: 280–281) also assessed coalition formation in West European democracies, but rather than focusing on coalition composition, they used individual parties as the dependent variable. Both studies have shown that coalition partners are likely to be ideologically closer to the formateur or Prime Minister’s party. The latter demonstrated that the median party has a significantly greater likelihood of become a member of a coalition government (Matilla and Raunio, 2004: 281), a finding that also emerged from Martin and Stevenson’s study (2001: 41–43).
The enduring intuitive appeal, and indeed empirical evidence, of the influence of ideology in the government formation process has meant that it is usually included in research designs that look at discrete aspects of coalition formation. For example, Tavits (2008a) has considered the extent to which defectors from coalition governments are punished in subsequent formation opportunities. She finds that while ideological distance from either the median or formateur is irrelevant, a party is significantly more likely to coalesce with parties they had previously defected from if that party is at the ideological median (Tavits, 2008a: 501). Others have compared the effect of ideology to factors such as the role of incumbency in coalition formation (Martin and Stevenson, 2010), while Bäck et al. have shown that the median party is a significant predictor of portfolio allocation in the government-formation process (Bäck et al., 2011: 460).
This overview of the literature on the role of ideology in coalition formation demonstrates that ideological proximity to the formateur and both the median party and distance from the median party are significant indicators of who gets into government in established democracies. The former is intuitively logical, as it makes sense that formateur parties will seek to minimize the distance between themselves and their coalition partners as a safeguard against potential policy conflicts. The effect of the median party, and proximity to the median, is not quite as obvious, but spatial models of coalition-formation suggest that in a unidimensional policy space the party located at the median has an advantage due to its strategic position in the coalition bargaining environment. Laver and Schofield (1990: 111) suggest that this is true regardless of the size of the party. From that point we also understand that, in theory, parties that are closer to this privileged position should be more likely to participate in government.
Ideology in Central and Eastern Europe
The literature on coalition membership leaves little doubt that ideology has been an important factor in established democracies. But does it follow that ideology should be similarly important in new democracies that do not have deeply entrenched institutions and political norms? There are a number of reasons for believing that ideology may be less important for coalition membership in CEE than it is in Western Europe. Chief among these are perceived high levels of party system instability and a perceived lack of programmatic political parties for much of the period since the transition to democracy. However, these reasons are based on arguably dated preconceptions of party competition in the region.
Turning first to the party system, this is the key institution that structures the government formation process. The party system structures the interactions of political parties and, more importantly, it sets the parameters of the bargaining system within which governments are constructed. As party systems stabilize, party leaders can start to make assumptions about future alliances or the likely behaviour of individual parties and their potential ideological compatibility. This is because the ideological structure of the party system becomes familiar to the actors in the political process. When party systems are more unstable, such future orientation becomes more difficult as the range of potential governing partners may change from one election to the next. A number of scholars have noted that party systems in CEE have been unstable over the democratic period since 1990 (Markowski, 1997; Olson, 1998; Tavits and Letki, 2009), while others have compared the level of stability unfavourably with that found in Western European democracies (Lewis, 2000: 148–149; Mair, 1997: 197). Rose and Munro show that, on average, 30 parties have contested elections in CEE countries between 1990 and 2007, but an average of just two parties have fought every election in each country (Rose and Munro, 2009: 48). 1
However, it is arguable that the idea of party system instability in CEE countries is a hangover from the early part of the democratic transition. As early as 2000, Toole (2000) found evidence that party systems in some countries were stabilizing, while more recent studies provided further evidence of party system stabilization in the region, particularly in Poland, Hungary and the Czech Republic (Bakke and Sitter, 2005; Tavits, 2008b: 548). Some disagree with this assessment, asserting that the appearance of stability in CEE is a result of the cartelization of the party system which prevents ‘genuinely new’ parties from entering while levels of volatility within the system have remained relatively high (Sikk, 2005).
The development of individual parties may have had a similar effect in terms of the ability to cement governing alliances and build coalitions on a substantive ideological basis. In the early stages of democratization, parties were often characterized as having weak programmatic identities (Kitschelt, 1995) and in some countries there is evidence that parties still do not have well-defined policy platforms. Part of the reason for this may have been the restrictions that the European Union accession process in effect placed on competition on certain areas of policy together with the impetus to liberalize national economies (Innes, 2002). However, Kitschelt et al. (1999) have argued that party programmes crystallized over time, a view reinforced by recent research (Hanley et al., 2008; O’Dwyer and Kovalčik, 2007). Furthermore, Whitefield and Rohrschneider have shown empirically that there has been a great deal of stasis in party programmes, which ‘indicates that by now issue positions are connected in predictable and stable ways’ (Whitefield and Rohrschneider, 2009: 681).
While party platforms may have become more identifiable in CEE it is also true to say that they do not mirror those that are familiar to observers of Western European party politics. Furthermore, conceptions of ‘left’ and ‘right’ can have ‘nationally distinct, unstable, or unclear meanings’ in the region (Evans and Whitefield, 1998). In particular, the definitions of ‘left’ and ‘right’ are not necessarily tied as closely to economic policy in some CEE countries as they are in Western Europe (Kostelecký, 2002: 170). Parties of the left have grasped the liberal economic mantle from the right, particularly when those parties have ascended to government (Tavits and Letki, 2009: 567), while the right have often sought to demarcate their ideological territory in social or cultural terms (Hanley, 2004: 17–19). Vachudova and Hooghe (2009) have highlighted a further distinction of the ideological space in CEE. They used the 2002 Chapel-Hill survey of party policy positions to demonstrate that party competition in CEE took place along an axis which at one pole combined a conventional left-wing economic outlook with traditional cultural values. At the opposite pole liberal economics was bundled together with a greater social and cultural openness (pp. 206).
Even though party systems are generally stable – or stabilizing – in CEE, and party programmes are becoming more coherent, empirical studies of party politics are continue to find that ideological considerations have only a marginal impact on political outcomes, including on coalition formation. The earliest cross-national comparative study on this subject found that ideological considerations were secondary in government formation to whether or not parties had roots in the old communist regime (Grzymała-Busse, 2001: 87). In examining the impact that previous defection from a coalition had on a party’s chances of forming a future coalition with the same partners, Tavits found that both distance from the formateur and distance from the median were statistically insignificant. However, the median party itself was more likely to be asked to participate in a future coalition even if it had previously defected from a government (Tavits, 2008a: 501, 503–505). 2 Conversely, Glasgow et al. have shown that the median party is not more likely to provide the Prime Minister of governments that have formed in CEE, which is in contrast to their findings for Western Europe where the median party is a significant factor in determining which party provides the Prime Minister (Glasgow et al., 2011: 945). In a more limited examination of the role of ideology, Druckman and Roberts have shown that the current ideological persuasion of former communist parties is irrelevant to their chances of becoming a member of the government (2007: 17). Furthermore, the ideological diversity of coalition governments has been demonstrated to be irrelevant to its longevity in CEE (Tzelgov, 2011: 544). Although this is only tangentially related to coalition membership it adds further weight to the evidence that ideology is not relevant to party competition in the CEE region.
Reassessing ideology and coalition membership in CEE
The literature on the role of ideology in the coalition process in CEE is puzzling. On the one hand there is a wealth of evidence that suggests party systems are more stable and party platforms are now generally coherent and have been for some time. In theory, this should mean that parties in the coalition bargaining process can make informed decisions about potential coalition partners which, one would assume, also means that they can choose partners that are ideologically more compatible. In that way, coalition-formation processes should resemble those in Western Europe, where ideological position and distance are significant factors. However, the evidence to date suggests that party ideology has had little impact on either coalition membership or government survival in CEE.
One of the possible explanations for this discrepancy is that party ideology is simply being operationalized incorrectly in empirical studies of CEE. The common thread in the studies by Tavits, Garett et al. and Tzelgov is that they rely on data from the Comparative Manifestos Project (CMP) (Klingemann et al., 2006). These studies use the standard left–right dimension provided in the CMP dataset. While that variable has been shown to be reliable in a West European context there is reason to doubt that it is an appropriate measure of the left-right space in CEE. This is due to the way in which policy dimensions are bundled together to make the left–right scale. The ‘right’ is associated more with free market economic policies and traditional cultural values in CMP data, while the ‘left’ is linked closely to a more statist approach to the economy and expansion of public services (Budge, 2001: 56).
As has already been outlined, this approach to constructing a unidimensional ideological space does not fit with the reality of politics in some CEE countries. This has been further demonstrated by Benoit and Laver (2007), who collected an expert survey of party policy positions while also asking experts to place parties on a left–right dimension. They found that economic policy best predicted left–right placements in six countries, while in others social policy dimensions were the most accurate predictors of left–right placements (Benoit and Laver, 2007: 93). Glasgow et al. (2011: 943) acknowledge the limitations of the CMP data in CEE, stating that their measures of the median party are less reliable for parties in CEE countries than in Western Europe.
This begs the question: does ideology matter in determining coalition membership once the ideological space in CEE is more appropriately measured? I answer this question by analysing the effect of ideology on coalition membership using a new dataset based on an expert survey of party policy positions that, I believe, more closely represents the one dimensional space within which party competition takes place in CEE.
Following Warwick (1996) and Matilla and Raunio (2004) I assume that parties forming a coalition are more likely to be ideologically proximate so that policy differences, and therefore prospects for intra-governmental conflict, can be minimized. As the formateur takes the leading role in putting together a government, this is specified as whether or not ideological proximity to the formateur increases a party’s likelihood of getting into government. I also examine the median party hypothesis, which states that parties at the ideological median are in a privileged position in the coalition-bargaining space and are therefore more likely to become members of the government (Laver and Schofield, 1990; Matilla and Raunio, 2004). As a natural extension of this, and once again assuming that parties seek to form ideologically compact coalitions, I test whether or not ideological proximity to the median increases a party’s prospects for becoming a member of the government. Formally, I test the following hypotheses:
Data and methods
This study draws on two datasets to assess the role of ideology in coalition-participation. The first uses Benoit and Laver’s (2006) expert survey of party policy positions and covers governments formed between 2000 and 2004 in 10 CEE countries. This dataset includes 120 parties and 20 governments. From here on this is referred to as the BL dataset.
I also assess the role of ideology over a longer time period in a subset of CEE countries. This is made possible by the collection of a new expert survey of party policy positions that covers the entire period from 1990 to 2004 in five CEE countries. These countries were all part of the first wave of European Union entrants from the post-communist region: the Czech Republic, Hungary, Poland, Slovakia and Slovenia. The choice of these countries is logical and pragmatic. First, these countries were considered to have progressed sufficiently along the path of democratization and at around the same rate if one considers EU accession as a benchmark. Second, these five countries belong to the Central European bloc of post-communist countries. Finally, the expert survey data from which party policy positions will be taken covers only these five countries due to resource limitations. This dataset is referred to as the LS dataset.
The final LS dataset includes every party to have served in the legislature at any point between 1990 and 2004; 3 a total of 226 parties across the five countries. The number of observations in the analysis varies due to lack of data on certain parties; this is largely where there is no information on ideological positions due to small splinter factions forming in parliaments during legislative terms.
In this article the criteria set out by Arend Lijphart are used to determine when a new government is said to have formed: if there is a change in the party composition of the government; a parliamentary election is held; the Prime Minister resigns for political reasons; or the cabinet resigns for any reason (Lijphart, 1984: 267). The European Journal of Political Research annual yearbooks, together with Blondel and Müller-Rommel’s edited volume, Cabinets in Eastern Europe (2001) and data from Conrad and Golder (2010), were used to determine when any of those conditions came to pass. Caretaker cabinets are excluded from the data as they are qualitatively different from standard governments. A summary of the data is provided in Table 1.
Parties and governments in Central and Eastern Europe.
The focus of this article is how ideology influences which parties get into government. Therefore, the unit of analysis is the party rather than the overall configuration of the coalition that eventually forms. As this is a discrete choice outcome a logit model is used for all analysis in this article, though a probit model could equally be used. In recent years a number of authors have used one of these models to determine which parties get into government (Bäck, 2009; Druckman and Roberts, 2007; Matilla and Raunio, 2004; Tavits, 2008a; Warwick, 1996) though some have raised concerns about the conditional dependence of the data. This can be accounted for by using robust standard errors in the reporting of results (Bäck, 2009; Tavits, 2008a; Zorn, 2006) which is the method employed here.
A new expert survey of party policy
Expert surveys have been used in political science frequently over the past few decades. Most have been used to gather data on party policy or ideological positions (Benoit and Laver, 2006; Franklin and Mackie, 1984; Hooghe et al., 2010; Huber and Inglehart, 1995; Laver and Hunt, 1992; Ray, 1999), while others have been used to gather data on subjects from Prime Ministerial power (O’Malley, 2007) to portfolio salience (Druckman and Roberts, 2005). The arguments for and against using expert surveys to evaluate party policy positions have been debated at length over the years. Budge (2000) has cogently set out the case for caution when using expert surveys in political research, and Dinas and Gemenis (2010) have sounded a similar call for caution when using the CMP data. Marks et al. (2007) set out the strengths and weaknesses of both expert surveys and manifesto data. The concern of this article is that the manifesto data are not the best source of party left-right positions in the CEE region in particular.
In order to test the ideological distance and centrality hypotheses of coalition membership in CEE it is essential that measures of the unidimensional space are accurate and appropriate. As suggested previously, there is reason to believe that these measures have been inappropriate in some studies of party behaviour and coalition outcomes in CEE. Most studies relied on CMP data as the only source of party policy positions available over time. However, it has been suggested that expert survey data may more accurately reflect the ideological space in the CEE region (Benoit and Laver, 2006, 2007; Whitefield et al., 2007). The drawback of this type of data is that they tend to be a snapshot of party policy at one or two time-points.
In order to capture party policy positions within the specific context of CEE, I collected a new expert survey during 2003–2004. 4 This survey was a replication of the Laver and Hunt (1992) survey of party policy positions with some adjustments to account for contemporary political circumstances. 5
Crucially, the survey used in this research followed the example of Ray (1999) and asked respondents to place political parties on each policy dimension at each election since 1990. This allows us to assess how party policy has changed over time and provides a basis for analysing party access to government over the entire period from 1990 to 2004. This approach to the survey has the benefit of producing one of the few estimates of how party policy has changed over in time in CEE, though it does have attendant drawbacks. The pool of respondents who would be capable of answering such a comprehensive survey is limited. Rather than sending the survey to every political or social scientist in each country, the survey was targeted at experts in the party politics of each country of some years standing. 6 Furthermore, due to the somewhat exhaustive and possibly time-consuming nature of the survey, preliminary contact was made with potential respondents to ensure that responses would be as complete and considered as possible. 7
The response rate to the survey was 23 percent, which is slightly lower than the 28 percent response rate obtained by Benoit and Laver for their similarly themed survey. Overall, for such a comprehensive survey this is a reasonable return from 110 initial questionnaires, all of which were sent out only in English due to limited resources. 8 Limited resources led to another drawback of the survey in that it does not cover every country in CEE. The initial hope was to gather data on the entire 2004 wave of EU accession states; however, responses from the Baltic States proved to be unforthcoming and without either the time or capacity to boost responses from those countries a decision was made to focus on the Central European post-communist states: the Czech Republic, Hungary, Poland, Slovakia and Slovenia. 9 The mean and standard errors of each policy dimension are given in Table 3. As a rule, only parties that gained a minimum of four expert ratings are included in the analysis. In practice, this means that the observations from the early Slovenian legislatures are excluded which reduces the number of observations to 180. There are no other exclusions due to insufficient responses.
Correlations between expert survey and corresponding dimensions from the Benoit and Laver survey.
**p < 0.01, *p < 0.05.
BL = Benoit and Laver dimension.
The negative correlations between our variables and the Benoit and Laver variables on Foreign Policy and Religion arise due to the inverted scales used by Benoit and Laver. For example, Benoit and Laver’s foreign policy variable indicates that parties which strongly oppose joining the EU are given a score of 1 while our survey states that parties strongly in favour of joining the EU are given a score of 1.
Means and standard errors of party policy positions by country.
Standard errors in parentheses.
The survey is likely to have overlapped with the Benoit and Laver survey that took place at around the same time and which was a similarly modified version of the 1992 Laver and Hunt survey. However, this does provide some basis for validation of our survey data, as correlations of similar dimensions can be used. Table 2 shows that correlations between similar dimensions were significant at the p < 0.01 level on all dimensions, while the coefficients were below 0.75 only on the Environment and Urban–Rural dimensions.
Party competition in one-dimensional policy space
This article is concerned with competition for government in a one-dimensional space. We therefore need to produce a meaningful measure of that space for each country in CEE rather than seeking to impose a particular conception of that space that we know applies in established democracies. As has been outlined earlier in this article, competition between parties can be conceptualized as a one-dimensional space in CEE countries and the endpoints of that space can be understood in terms of ‘left’ and ‘right’. However, the policy dimensions that underpin the conceptions of left and right are not identical across countries. In the Czech Republic and Slovakia, left–right competition is bound up with positions on economic policy, whereas in Hungary, Poland and Slovenia notions of left and right are more closely associated with the social and religious outlook of parties. Therefore, if we operationalize the ideological space uniformly across CEE, as, for example, an economic divide, then it is almost certain that the true nature of left–right competition in Hungary, Poland and Slovenia will be incorrectly specified and would fail to be captured accurately.
The BL data contain a left–right variable that is based on a question respondents answered in their survey. In the LS data, this left–right variable needs to be derived from policy scales. In this analysis, I define the left–right ideological space as that which is most relevant to each country. This is done in a two-step process in the LS data. First, the policy dimensions from the LS expert survey that are most closely associated with notions of left and right in the BL survey are identified. In the Czech Republic and Slovakia, party policy positions on tax and public ownership are almost perfectly correlated with expert placements of party positions on the left–right dimension. In Hungary, Poland and Slovenia party positions on social policy and the role of religion in society are most closely associated with left–right placements. 10 Second, these four policy dimensions are reduced to two dimensions using principal components analysis. This gives us a left–right dimension that is tied to positions on economic policy which is used to array parties in the Czech Republic and Slovakia and a left–right dimension that is more closely linked to social policy that is used to position parties in Hungary, Poland and Slovenia.
The final left–right measure can be compared to the same measure derived from two other expert surveys of party policy positions – Benoit and Laver (2006) and the 2002 Chapel-Hill survey (Hooghe et al., 2010). The left–right measure from the LS data correlates extremely highly with both of these surveys; the correlation with the left–right dimension from the Benoit and Laver survey is r = 0.86, p < 0.01; with the Chapel-Hill survey it is r = 0.82, p < 0.01.
The independent variables that are used to assess the effect of ideology in CEE are drawn from these data and the BL left–right measure. The three independent variables that are tested in this article are ideological distance from the formateur, ideological distance from the median party and whether or not a party has the median legislator.
Control variables
The literature has demonstrated that numerous other factors are also influential in deciding which parties make it into government. Party size is often hypothesized to influence coalition membership (Matilla and Raunio, 2004; Tavits, 2008a; Warwick, 1996). Many models assess whether or not the largest party becomes a member of the coalition. However, in 25 of 31 cases of government formation in the LS dataset, the largest party was also the formateur; therefore this is not appropriate for the present research. Instead, a control for party size based on the number of seats each party holds is included, since very small parties are unlikely to get into government barring exceptional circumstances.
Controls are also included for previous experience of government, specifically whether or not a party was a member of a previous government that terminated before the end of its constitutionally mandated tenure. There are two possible interpretations of this variable: first, a party that was a member of a government that previously failed to complete its tenure may be punished by exclusion from future coalitions. Second, and conversely, a party in similar circumstances may be more likely to become a member of government if that government is a slight reconfiguration of a previous incarnation. This occurs when, for example, a party leaves a minimal winning coalition but no alternative government can gain the approval of the legislature so the remaining coalition partners continue in a minority government.
The latter occurrence also suggests that a control is needed for parties that enter governments that form mid-term rather than after an election, since mid-term government formations are more likely to be reconstitutions of the previous government rather than entirely new configurations.
The bargaining environment within parliament also needs to be considered, as this constrains the number of potential partners the formateur can choose from. With fewer parties in the legislature, the likelihood of any one being included in the government increases (Tavits, 2008a: 499). This is operationalized as the Effective Number of Parties (ENP) in parliament (Laakso and Taagepera, 1979).
Finally, controls for certain institutional arrangements are also entered into the model. Whether or not a government must undergo a vote of investiture is an indicator of the ease of access to government. Governments that must pass a vote of investiture require a greater level of explicit support than those that do not. Where no vote of investiture is required, governments must only be tolerated (Glasgow et al., 2011: 939). A dichotomous variable is also included for those countries which operate a constructive vote of no confidence. In such circumstances access to government during a parliamentary term is potentially more restricted, particularly for parties that were not part of the original government that took office during that legislative term.
Results of the analysis
Four separate models of party participation in government have been estimated. 11 The first considers ideological proximity to the formateur, while the second focuses on the median party hypotheses. The first and second models exclude formateur parties from the equation; including formateur parties in these models would artificially inflate the coefficients for ideological proximity. The third and fourth models consider the median party hypotheses. These models contain no exclusions except where data are missing. 12
The formateur and ideology
We begin by considering the hypothesis relating to the ideological distance from the formateur party. Coalition theory suggests that parties seek to form ideologically compact coalitions in order to minimize policy difference, which, in turn, reduces the potential for inter-party disputes during the government’s tenure. This is a logical assertion, as even parties that are numerically strong in the parliament still need to possess a reliable majority in order to pass legislation, which is best achieved through negotiation and cooperation rather than attempting to dominate and coerce coalition partners. Model one shows the results for the LS dataset, which covers the period 1990–2004, and model two the results using the BL data which cover more countries over a narrower timescale.
The results of models one and two, displayed in Table 4, provide strong support for the first hypothesis. The negative coefficient for the distance from the formateur variable in both models shows us that parties in CEE that minimize the ideological gap between themselves and the formateur are more likely to become partners in the government. These results are highly significant (p < 0.01) and, furthermore, distance from the formateur is one of only two variables in the model that obtained this level of significance. Parties in a previous government that terminated prematurely were also significantly more likely to get into government. One explanation for the significance of this variable is that when a government dissolves before the end of its constitutional mandate, one or more of the constituent members of that government re-form a new, possibly minority, administration to continue governing until the next scheduled election. There are numerous examples of such occurrences in CEE; for example, the 1997 Solidarity Electoral Action (AWS)/Freedom Union (UW) government in Poland which under the terms of our analysis dissolved in 2000 following the UW’s exit from the coalition. Despite not formally participating in the government, the UW agreed to support a minority AWS administration until the 2001 elections. This happens particularly in countries where the constructive vote of no confidence operates and alternative governments must be constructed before the incumbent is ousted.
Logit estimation of parties’ likelihood of getting into government.
***p < 0.01, **p < 0.05, *p < 0.10.
Dependent variable is whether or not a party got into government.
Somewhat surprisingly, the proportion of seats that a party holds is insignificant in models one and two, which suggests that party size is not important in determining who becomes a partner in a coalition in CEE. However, this is also likely to be partly a result of the way the model has been specified; with formateurs excluded the second largest party becomes the largest party in the model and, in most cases, is the principal opponent of the formateur.
Overall, models one and two provide strong support for the assertion that ideology is highly important in determining which parties get into government in CEE. In both the BL data, which cover governments formed over a narrow timescale in 10 countries, and the LS data, which consider governments formed over a long timescale in five countries, ideological distance from the formateur attains the highest level of significance. In this respect, ideological factors are one of the principal determinants of which parties get into government in CEE.
However, considering the role of ideology in this way implies a degree of risk for parties seeking to get into government. It is often uncertain which party will become the formateur of the next legislature, so parties that seek to strategically place themselves next to one or other potential formateur party risk alienating alternative candidates and thereby reduce their prospects for participating in government. For some parties this is not a problem, as they seek to form long-term alliances and prefer not to cooperate with alternative parties. On the other hand, some parties may be less attached to one particular governing partner. There may also be a risk for potential formateurs, particularly those located at a relatively extreme ideological position; if parties tend to coalesce with ideologically proximate partners and the largest party is isolated in the ideological space it may forfeit its opportunity to form the government.
The advantages of an ideologically moderate position have been hypothesized for decades (Downs, 1957) and it is likely that parties in the coalition-bargaining game will seek to ‘play the odds’ by adopting a position in the ideological space that offers the potential to cooperate with the widest range of partners possible – the median.
The median party and participation in government
The median party proposition is important when assessing party competition in a one-dimensional space, as is the case in this article. In such instances, the median ‘voter’ is said to be in a dominant position in the bargaining process (Laver and Schofield, 1990: 111; Warwick, 1996: 475). Furthermore, since parties seek to form ideologically compact coalitions to reduce the potential for intra-governmental disagreements, parties that are closer to the ideological median should be more likely to become part of any coalition that forms.
Models three and four of Table 4 show the results of this analysis; model three uses the LS expert survey data, while model four is based on the BL dataset. The median party in each parliament was coded as a dummy variable (1 = median party), but although the coefficient indicates that median parties were more likely to get into government in CEE, the effect of this was statistically insignificant in all models. There is therefore no support for hypothesis 2.
On the other hand, models three and four offer strong support for hypothesis 3. Parties that are closer to the median are significantly more likely to become part of the government. The only other variable in the model that obtains a similar level of significance is the proportion of seats held by a party. This is due primarily to the inclusion of formateurs in these models, which generally hold more seats than any other party. Similarly to models one and two, parties that previously participated in a government that terminated prematurely also entered models three and four significantly.
Parties that remain ideologically proximate to the median therefore have an advantage when seeking to become part of the government. This could indicate perfectly rational behaviour by parties which seek to minimize the distance between a range of potential coalition partners rather than just potential formateurs. It could also indicate that formateurs are more likely to be located at the median and parties are simply minimizing the distance to the formateur. However, the data show that in only 4 percent of cases was the formateur also the median party in the LS dataset. It is therefore apparent that although the median voter theorem does not hold true in the coalition-bargaining process in CEE, parties which are closer to the ideological median of the legislature have a greater chance of participating in government.
Figures 1 and 2 demonstrate the model-based probability of getting into government associated with both ideological distance from the median and distance from the formateur in CEE with both clearly showing the diminishing likelihood of participating in government as parties move further away from the median and formateur.
Effect of ideological distance from the formateur on probability of getting into government.
Effect of ideological distance from the median on probability of getting into government.
Model predictions in context
The results in Table 4 suggest that ideology is highly important in CEE in determining which parties get into government, but we can also show how this has impacted on individual parties. By comparing the predicted probabilities generated by the models in Table 4 with those generated by similar models with the ideological variables removed, we can develop a picture of the effect that ideology has had on coalition-participation. Table 5 shows the effect of this for Model 1, which evaluates distance from the formateur in the LS dataset. Overall, of those parties that got into government, in 71 percent of cases their predicted probability of getting into government increased once ideological distance from the formateur was added to the model compared to their predicted probability with the ideological variables excluded.
Change in predicted probability of participating in government.
Formateur parties not included in models that assess distance from the formateur.
> 0.09 increase
0.00 to 0.09 increase
0.00 to −0.09 decrease
< −0.09 decrease
On a country level, if we consider the 2002 Slovak coalition of SDKU, SMK, KDH and ANO each of the parties in that legislature had a 25 to 28 percent probability of joining the SDKU in government according to model-based estimates that exclude ideological variables. Based on the predicted probabilities derived from Model 1, which includes the distance from the formateur variable, the probability of the SMK, KDH and ANO entering government increased to 53, 42 and 39 percent, respectively.
Ideological factors can also add to our understanding of some of the more interesting cases of government participation in CEE. In 1994, Hungarian voters saw an oversized coalition of the MSzP and SzDSz take office. A large part of the reason for this surplus majority coalition was assuaging doubts about a communist successor party taking sole charge of the government so soon after democratization. However, the predictions of model three show that the SzDSz was the party with the second highest probability of participating in that government at 41 percent. This is a large increase on the 31 percent predicted probability of SzDSz participation based on a model with ideological variables excluded. The only other party which saw an increase in the probability of participation between the two models was the Christian Democrats (KDNP) which rose from 16 percent to 25 percent.
Of course, there are always cases that lie outside statistical models. One such case is the 1998–2002 single party minority government in the Czech Republic. On the basis of both models one and three, the Christian Democrats had an 85–87 percent probability of joining the government. The Christian Democrats were the median party in the legislature and ideological neighbour to the formateur, the Social Democrats. However, the Christian Democrats fought the 1998 election as an alliance with the Freedom Union which was an offshoot of the ODS. 13 The Freedom Union could not entertain forming a coalition with the largest party in the legislature, the Social Democrats, while the second largest party in parliament, the ODS, refused to bow to the demands of the Freedom Union (Roberts, 2003: 1277). 14 The deadlock was broken when the Social Democrats and ODS then agreed an extraordinary covenant which allowed the former to govern as a minority administration. Such occurrences are difficult to account for in statistical models, but these are relatively rare exceptions and, overall, the ideological variables considered in this article add considerably to our understanding of which parties get into government and which do not in CEE.
Conclusion
One of the puzzles of the current political science literature on Central and Eastern Europe is that studies have shown that ideological variables are irrelevant in determining which parties get into government in the region. For some this would stand to reason given some of the common preconceptions about unstable party systems and a lack of programmatic political parties in CEE. However, these notions are based on evidence from the early stages of democratization; recent research has shown that in fact party systems became more stable relatively quickly and that parties have competed on the basis of coherent and identifiable platforms for some time. Why, then, do ideological factors seemingly not matter when forming a government? Are parties in the region simply archetypal office-seekers?
I have argued in this article that the problem lies in the way ideological variables have been brought to data in most studies of government participation in CEE. Almost without exception, research in this field has used manifesto data from the CMP, in particular the derived left–right variable, despite the well-established limitations of these data for parties in CEE. In short, the CMP left–right variable fails to capture the true meaning of ‘left’ and ‘right’ in each country in the region. While it can be argued that party competition does take place in a one-dimensional ideological space in every country, the policy issues that are rolled into that space vary across countries and this is not accounted for in empirical studies that have used the CMP data for CEE.
To gain a more accurate assessment of party ideological positions and the effect that these have on government participation, this article used data from two expert surveys. The first, the Benoit and Laver survey, covers every country in the CEE region over a narrow period of time, while the second, the LS survey, covers the five Central European countries over the period 1990–2004. The left–right variable from each of these surveys uncovers the same one-dimensional ideological space across the region, meaning that the two surveys are therefore comparable.
The analysis of government participation using these data shows that contrary to expectations based on previous empirical research, ideological variables are highly significant in determining which parties get into government in CEE. Logit models that used both the BL dataset and the LS dataset produced almost identical results. First, parties that are ideologically closer to the formateur in CEE are more likely to get into government than those positioned further away. This indicates that parties in the region conform to traditional expectations of coalition formation by seeking to form governments that minimize the distance between governing parties and thereby reducing the potential for intra-governmental disagreements over policy. Second, ideological distance from the median also significantly affects a party’s prospects for participating in government. Those parties that are closer to the median are much more likely to become a member of the government than those located further away. This indicates that ideological moderation is an advantage for parties seeking to become part of the government.
These ideological variables were the most significant determinants of which parties get into government in CEE, along with previous experience of participation in a government that terminated prematurely. This demonstrates that when the left–right dimension is operationalized with due consideration to the context of CEE, ideology matters: it structures party competition for government right across the region and has also been important over time.
Footnotes
Acknowledgements
I thank participants at the Research in Progress Seminar in the Department of Politics at the University of Sussex and the anonymous reviewers for their helpful comments on an earlier version of this article.
Funding
This research and expert survey was funded by the Economic and Social Research Council grant number ES/J003425/1.
Notes
Appendix
Correlations between party policy positions and Benoit and Laver’s left–right dimension.
| Tax | Foreign policy | Public ownership | Social policy | Religion | Urban vs. rural | Centralization | Environment | |
|---|---|---|---|---|---|---|---|---|
| Czech Republic | 0.99 | −0.41 | 0.93 | 0.51 | 0.58 | 0.25 | 0.00 | 0.04 |
| Hungary | −0.26 | 0.86 | −0.91 | 0.92 | 0.93 | 0.89 | 0.85 | −0.91 |
| Poland | 0.47 | 0.13 | 0.35 | 0.54 | 0.78 | −0.35 | −0.11 | −0.72 |
| Slovakia | 0.98 | −0.85 | 0.88 | 0.36 | 0.59 | 0.22 | −0.57 | −0.85 |
| Slovenia | 0.78 | −0.37 | 0.65 | 0.86 | 0.90 | 0.82 | 0.83 | 0.66 |
Entries in the table are Pearson correlations with Benoit and Laver's (2006) left–right dimension for each country.