The relationship between party policy preferences and coalition formation – that is, whether parties that were close in these terms were more or less likely to form coalitions – has been extensively analysed by the Comparative Manifesto Project. In their seminal work that relied on 11 country studies, Laver and Budge (Reference Laver and Budge1992) proved that ideological connectedness is a strong predictor for coalition building. Many authors followed this path and party preferences have often been used in explanations of coalition and government formation, including the content of the coalition agreements (see, for example, Aarts et al. Reference Aarts, MacDonald and Rabonowitz1999; Bara Reference Bara2001; Bräuninger and Debus Reference Bräuninger and Debus2009; Budge Reference Budge1992; Budge and Laver Reference Budge and Laver1993; Keman Reference Keman2007; Kleinnijenhuis and Pennings Reference Kleinnijenhuis and Pennings2001; Quinn et al. Reference Quinn, Bara and Bartle2011; Rihoux et al. Reference Rihoux, Dumont and Dandoy2005; Tegenbos Reference Tegenbos1974; de Vries et al. Reference de Vries, Giannetti and Mansergh2001), government declarations (Hearl Reference Hearl1992) or throne speeches (Bara Reference Bara2005). In their case studies on the Netherlands and Belgium, Thomson (Reference Thomson1999) and Timmermans (Reference Timmermans2003) have shown that coalition agreements contain issues that are salient to the coalition parties and for which they have formulated policy pledges, even when they disagree on the solution to such issues. The coalition agreements do not focus only on non-divisive issues;Footnote 1 on the contrary, during the negotiations the pledges most salient to each party are included in the agreement to avoid disagreements along the way and to present an image of a decisive body that will last its full term.
This article challenges conventional wisdom since it argues that, rather than party policy or ideology explaining government participation, government participation explains party policy preferences. Party preferencesFootnote 2 – consisting of the relative amount of attention allocated by parties to specific policy issues – are known to be important. But the analysis of the factors that shape these policy preferences is rarely dealt with in the literature.Footnote 3 Most studies in political science consider party policy preferences as an independent variable. Party preferences are crucial in the explanation of, for example, policy outputs and election results. But what explains party preferences? In my research design, party policy preferences are used as a dependent variable. The explanation of party policy preferences is rather complex since, following Budge and Farlie (Reference Budge and Farlie1983: 129), ‘many factors shape a manifesto other than election considerations’. These potential explanations are numerous and very diverse, such as internal dissent, party organization, median voter position, party size, referendum, public opinion, real-world events, party system fractionalization, party competition, electoral results, popularity of party leaders, party activists, state organization and decentralization, and so on. Even if the party preferences and priorities are a complex issue for political parties, the role of political science is, in this regard, to identify the most important phenomena at stake. This study intends to contribute to the identification of the most important variables in party policy preferences by outlining the role of one specific phenomenon: government participation.
Starting from the basic assumption that party policy preferences are not alike over time (that is, they vary from election to election) or over space (that is, they vary between parties), hypotheses can be derived from the political science literature in order to explain these variations. The main goal of this article is to explain the preferences of political parties at the time of elections with the main assumption that their government participation and their desire to enter the next government influence these preferences. Indeed, government-participation and -formation processes are closely linked to policy issues. Voters and the media link parties in the government with specific policy issues while they link parties outside the government with other issues.
In this study, I will develop an analytical framework that allows us to explain party policy preferences in Belgium in the context of participation in government. I developed my study around the analysis of party manifestos for parliamentary elections, hypothesizing that the status of a party (whether it is in government or in the opposition) and its ability/willingness to enter the next government have an impact on how close its policy preferences are to those of the other parties. In other words, I expect to observe an ideological proximity for parties that governed together as well as for parties that would like to enter government together. The hypotheses that the executive (governing party or potential coalition partner) will influence party policy preferences may seem trivial, but this issue has seldom been discussed in political science and is crucially lacking empirical evidence.
This article is in three parts. In the first section, the literature on the impact of government-related factors on party policy preferences is analysed. Three main hypotheses will be derived from this review, focusing on three different aspects of the party–executive relationship: the party as a negotiator or a potential coalition partner, the party as an incumbent that is part of a coalition (with other governing parties) and the party as an incumbent that is distinct from opposition parties. The second section presents the data and methodological elements of my study of Belgian party manifestos, while the third part is devoted to the empirical test of the three main hypotheses. The conclusion will summarize the main findings of this study while outlining some methodological considerations about the use of party preferences in analysing coalition formation.
GOVERNMENT AND PARTY PREFERENCES
Government participation is a key variable in the explanation of party policy preferences, even if it can also be related to other political phenomena such as party size and electoral success. The variables of government formation and participation have often been used as dependent variables explained by the content of the manifestos – that is, parties that are close ideologically are more likely to form a coalition – including the understanding of the content of the coalition agreements, government declaration or throne speeches, but much more rarely as an independent (for an exception, see Daubler Reference Daubler2010) or control variable (for an exception, see Tavits Reference Tavits2008).
The party–government relationship and its effect on the content of manifestos have to be assessed at two distinct times:Footnote 4 the process of government formation and negotiation (that is, mainly after the elections and the drafting of the manifestos) and the process of incumbency and accountability for government record (that is, mainly before the elections and the drafting of the manifestos).
Government Formation
Ideological positions are important in order to explain government formation (see the abundant literature on this aspect of the government formation process), but this process in turn has an impact on party policy preferences. A manifesto consists of a sum of various intentions, emphases, promises, pledges and goals that will be activated once the party enters government (Bara Reference Bara2005: 290, 297). The content of manifestos is sometimes affected by expectations that a coalition will be formed (Hearl Reference Hearl1987a: 258), and ‘office opportunities affect party preferences’ (Kitschelt Reference Kitschelt1994: 127). Most of the parties are office seeking, and the fact that they perform well in elections is often not sufficient for them to enter a coalition. As a result, the parties have to adopt positions in their manifestos that render them ‘acceptable’ to the other potential partners.
This modification of the ideological profile can be done in different ways. First, the party may want to soften the content of its manifesto or ideologically moderate its policy stances when bargaining with other parties (Mattila et al. Reference Mattila and Raunio2007: 283). For Dittrich (Reference Dittrich1987: 228), the ‘need to build coalition governments pushes parties who actually want to enter government to adopt a rather moderate attitude in their manifesto’. This author adds later that parties show a tendency to move towards the centre of the ideological dimension.
Similarly, party manifestos tend to be rather vague because they seek to maintain their viability as coalition partners (Aarts et al. Reference Aarts, MacDonald and Rabonowitz1999: 64). Issue positions displayed in manifestos are therefore expected to remain vague. Even though some argued that policy pledges are expected to be less vague or ambiguous than other manifesto statements (Pétry and Landry Reference Pétry and Landry2001), analysts of pledges made by winning parties confirmed this (increasing) vagueness of manifestos (see, for example, Bara Reference Bara2005: 590). In order for the party to be accepted as a credible coalition partner, the policy preferences of a party should not only be moderate and/or vague but they should also be realistic (Heinisch Reference Heinisch2003: 101; and for the Belgian case, Anon. 1973: 1031). In other words, parties are expected to focus on issues and pledges that can be realistically achieved by the forthcoming cabinet.
The vagueness and moderation of a party manifesto may be related to the type of issue at stake. Two main types of issues can be distinguished: position and valence issues. Valence issues are often considered to be consensual issues – that is, issues on which there is agreement on the desired outcome across the whole electorate and one clear ideological position for all political parties (Katsanidou and Gemenis Reference Katsanidou and Gemenis2010). Unlike position issues, party preferences on valence issues may remain vague, moderate or even unrealistic. Subsequently, one might think that parties wanting to enter government would avoid emphasizing potentially polarizing position issues and would favour stressing valence issues.
However, party preferences are also a relative phenomenon when we are considering government formation: moderate and realistic preferences do not fully guarantee government participation since the preferences of the parties that wish to enter the government should not be too far from each other. The parties should not confront issues that would exclude them from participating in government. Parties have a clear incentive not to put on the agenda issues about which they disagree with potential coalition partners (van der Brug Reference van der Brug2001: 116).
In this regard, Hearl (Reference Hearl1987a: 268) hypothesized that parties that share a governmental vocation demonstrate a high degree of consensus. Since parties are able to choose their own policy preferences, they should tend to minimize the ideological distance between the different coalition partners (Kitschelt Reference Kitschelt1994: 125; Ray Reference Ray2007: 17). More precisely, Bale (Reference Bale2003: 75) argued that when centre-right parties want to enter a government together with the extreme right, they focus their manifestos on issues belonging to the extreme-right parties’ agenda: anti-immigration and welfare-chauvinist issues. Another example is to be found in the Dutch case in 1998. De Vries et al. (Reference de Vries, Giannetti and Mansergh2001: 207) observed that the green party presented a manifesto that was not distinct from other parties’ non-environmental issues as this party was seeking government participation and wanted to be considered as a potential coalition partner. Inversely, when the incumbent party anticipates a change of majority, it should focus on issues that will constrain the policy choices of the next policymaker – that is, the next party in government (Dellis Reference Dellis2009: 206). As a result, the government formation hypothesis is formulated as follows:
Hypothesis 1: Parties willing to enter government display similar policy preferences.
Incumbency and Government Participation
The second period of the party–government relationship deals with the effect of incumbency. Whether a party is in opposition or is part of the governmental coalition at the time of the manifesto release is a crucial factor when we look to explain the content of this document. Several phenomena are at stake here, relating to the government's record. Indeed, the government's record or particular actions and decisions made while in government are important for our understanding of the issues addressed during an electoral campaign and are a predominating concern in manifestos (Budge Reference Budge1992; Budge and Farlie Reference Budge and Farlie1983; Klingemann et al. Reference Klingemann, Hofferbert and Budge1994: 24; Netjes and Binnema Reference Netjes and Binnema2007: 43).
On one hand, government action covers a large range of different policies and policy domains. Yet, since one of the functions of a manifesto – and more largely of an election campaign – is to discuss and evaluate the government's outputs, Marks et al. (Reference Marks, Hooghe, Steenbergen and Bakker2007: 28) expect that the manifestos of parties in government have to develop policy positions on a wider spectrum of issues. Governing parties are often linked and identified with the policy issues discussed and decisions taken while they are in government and are expected to defend the government record – particularly when it concerns successful policy outcomes – on a variety of issues. As a result, one could expect that the positions of the governing parties would be less radical and less innovative as they are constrained by the government record.
On the other hand, manifestos from governing parties display an ‘inability to provide distinct programmatic positions in the government functions’ (Kritzinger and McElroy Reference Kritzinger and McElroy2010: 24). Indeed, if all governing parties are similarly bound by the government record – particularly as they all want to discuss successful policy outcomes – they might be expected to look alike. In other words, manifestos of incumbent parties should be similar (Louwerse Reference Louwerse2009: 17). Nonetheless, even if parties in government find it difficult to dissociate themselves from the actions of the government on certain policy stances, it does not mean that policy changes are excluded. Shifts of party position are possible but only in a gradual way (Budge Reference Budge1992: 13).
Even though in the case of France, Pétry and Pennings (Reference Pétry and Pennings2006: 105) observed changes in overall means for some policy issues following government participation, empirical evidence demonstrated that this hypothesis is not always verified. The hypothesis has been rejected by Alonso and Gomez (Reference Alonso and Gómez2010: 23), for example, concerning the decentralization issue. For the Spanish regional case they found that, even if the incumbent party has incentives to convey an image as defender of regional interest vis-à-vis the central government and the rest of the regions in order to stay in office, there is no difference between majority and opposition manifestos. Netjes and Binnema (Reference Netjes and Binnema2007: 44) believed that government participation contributes to explaining the salience of an issue in party manifestos since it might be expected to force governing parties to take clear-cut stances on that issue, but they found no empirical evidence to support that hypothesis. Ray (Reference Ray2007: 20) similarly found that cabinet participation had no impact on preferences expressed in party manifestos.
One can also identify an incumbency effect within the cabinet – that is, regarding ministerial portfolios. A party is influenced not only by the government record as a whole but also by the fact that its ministers hold specific policy portfolios. This party is therefore expected to deal in its manifesto with the policy domains covered by its ministers (Marks et al. Reference Marks, Hooghe, Steenbergen and Bakker2007). For example, in the Dutch case, Kleinnijenhuis and Pennings (Reference Kleinnijenhuis and Pennings2001: 173) observed that the fact that the Labour Party (Partij van de Arbeid – PvdA) held the Ministry of Environment explained its heavier emphasis on issues concerning the environment.
Not only is the content of the manifestos of the governing parties expected to resemble – in terms of policy preferences – the cabinet's agenda, but governing parties’ manifestos are therefore expected to look alike. Parties are closer to each other when they have previous experience of governing together (Budge Reference Budge1992: 18) and they should display more similar manifestos than any other pair of parties (whether in government or in opposition). This was the case for Norway, where the parties that were in coalition between 1965 and 1970 moved closer (Budge et al. Reference Budge, Crewe and Farlie1976: 16). Concerning the Belgian case, Hearl (Reference Hearl1987b: 240) empirically observed that government participation explains why some parties (namely, the Christian People's Party–Social Christian Party (Christelijke Volkspartij–Parti Social Chrétien – CVP-PSC) and the Socialist Party (Socialistische Partij–Parti Socialiste – SP-PS)) have similar concerns for issues such as technology, social justice, social services, education and minorities. Likewise, regarding positions on the EU, Bornschier (Reference Bornschier2010: 24) observed that governing parties – since the EU is the project of the governing parties, and these parties have participated in the making of the European polity – do not differ significantly in their posture regarding the integration process. As a result, my second hypothesis is formulated as follows:
Hypothesis 2: Incumbent parties display similar policy preferences
The other side of the incumbency phenomenon is related to the party's opposition status. According to the literature on party manifestos, the fact that the party is in the opposition has a significant effect on the content of its manifesto. Opposition parties have a structural advantage over government parties as they benefit from having more liberty to address the issues they prefer (Dalton Reference Dalton1985: 290; Green-Pedersen and Mortensen Reference Green-Pedersen and Mortensen2010). Similarly, Marks et al. (Reference Marks, Hooghe, Steenbergen and Bakker2007: 28) hypothesized that the positions of opposition parties tend to be more focused and more innovative than those of governing parties, and Klingemann et al. (Reference Klingemann, Hofferbert and Budge1994: 28) stated that ‘opposition parties have strong incentives for innovative framing of alternatives to current policy’. Since governing parties have to develop policy preferences on a wider spectrum of issues, they have less room for strategic issue targeting. One can therefore argue that opposition parties may choose the topics they focus on in their manifesto more freely and in a more radical way than incumbents. Following the argument regarding the type of issue at stake, opposition parties might contemplate making space for potentially polarizing (position) issues rather than consensual valence issues. Indeed, Louwerse (Reference Louwerse2009: 17) found that governing parties are closer to each other than opposition ones.
More precisely, a party in opposition will emphasize specific issues as a way to put the government in difficulties. Opposition parties use their manifestos as a public platform in order to criticize the government record and policy outputs – while maybe also targeting specific parties – and to present alternative policy solutions. Similarly, Green-Pedersen and Mortensen (Reference Green-Pedersen and Mortensen2009: 23) believe that opposition parties can force governing parties to pay attention to the issues that favour the opposition. This appears to be particularly true in the case of the moral issues in Belgium (Varone et al. Reference Varone, Jacob and De Winter2005: 263). Nonetheless, analyses of the Spanish case taught us that government participation has no influence: compared to governing parties, a party in opposition in a region does not particularly use decentralization issues as a way of putting the government in a difficult position (Alonso and Gómez Reference Alonso and Gómez2009: 12). Kerr (Reference Kerr1987: 131) added that ‘issues may unite parties in common opposition to governmental policies’ – that is, opposition parties should share policy priorities while criticizing the government record. On these specific policy issues, opposition parties are expected to look alike.
Overall, these different aspects regarding the content of the manifestos of opposition parties (freedom of issues, radical issues, innovation, critique of the government record, and so on, as well as elements mentioned regarding incumbency effects) lead to the hypothesis according to which one should observe a clear distinction between the manifestos of the governing and opposition parties:
Hypothesis 3: Governing and opposition parties display different policy preferences.
DATA AND METHOD
Several indicators are often used in order to measure party policy preferences. The particularity of the estimations based on party manifestos is that they clearly determine the position of a party at each election. By definition, the obtained information comes from an official document that was approved by the party: it is not even unusual that the majority of the party members and affiliates have formally to approve the document before the election. Electoral programmes are therefore considered as valid indicators of the positions of parties regarding certain domains of public policy at a given point in time (Volkens Reference Volkens2001).
In the framework of a research project – the Comparative Agendas Project (CAP) – aimed at analysing the positions of the Belgian political parties, a systematic thematic coding of the content of the party manifestos was set. Inspired by the original work of Baumgartner and Jones (Reference Baumgartner and Jones1993), the Belgian version of the Comparative Agendas ProjectFootnote 5 uses a slightly adapted version of their topical codebook to identify and track issue attention within several policy agendas (of parties, parliament, government, media, and so on) from 1977 to 2008. The codebook contains over 250 topic codes, organized by main topic category. The three most important issues in the Belgian manifestos for that period are intergovernmental relations (that is, the relations between the different policy tiers of government), social affairs and fiscal policy.
Party manifestos were coded per quasi-sentence, a procedure used by the Manifestos Research Group's Comparative Manifesto Project handbook (Budge et al. Reference Budge, Klingemann, Volkens, Bara and Tanenbaum2001). This coding procedure consists of identifying all the issues put forward in a document. If several ideas are expressed within a larger sentence, they would all be coded. This procedure applies very strict rules with respect to identifying such core ideas or quasi-sentences and was carefully executed. In addition, coding procedures allows a multiple coding for the same quasi-sentence, and various dummy variables were also added, allowing us to identify pledges, titles, references to other parties and so forth. About 250,000 quasi-sentences were coded, with an average of 2,435 per manifesto.
Regarding the time frame, 1978 can be considered as the starting point of contemporary Belgian politics. Up to 1968, the three mainstream parties displayed united and bilingual structures but, in a period of 10 years (1968–78), the Belgian unitary party system became completely regionalized as each mainstream party split into two independent parties (one Flemish, one French-speaking). These events had an important impact on various aspects of the Belgian political arena, among which are party positions. In this research, I therefore decided to focus on the contemporary period, starting with the elections preceding the split of the last mainstream party in 1978 up to the 2007 federal elections – that is, the 10 legislative elections that occurred from 1977 to 2007.
In regard to case selection, I included all parties that were at least represented in three elections in the federal House of Representatives – 13 parties: two Christian democrats, two social democrats, two liberals, two greens, two extreme right and three regionalists. With the exception of the French-speaking Socialist Party, ecologist Ecolo and National Front (Front National – FN), all other parties changed their name within the observed period. In addition, several electoral alliances were formed, as well as a few party splits, but the structure of the party system remained quite stable over time. The content of the manifestos of about 9 to 11 parties for each election year has been coded. The database obtained therefore consists of the content of exactly 100 party manifestos on about 250 different policy issues and spread over 31 years.
In order to measure similarities between party manifestos, correlations appear to constitute the proper method of analysis.Footnote 6 To put it simply, when the correlation scores are high, it is assumed that the two documents are somewhat similar. Correlations are calculated based on issue emphases in different party manifestos, across parties and/or over time. Examples of the use of correlations in order to measure similarities between manifestos are to be found in several works (Gabel and Hix Reference Gabel and Hix2002; Janda et al. Reference Janda, Harmel, Edens and Goff1995; Netjes and Binnema Reference Netjes and Binnema2007; Ray Reference Ray2007). Similarly, Pétry and Landry (Reference Pétry and Landry2001: 133) assess interparty distances by correlating the shares of party programmes devoted to particular issue categories.
Yet, some aspects of the use of correlations have to be pinpointed. First, correlations have been made on the whole manifesto – that is, all policy categories taken together and excluding non-codable items. Each manifesto has been used as a unit of analysis in the calculation of the different correlation scores for each pair of manifestos. Secondly, if significance is not an issue when dealing with the whole period under scrutiny, disaggregating the data per type of party or per election year produces important drops in the number of observations. One has to remain cautious with the results obtained with such a small N. In addition, much has been discussed about the minimally acceptable correlation scores. I follow McDonald and Mendes (Reference McDonald and Mendes2001: 111), who considered correlations of 0.80 as acceptable. Correlations between 0.70 and 0.50 will be considered as moderate, while correlations below 0.50 will be considered as weak.
In order to test the hypothesis concerning parties that want to enter government (Hypothesis 1), I used the proxy of an indicator based on whether the party was associated with any of the negotiation talks before the actual formation of the government. Indeed, one of the main characteristics of the Belgian government formation process is its long duration. The formation of an executive is sometimes a long-term process, and it is not rare that different coalition formulas are tested – including different parties – before the actual government formation. For example, the formation of the Martens I cabinet in 1978 was preceded by 11 unsuccessful attempts gathering different parties. I believe that a party that accepts to participate in the negotiation talks is by definition willing to enter the government. This proxy for the parties that want to enter the cabinet is based on data from Dumont and De Winter (Reference Dumont and De Winter1999) and on my update for the governments formed after 1999.
Like Klingemann, Hofferbert and Budge (1994: 47) and Daubler (Reference Daubler2010: 20), I operationalized the two government-related variables (that is, concerning the two hypotheses of incumbency and government participation) using dummy variables. The variable related to the negotiation process concerns the parties that were included in the government formation talks, while the incumbency variable measures whether the party was part of the governmental coalition at the time of the manifesto release, independently from the number of portfolios held by each party and from its participation in the regional cabinets. The list of parties included in both variables is to be found in the Appendix.
RESULTS
Government Formation
Three different hypotheses have been derived from the literature (see above), each focusing on a different aspect of the government formation and participation. The very first hypothesis relies on the fact that parties that want to enter cabinet not only display realistic and moderate party preferences but also that these preferences should be close to their potential coalition partners'. In other words, it is expected that the manifestos of parties that want to enter government are similar. This hypothesis is difficult to verify since one can hardly identify which parties want to enter government and which ones wish to remain in the opposition. As indicated above, I used a proxy based on the parties that were formally included in the government formation negotiations.Footnote 7
The problem with the use of this proxy is that sometimes it leaves few parties outside the negotiation talks. For example, the negotiations for the formation of the Tindemans IV cabinet included all Belgian parliamentary parties, with the exception of the Walloon Rally (Rassemblement Wallon – RW) and the Belgian Communist Party (Parti Communiste Belge – PCB). As a result, I could not test this first hypothesis for all elections under scrutiny: elections of 1977, 1978 and 1991 are left aside. I first calculated the average correlation figures for each group of parties per party system (Flemish and French-speaking) and, in a second step, the mean for the whole country. The obtained results are show in Figure 1.
Average Similarity between Manifestos, per election year (1981–2007)
Overall, parties that are included in the negotiation talks are much more similar to each other than parties that do not participate in such meetings. On average, correlation figures reach 0.528 for the first group of parties while it drops to 0.362 for the latter. In other words, parties that want to enter government look more alike than the other parties, meaning that these parties share policy preferences before the start of the actual process of government formation. This is not without consequence for the studies of government formation based on ideological proximity between parties. Such analyses hypothesize that parties that are the most similar would enter the cabinet together. However, the results demonstrate that parties that negotiate are similar, independent of their entry into government. Proximity assessed in terms of policy preferences is thus essential for parties to be considered for participation in negotiation talks and not for entrance into the cabinet.
With the exception of the elections of 2007, a clear pattern is observed in our time period. The differences between the two types of parties decrease over time. In the 2007 elections, significant differences are not observed and the similarities between manifestos are shared by both types of parties. There are no longer clear party policy differences, mainly because of two convergent phenomena. First, parties that enter into negotiations look less alike than in the 1980s. Compared to previous elections, they have less in common and the correlation figures indicate rather low similarities between these documents. Second, the parties that were not included in such talks – partly because some of them were never accepted as partners, such as the extreme-right partiesFootnote 8 – look more alike over time. One might believe that they adapt and maybe soften their party manifestos in order to be accepted by other parties.
A characteristic of the Belgian electoral system is its schism along linguistic lines. Broadly speaking, Flemish parties do not address French-speaking voters and vice versa, even in the officially bilingual city of Brussels. Since the split of the traditional political parties into language-based sister parties (between 1968 and 1978), manifestos are no longer released in both languages and not even translated. Language is in this framework an important variable as it allows us to distinguish Flemish from French-speaking parties. I therefore test whether the behaviour of the Flemish parties regarding the government formation is different from the French-speaking ones within the two distinct party systems.
Disaggregating the data per party system provides more or less the same results. For both Flemish and French-speaking party systems, average correlation figures are very similar, confirming the larger proximity of parties included in government formation talks compared to the other parties. If the same trends – that is, an increase in the proximity of parties that did not participate in the negotiations – are observed, there are some small differences. In Flanders the elections of 1999 and 2003 witnessed a closer proximity of non-negotiating parties, compared to that of the parties included in the talks. As regards French-speaking parties, the differences between the two types of parties remain to the advantage of the parties included in the formation negotiations, even if these differences clearly decrease over time.
Government Participation
The second hypothesis deals with the impact of government participation on party policy preferences. Based on previous works and empirical findings, it is expected that the content of the manifestos of incumbent parties would be similar. To verify this hypothesis, I calculated the correlation coefficient of each pair of parties that were in government at the time of the manifesto's release (see Table A2 in the Appendix). When the cabinet includes more than two parties from the same party system (as in the case of the elections of 1978 and 2003), I calculated the language-based average of each pair of parties.
Figure 2 shows that manifestos from parties that were in government at the time of the manifesto release are moderately similar. On average, both Flemish and French-speaking governing parties display correlation coefficients around 0.6 (with, respectively, average coefficients of 0.603 and 0.592). In addition, both party systems display similar figures of individual elections, with the exception of the elections of 1981, 1985 and 1999, and no general trends can be observed.
Average Similarity between Manifestos of Governing Parties, per election year (1977–2007)
The comparison of these figures with the ones obtained at the level of all the parties for one specific election – that is, including parties in the majority and in the opposition – delivers interesting and complementary information (see Table 1). Over the observed period, manifestos from governing parties look more alike than those from all parties taken together. Correlation coefficients for the latter are smaller by 0.109 and 0.124, respectively, among Flemish and French-speaking parties. At the election level, almost all coefficients concerning the content of the manifestos from governing parties indicate a closer proximity than figures for all parties for one specific election. Exceptions are to be found in 1999 and 2003 in Flanders and in 2003 for French-speaking parties. Even if these three exceptions do not show large differences (up to −0.068 in 1999), they indicate that similarities between governing parties are smaller than similarities observed between all parties. In these cases, the hypothesis of a high similarity of the manifestos from governing parties has to be rejected.
Average Correlations between Manifestos, per election year (1977–2007)
Similarly, some elections display a relatively high difference in the correlation coefficients when comparing governing parties with all parties, as in the case of the Flemish party system (+0.338 in 1985, +0.304 in 1981 and +0.293 in 1991) as well as, in the case of the French-speaking one (+0.280 in 1991). For these elections, governing parties look alike not only in absolute terms but also in relative terms since they are more similar than the other parties competing for the same election. Overall, these elements – average correlations taken in absolute and relative terms – indicate that the incumbent hypothesis may play a role for some specific elections but that similarities between governing parties remain relatively moderate. If incumbency demonstrates that it may explain some of the similarities between pairs of manifestos, this hypothesis does not cover all the observed similarities between manifestos.
The last hypothesis on government participation concerns the fact that the manifestos of governing and opposition parties are expected to be different. In order to test this hypothesis, I created three different indicators allowing us to measure the difference between governing and opposition parties. The first one, based on correlation coefficients between governing parties, has been presented above (Gov-Gov); the second one concerns correlation coefficients between opposition parties (Opp-Opp); while the third one concerns correlation coefficients between governing and opposition parties (Gov-Opp). Each indicator is based on pairs of individual manifestos and then averaged for each election and party system – that is, 137 pairs for the Flemish parties and 81 pairs for the French-speaking ones.
The results displayed in Figure 3 are rather intriguing for Flanders. Two different periods are to be distinguished. Between 1977 and 1995 – that is, in no fewer than seven elections – the three indicators remain distinct and allow the identification of different processes. The line representing the coefficient of correlations between governing parties (Gov-Gov) is always on top during that period. On average, pairs of Flemish governing parties tend to be more similar than any other pair of parties. Below Gov-Gov, the second line represents the correlation coefficients between governing and opposition parties (Gov-Opp). This line remains relatively stable throughout the whole observed period and indicates a fairly modest – if not weak – similarity between governing and opposition manifestos. The last line represents the correlation coefficients between pairs of opposition parties (Opp-Opp). Of the three indicators, this indicator is always at the bottom and demonstrates that opposition parties have almost nothing in common in their manifestos. During that period, similarities between opposition parties are rare and they share more party preferences with governing parties than with other opposition parties. As a result, the hypothesis according to which manifestos from governing and opposition parties are not alike somehow found confirmation, with the important observation that opposition parties are even less similar.
Average Similarity between Flemish Manifestos, per election year (1977–2007, N = 137)
The last three elections (1999–2007) present a completely different pattern. There is no longer a clear upper line and one can hardly distinguish between averaged pairs of manifestos. The correlations between governing parties (Gov-Gov) are surprisingly low, while correlations between opposition are rather strong (Opp-Opp). In 1999, the weak similarity between governing parties probably finds its explanation in the fact that the Christian democratic party Christian People's Party/Christian Democratic and Flemish (Christelijke Volkspartij/Christen-Democratisch en Vlaams – CVP/CD&V) moved into opposition for the first time, leading to a larger left–right ideological gap between the governing parties. Indeed, this so-called ‘rainbow’ cabinet gathered the green party but also the centre-left (socialist) and the centre-right (liberal) parties while excluding the central Christian democrats. Another part of the explanation relies on the fact that the green party (Agalev/Groen!) is no longer in the opposition, leading to a more homogeneous right-wing opposition (Christian People's Party/Christian Democratic and Flemish, Flemish Block/Flemish Interests (Vlaams Blok/Vlaams Belang – VB) and People's Union/New Flemish Alliance (Volksunie/Nieuw-Vlaamse Alliantie – VU/N-VA)). In 2003, manifestos from the opposition parties were more alike than the ones of the governing parties. This is likely to be the result of the Dutch-speaking versus French-speaking polarization of the country and the salience of these linguistic issues in the manifestos of the opposition parties. With the exception of the green party, all other opposition parties (Christian People's Party/Christian Democratic and Flemish, Flemish Block/Flemish Interests and People's Union/New Flemish Alliance) strongly emphasized the need for more autonomy (if not independence) for the Flemish region.
Note finally that the line representing the correlation coefficients between governing and opposition parties (Gov-Opp) remain relatively stable at around 0.5. The last hypothesis related to the government and opposition divide is only clearly confirmed in the case of the last elections of our period (2007) since manifestos from opposition and governing parties are less similar than any other pair of manifestos.
On the French-speaking side, the hypothesis according to which manifestos from governing and opposition parties are not alike found almost perfect confirmation, together with the fact that opposition parties are even less similar (see Figure 4). As expected, the line representing the coefficient of correlations between governing parties (Gov-Gov) is always above, with one exception (the elections of 2003). On average, pairs of French-speaking governing parties tend to be more similar than any other pair of parties. The second line represents the correlation coefficients between governing and opposition parties (Gov-Opp) and constantly increases from 1981 to 1999. Even if different coalition formulas have been implemented, governing and opposition parties are gradually becoming more alike and the indicator even exceeds the correlations between governing parties (Gov-Gov) for the 1999 elections. The last line represents the correlation coefficients between pairs of opposition parties (Opp-Opp) and is always situated at the bottom of the graph. French-speaking parties that sit together in the opposition have barely anything in common in their manifestos. Overall, manifestos from opposition and governing parties are not similar, confirming the hypothesis, but the similarities between opposition parties are even smaller.
Average Similarities between French-speaking Manifestos, per election year (1981–2007, N = 81)
Yet, the three indicators demonstrate an important convergence of policy preferences over time, with average correlations between 0.2 and 0.5 in 1981 to average correlations between 0.5 and 0.7 in 1999. Independent of their position – majority or opposition – French-speaking parties are becoming more alike. Unlike Flemish parties, the rejection of the central Christian democratic party Social Christian Party/Humanist Democratic Centre (Parti Social Chrétien/Centre Démocrate Humaniste – PSC/CDH) from government and into the opposition in 1999 did not lead to smaller similarities between the coalition partners that were no longer connected on the left–right dimension. The observed convergence stops abruptly in 1999 and the manifestos tend to diverge after that date. This trend needs to be confirmed by analysing the manifestos for the 2010 elections. The explanation of the poor performance of the indicator of the similarities between governing parties in 2003 (Gov-Gov) is probably to be found in the larger polarization of the two coalition partners – that is, the centre-left (socialist) and centre-right (liberal) parties – on socioeconomic issues.
DISCUSSION
This article aimed to analyse party policy preferences and hypothesized that government-related phenomena could have an impact on the similarities between electoral platforms of political parties. Indeed, the drafting of a party manifesto is a very complex exercise that requires an arbitrage between subtle equilibriums, depending on the party's objective, including for the post-election period.
If the party aims to maximize its electoral result by attracting the electorate, the voter is the target of its manifesto. If the party aims to enter government and to be part of the coalition, its platform is also targeted at the other parties, and more particularly at its potential partners in the future coalition. In these cases, the voter does not remain the primary target of a manifesto, and the parties are driven away from the concerns voiced by the public (see Bara Reference Bara2001, which linked party manifestos and the most important issue in opinion polls, and also Bartle et al. Reference Bartle, Dellepiane-Avellaneda and Stimson2011).
Parties that want to enter government should adopt positions in their manifestos that render them ‘acceptable’ by the other potential partners, by ideologically moderating policy stances in their manifestos and by minimizing the ideological distance with the different coalition partners. Yet, the parties which enter coalitions may not be those with the closest ideological proximity. In countries with fragmented party systems, parties would consider entering a coalition based on other considerations, such as previous governing experiences with these parties, political culture or simple parliamentary arithmetic (see the numerous examples in Laver and Budge Reference Laver and Budge1992).
Analysing the content of party manifestos in Belgium during three decades, I demonstrated the closer proximity of parties included in government formation talks compared to the other parties. I also observed that the differences between the two types of parties decrease over time and that, in more recent elections, parties tended to display similar preferences, independent of their ability/willingness to be included in the coalition formation talks.
The empirical verification of the first hypothesis leads to an important theoretical consideration. The literature on government formation often uses policy preferences as an independent variable that helps to explain the coalition formation process. Proximity models hypothesize that parties that have similar policy preferences are likely to form a coalition. However, I demonstrated that government participation prospects have an impact on the content of these policy preferences. If parties strategically adapt their electoral platforms in order to enter government, any study aiming at explaining coalition formation and/or government agreements based on party preferences is biased. To put it simply, studies on coalition formation analyse the capacity of preference adaptation of a party to other potential partners rather than its ‘real’ policy preferences. Not only does the party soften its discourse and moderate its policy stances, but it also artificially renders its manifesto ‘acceptable’ by adopting the other parties’ favoured policies. Coalitions are obviously made up of parties with similar policy preferences, not because they ‘are’ alike but because parties strategically ‘make’ them alike.
As a result, one should be more cautious when using party preferences in order to explain the government formation process, and more particularly the hypothesis according to which the parties that are located close to each other would form a coalition. Even if I do not argue that ‘electoral programs present an initial obstacle rather than an aid to the formation of the government’ (Klingemann et al. Reference Klingemann, Hofferbert and Budge1994: 33), I still believe that the use of party preferences – and more particularly party manifestos – in the understanding of government formation is particularly inadequate. Instead, future models should consider party policy preferences as the dependent variable rather than the independent variable.
Because governing parties experienced the fact of governing together and went through the same major political events and because they are similarly bound by the government record, I expected that such parties would display more similar policy preferences than any other pair of parties (whether in government or in opposition). In my analyses of the Belgian case, I observed that incumbency plays a role for most elections but that similarities between governing parties remain relatively modest and exceptions are to be found. Yet, governing parties tend to be more similar to each other than any other pair of parties. Similarities between opposition parties are rare and they share more party preferences with governing parties than with other opposition parties. As a result, the hypothesis according to which manifestos from governing and opposition parties are not alike somehow found confirmation, with the important observation that opposition parties are even less similar.
Even if my analyses proved that the distinction between three different pairs of parties (the so-called Gov-Gov, Opp-Opp and Gov-Opp measurements) is more relevant than the mere distinction between governing and opposition parties, some issues remain unsolved. In particular, the relevance of my findings is questioned since 1999 and 2003 constitute peculiar election years. In those elections, the policy preferences of the governing parties and of the parties involved in coalition talks were not very different from other parties’ preferences. One explanation may reside in the fact that those elections witnessed the emergence of the first cabinets since 1958 to send the centrist Christian democratic party family into the opposition and instead mainly gathered parties from the centre left and centre right. The consensus around the creation of a government without this party family was rather strong among the other parties, leading to a greater similarity in terms of policy preferences among those parties, independent of their governmental status (in government or in opposition) or their linguistic status (Flemish or French-speaking).
APPENDIX
Parties Included in Government Formation Process (1977–2007)
Parties in Government (at the expected time of manifesto release, 1977–2007)
