Dimensions of disagreement in EU affairs: is parliamentary opposition driven by left-right or European integration contestation?

Abstract

This study provides new insights into how the EU dimension relates to the left-right dimension in national politics. It does so by examining to what extent parties’ oppositional behaviour in EU affairs in parliament is a function of their distance from the government, other parties, the voters and party supporters on the EU dimension and the left-right dimension, respectively. The results suggest that oppositional behaviour is determined by two parallel dynamics: when expressing opposition, political parties are receptive both to the positions of the government and other parties on the EU dimension, and to the positions of voters and supporters on the left-right dimension. The results of the study thus indicate that the politicisation of EU affairs in national parliaments primarily is an elite phenomenon. In their relation to the voters and supporters, parties still principally navigate along the traditional left-right dimension.

Keywords:

• Oppositional behaviour
• political contestation
• party competition
• European Union affairs
• national parliaments

An extensive literature suggests that the issue of European integration has become of increasing importance for understanding domestic party competition in contemporary Europe (De Vries 2007; De Vries and Hobolt 2020; Hobolt and Rodon 2020; Van der Wardt et al. 2014). Yet, the empirical evidence regarding how the European Union (EU) dimension relates to the traditional left-right dimension in national politics remains mixed. In one view, political parties still base their behaviour in the domestic arena on conflicts along the well-known left-right dimension, leaving voter views on the potentially ‘giant’ issue of European integration largely unrepresented (Green-Pedersen 2012; Hoeglinger 2016; Mair 2007; van der Eijk and Franklin 2004). A contrasting view holds that the question of European integration is strongly politicised in the domestic arena and rivals or even surpasses the left-right dimension in importance (Hooghe and Marks 2009; Hutter and Grande 2014).

The aim of the current study is to contribute new empirical insights to this debate by examining oppositional behaviour in national parliaments. We focus on how political parties channel opposition in EU affairs because opposition is of great importance for the democratic quality of the EU polity as a whole. Political opposition, i.e. the voicing of disagreement with the government (Dahl 1966), is a key component of democracy because it forces the government to publicly defend its policies, but also because it provides voters with alternatives to the sitting government’s policies (Demirkaya 2019). In the absence of an active opposition that affords voters with choice, the right to political participation loses much of its relevance (Dahl 1971). Our investigation focuses on national parliaments where the authority over decisions regarding the future direction of European integration is centred. As pointed out by Peter Mair (2007), it is the domestic arena which holds ‘exclusive authority over constitutional questions’ in the EU. We draw on classical work on dimensions of contestation in the European political space (Hooghe et al. 2002; Marks and Steenbergen 2004), use spatial modelling and exploit a large hand-coded dataset on European Affairs Committee (EAC) deliberations in six countries (Austria, Denmark, Germany, Ireland, Sweden and the United Kingdom), to investigate the dimensions of domestic disagreement in EU affairs.

Given that political contestation in contemporary Europe no longer is organised along a single left-right dimension (Hobolt and Rodon 2020; Jackson and Jolly 2021), we must be open to the possibility that parties’ oppositional behaviour in EU affairs is affected by contestation along several dimensions. Thus, we develop two competing models of oppositional behaviour. According to ‘the EU contestation model’, opposition in EU affairs is seen as a function of the distance between a party and other relevant actors on the EU dimension. According to ‘the left-right contestation model’, opposition in EU affairs is instead assumed to be a function of the distance between a party and other relevant actors on the left-right dimension. Departing from these models, this study asks: To what extent is parliamentary opposition in EU affairs driven by contestation over European integration and/or conflict along the left-right dimension?

Following a theoretical understanding of parties as rational actors who engage in strategic behaviour to further certain goals (De Vries and Hobolt 2020; Downs 1957), we argue that the decision to express opposition should be seen as a deliberate choice made by party actors to promote vote-seeking, office-seeking and policy-seeking objectives (Müller and Strøm 1999). This further implies that it is not only the position of a party relative to that of the government that is of relevance as we seek to uncover the drivers of oppositional behaviour, but also that relative to other key actors in the political space. For example, a party that is far from the government’s position on a certain dimension of political contestation may strategically refrain from expressing opposition on a certain issue, if it knows that a majority of voters side with the government. In order to provide a comprehensive understanding of what drives oppositional behaviour in EU affairs, therefore, our examination takes into account the position of parties vis-a-vis four key actors: the government, other parties, the voters and party supporters.

Our findings suggest that oppositional behaviour in EU affairs is associated with two parallel dynamics. First, the greater the distance between a party and other elite actors (the government and other parties) on the EU dimension, the more likely it is to engage in oppositional behaviour in domestic EU affairs. Second, the greater the distance between a party and non-elite actors (the mean voter or party supporter) on the left-right dimension, the more likely it is to engage in oppositional behaviour. Taken together, the dimensions of disagreement we uncover in connection to parliamentary opposition in EU affairs confirm that the EU is politicised in the domestic arena (Hooghe and Marks 2009; Hutter and Grande 2014). However, the politicisation that we observe appears to be an elite phenomenon. As for their connection to non-elite actors, our results instead support previous findings (De Vries and Edwards 2009; Rovny and Edwards 2012), which suggest that parties still navigate along the traditional left-right dimension.

The article proceeds as follows: In the next section, we present the theoretical foundation of our study and derive our hypotheses. Thereafter, we discuss the research design, data and methods. Next the results of our analysis are presented, and the study concludes by highlighting the key results and implications of our findings.

Theory and hypotheses

The literature on parliamentary opposition has shed light on existing variation in the institutional ‘opportunity structure’ for opposition (Garritzmann 2017; Wegmann 2022), and has also identified other factors that impact opposition parties’ behaviour, such as ideology (Louwerse et al. 2017; Persson et al. 2023), previous government experience (Louwerse et al. 2021; Tuttnauer 2018), the legislative cycle (Schwalbach 2022) and the arena where deliberations take place (Karlsson et al. 2024). Both the literature on parliamentary opposition in general, and that on opposition in EU affairs have grown rapidly in recent years, advancing our knowledge on the subject (see Helms 2023 for a good overview).

Two previous studies are of special interest to us as they have sought to understand what drives oppositional behaviour in parliament, by drawing on spatial theory. Tuttnauer (2018) examines which factors make opposition parties more likely to confront or support the government by comparing voting patterns in 16 parliaments. He finds that being on the same ideological side (left or right) as the government makes opposition parties slightly more likely to vote against it. In an attempt to uncover what drives ‘government-opposition, voting’ Louwerse et al. find that the ‘government’s ideological composition and the status of the cabinet’ are key determinants of voting patterns in the parliament (2017: 756). Both these studies provide important insights for understanding what drives oppositional behaviour, but we believe our examination will move beyond these studies in two important ways.

First, both studies, like most prior work on oppositional behaviour in parliament, are concerned with voting behaviour. However, as the literature on opposition underscores, opposition is crucially about presenting alternatives to the positions taken by the government (Demirkaya 2019; Garritzmann 2017). This is why we turn to the content of parliamentary deliberations as a crucial data source for capturing oppositional behaviour. Like Peter Mair, we believe that ‘opposition demands political debate’ (2007: 7), and that research on oppositional behaviour needs to include parliamentary speech as well as voting behaviour. Second, we seek to provide an exhaustive account of what drives opposition in EU affairs, by examining the importance of both the EU dimension and the left-right dimension, as well as four spatial relations: the distance between a party, on the one hand, and the government, the mean party, the mean voter and the mean party supporter, on the other.

Spatial theory, dimensions of contestation and oppositional behaviour

According to Downs (1957) parties are vote maximisers and later research has confirmed that the mean voter position is a key concern for parties as they decide how to position themselves in the electoral market (De Vries and Hobolt 2020; Ezrow et al. 2011). Research on party behaviour has broadened the perspective by arguing that parties are ‘vote-seeking, office-seeking and policy-seeking’ (Müller and Strøm, 1999). Against this backdrop, we argue that the decision to express opposition or refrain from doing so should be seen as a deliberate choice by MPs with the intention to further their party’s interests. Expressing opposition is a way for parties to distance themselves from the government (Norton 2008: 238), making themselves visible to the electorate in an attempt to increase voter support. It may, however, also further office-seeking goals to the extent that criticising the government can weaken its position, or promote policy-seeking goals by presenting alternatives to the government’s policies that may gain widespread support in parliament.

The EU contestation model and the left-right contestation model introduced above rest on a common ground in terms of four basic assumptions: 1. Parties have vote-seeking, office-seeking and policy-seeking objectives. 2. Expressing opposition is a strategic choice that has consequences for a party’s chances of meeting its objectives. 3. The predicted consequences of expressing opposition depend on the positions of other relevant actors in the political space. 4. Parties have adequate information regarding the positions of other relevant actors. The models diverge when it comes to which underlying issue dimension they see as fundamental for a party’s choice to express opposition in EU affairs. The models are not contradictory, but they do represent competing views on which issue dimension is more important for understanding oppositional behaviour. According to the first model, it is the position of a party in relation to other actors on the EU dimension that determines whether or not it will engage in oppositional behaviour, whereas the second model sees the position of a party in relation to other actors on the left–right dimension as the key factor for the choice to express opposition.

Previous research has shown that the EU issue has become an increasingly important dimension of political contestation (De Vries and Hobolt 2020; Van der Wardt et al. 2014). Mainstream parties, however, have sought to prevent the politicisation of EU affairs by continuing to compete for votes mainly on the left-right dimension (De Vries 2007). However, this strategy has not worked as political entrepreneurs, ‘challenger parties’ (De Vries and Hobolt 2020), have successfully acted to increase the salience of European integration. By now we have a space of political contestation where both the European and the left-right dimension are important for understanding voter and party behaviour. Vote choices are thus likely to be affected by both the EU issue and the left-right dimension, as shown in a recent study of the 2017 general election in the UK (Hobolt and Rodon 2020). The fact that political contestation in Europe no longer is centred around one dominant dimension is something political parties need to take into consideration as they are confronted with key strategic choices (Dalton 2018). As argued by Bakker, Jolly and Polk, multidimensionality ‘fundamentally changes the dynamics and strategies of politics’ (2012: 220). Focusing specifically on the strategic choice to express opposition, we would naturally assume that this will be a function of the position of a party in relation to that of the government. To the extent that a party’s position is close to the government’s position it will have limited reasons to engage in oppositional behaviour. If, however, a party’s position is far from the government’s position its incentives for expressing opposition will increase. Two mechanisms may come into play here. First, by expressing opposition the party may hope to gather sufficient support to force the government to move its position closer to the party’s preferred outcome, thus furthering its policy-seeking goals. Second, expressing opposition may be a way to distance itself from the government, present voters with alternatives and compete on the electoral market, thereby furthering its vote-seeking and office-seeking goals.

Now, the fact that party competition takes place in a political space where the EU dimension is only partly aligned with the left-right dimension will have consequences for the strategic choice to express opposition. As stated above, we would consider a party that is far from the government on the EU dimension to be a likely candidate for expressing opposition when debating EU affairs in parliament. However, if we learn that this party happens to be close to the government on the left-right dimension, it may no longer be so obvious what to expect in terms of oppositional behaviour. If this party has a close cooperation with the government on economic policy it may very well refrain from expressing opposition when debating EU affairs in order to not put strain on its relationship with the government. In short, we need to be open to the possibility that oppositional behaviour in EU affairs is affected by a party’s position on the EU as well as the left-right dimension.

Previous work has demonstrated that parties are prone to revise their positions on European integration in light of changes in the behaviour of competing parties (Williams and Ishiyama 2018), or shifts in the preferences of voters and supporters (Dalton 2018; Ezrow et al. 2011; Meijers 2017). Building on this literature we argue it is reasonable to assume that oppositional behaviour is likely to be influenced by strategic considerations regarding a party’s position in relation to, not only the government, but three additional actors in the political space: other parties, voters and party supporters. Consider for example a situation where two parties are positioned equally far from the government on the EU dimension, but where the supporters of party A are positioned much closer to the government’s position than the supporters of party B. In this situation party A may be less prone to engage in oppositional behaviour than party B in order to avoid alienating their supporters.

In conclusion, our theoretical argument posits that the likelihood of a party engaging in oppositional behaviour in EU affairs is a function of its position on the EU dimension and/or the left-right dimension, relative to the position of the government, other parties, the voters and its supporters.

Hypotheses

The central proposition underlying our first set of hypotheses is that parties will decide to engage in oppositional behaviour based on their proximity to the government’s position. It stands to reason that the greater the distance between their preferences, the more disagreement we would expect to find between actors (Tuttnauer 2018). Expressing opposition can be a way for parties to force minority governments into concessions by presenting alternatives that other opposition parties may support, thereby promoting its policy-seeking objectives. Expressing opposition may also be a way for parties to portray themselves as a viable option to the sitting government in the eyes of voters (Demirkaya 2019; Norton 2008).

When it comes to European integration, we typically find that governments are pro-European. According to the EU contestation model, we would expect that the more Eurosceptic a party is, the more likely that it will engage in oppositional behaviour. This is indeed a proposition supported by case studies confirming that Eurosceptic parties are prominent drivers of oppositional behaviour in EU affairs (Karlsson and Persson 2022). More generally, the EU contestation model predicts that parties whose position on the EU dimension deviates greatly from that held by the government – in either direction – are more likely to express opposition than those positioned closer to the government.

The competing view would suggest that it is the distance between a party and the government on the left-right dimension that matters most and point to studies showing that left-right voting remains more important than EU issue voting (Carrieri 2020; Hobolt et al. 2009). The fact that Eurosceptic parties are the key drivers of opposition in EU affairs has to do with the fact that these parties are positioned to the extreme left or extreme right on the left–right dimension rather than the distance between their position and that of the government on the EU dimension. Thus, the left-right contestation model would predict that parties whose positions deviate greatly from the government’s position on the left-right scale – in either direction – would be more likely to engage in oppositional behaviour than parties positioned closer to the government. It is in the job description, so to speak, of left-wing parties to voice disagreement with right-wing governments and vice versa, no matter what the issue debated happens to be.

We can now formulate our first set of theoretical propositions, which we label the government distance hypotheses:

H1a: Parties far from the government’s position on the EU dimension are more likely to express opposition in EU affairs than parties that are close to the government’s position, all else being equal.

H1b: Parties far from the government’s position on the left–right dimension are more likely to express opposition in EU affairs than parties close to the government’s position, all else being equal.

Our second set of hypotheses is founded on theoretical and empirical insights regarding how parties compete to mobilise voter support. The basic premise is simply that as they compete for attention and ultimately votes, parties are searching for issues to run on that will make them stand out from the crowd. Classical work on issue evolution (e.g. Carmines and Stimson, 1986) have explained how new issues arise on the political agenda in a two-party system. More recent work by Hobolt and De Vries (2015) introduces the concept of ‘issue entrepreneurship’ to show how parties that are losers on the dominant dimension of contestation have used European integration as an issue to increase their electoral support.

The work on issue evolution (Carmines and Stimson, 1986) and issue entrepreneurship (Hobolt and De Vries 2015) highlight that parties in order to attract attention will have an interest in competing on pre-existing issue dimensions where they occupy policy positions that deviate from the mean party position, or introduce a new dimension where they take a polarising position (c.f. Riker 1996). Against this backdrop we assume that parties that hold positions that deviate from the mean party position on a certain issue dimension have incentives to engage in oppositional behaviour to separate themselves from the crowd in order to attract new voters. The EU contestation model suggests that what matters here is the position of a party in relation to the mean party position on the EU dimension. Parties that deviate from the mean party position by being markedly more EU-sceptical or pro-European will have an interest in drawing attention to this fact by expressing opposition. The left-right contestation model will instead expect the distance of a party to the mean party position on the left–right dimension to be the determining factor for variation in oppositional behaviour. As shown by previous scholarship (De Vries and Hobolt 2020; Hobolt and De Vries 2015) parties that are far from the mean party position on the dominant left-right dimension have incentives to compete on other issue dimensions such as European integration. Parties close to the mean party position on the left-right dimension may, on the other hand, have an interest to be seen as sufficiently moderate to be candidates for being included in future coalition governments (Ibenskas and Polk 2022), which may lead them to refrain from expressing disagreement with the government even if they are placed far from the government’s position on the EU dimension.

The reasoning above leads to our second set of theoretical propositions, which we label the mean party hypotheses:

H2a: Parties far from the mean party position on the EU dimension are more likely to express opposition in EU affairs than parties close to the mean party position, all else being equal.

H2b: Parties far from the mean party position on the left–right dimension are more likely to express opposition in EU affairs than parties close to the mean party position, all else being equal.

Previous research has demonstrated that mainstream parties respond to changing views among the public by moving closer to the mean voter position (Dassonneville 2018; Ezrow et al. 2011). Based on the ‘general electorate model’ (Ezrow et al. 2011) it seems reasonable to assume that the decision to engage in oppositional behaviour is linked to the mean voter position. Thus, parties and their representatives will factor in information about the mean voter position and how this is likely to affect the parties’ chances of furthering their vote-seeking, office-seeking, and policy-seeking goals. The EU contestation model would suggest that parties far from the mean voter position on the EU dimension will moderate their behaviour and express less opposition to avoid drawing attention to the fact that they are out of tune with voters. The left-right contestation model will instead cast the distance on the left-right scale between a party and the mean voter as the decisive factor in the choice to engage in oppositional behaviour. By drawing on insights from the literature on issue entrepreneurship (Hobolt and De Vries 2015) and wedge issues (Van de Wardt et al. 2014), we would expect parties that are losers on the dominant ideological dimension to highlight the EU issue and express opposition in order to receive increased attention and attract new voters. This brings us to our third set of theoretical expectations, which we label the mean voter hypotheses:

H3a: Parties far from the mean voter position on the EU dimension are less likely to express opposition in EU affairs than parties close to the mean voter position, all else being equal.

H3b: Parties far from the mean voter position on the left–right dimension are more likely to express opposition in EU affairs than parties close to the mean voter position, all else being equal.

Finally, prior research also suggests that parties are concerned with the mean position of their supporters when they consider how to position themselves on key dimensions of contestation (Ibenskas and Polk 2022; Schumacher et al. 2013). Ezrow et al. (2011) refer to this view of the relationship between parties and voters as the ‘partisan constituency model’. The idea here is that in their pursuit of vote-seeking and office-seeking objectives (Müller and Strøm 1999), parties are fundamentally concerned with preserving the backing of their current supporters rather than coming up with strategies to catch new voters. Some studies suggest that this commitment to the mean party-supporter position may indeed vary between parties. Whereas mainstream parties seem to be more guided by the views of the mean voter, niche parties like Green parties or Eurosceptic parties seem to be more responsive to the views of their supporters (Ezrow et al. 2011). Although the extent to which different types of parties are committed to aligning themselves with the mean supporter position may vary, we expect this to be a concern shared by all parties. Recent work has indeed confirmed that even mainstream parties are aware of and respond to incongruence between themselves and their supporters (Ibenskas and Polk 2022). One way of bridging such incongruence would be for parties to shift policy positions, but they may also act strategically to shift attention away from the fact that they hold preferences not aligned with those of their supporters. Building on the insights from previous scholarship we assume that parties that are out of touch with their supporters on the EU dimension have an interest in keeping a low profile on EU affairs by moderating their oppositional behaviour. Parties that are far from their mean supporters on the left-right dimension, on the other hand, may be incentivised to politicise cross-cutting issues such as European integration, where they may be more in sync with their supporters, by taking the opportunity to criticise the government. We thus arrive at our final set of theoretical propositions, which we label the mean supporter hypotheses:

H4a: Parties far from the mean supporter position on the EU dimension are less likely to express opposition than parties close to the mean supporter position, all else being equal.

H4b: Parties far from the mean supporter position on the left–right dimension are more likely to express opposition than parties close to the mean supporter position, all else being equal.

Research design, data and methods

In order to test our hypotheses, we use an original dataset on EAC deliberations in six parliaments. Whereas most committees in national parliaments only provide short summary protocols of their deliberations, the EACs provide detailed protocols that allow a thorough examination of committee deliberations, in most cases word by word. For the independent variables, we draw on high-quality data on party and voter positions on the EU issue, from the Chapel Hill Expert Survey (CHES) and the European Social Survey (ESS 2020), respectively.

Our choice to focus on EAC deliberations rather than plenary debates has been motivated by two reasons. First, although the importance and strengths of EACs vary between EU member states (Auel et al. 2015), this committee is still the most important arena for formalised interactions between the government and party representatives in EU affairs. Second, by focusing on the everyday grind of committee deliberations rather than high-profile plenary debates, we seek to capture oppositional behaviour representative of the daily business in parliament. A possible drawback with focusing on EAC deliberations rather than plenary debates is that the they are less visible to the public. This could potentially affect the incentives for MPs to engage in oppositional behaviour in the EAC for the purpose of communicating with voters. However, detailed records of EAC deliberations are made public and the media has good opportunities to cover what is being said, which means that MPs may still use these meetings to indirectly communicate with voters. There are also data to suggest that the media coverage of EAC meetings has been increasing over time (Persson et al. 2019).

The six countries in our sample have been chosen for both practical and theoretical reasons. First, we have selected EU member states that provide detailed records from EAC deliberations. Second, we have included cases to create variation when it comes to factors that may reasonably impact oppositional behaviour, such as public support for the EU, the parliaments’ institutional strength, and the presence of Eurosceptic parties in parliament.

Our dependent variable, Opposition, is derived from a dataset of nearly nine thousand human-coded statements from EAC deliberations in our six countries for the period 2005–2016. In four countries – Denmark, Germany, Sweden and the UK – we have drawn random samples of statements from EAC meetings in two legislative periods. In Austria and Ireland, the EACs hold fewer meetings, so we include all meetings in Ireland and in the first period in Austria (see Section C in the Online Appendix for more information on sample sizes, legislative periods and cabinets).

During EAC meetings, members of the committee make statements in response to presentations that lay out the government’s position. We code each individual statement made by MPs during the sampled EAC meetings according to a detailed scheme. Statements that deliver critique against the government’s position or advocate an alternative course of action are coded as opposition, whereas statements that express support for the government’s position or remain neutral, for example, by requesting more information, are coded as not containing opposition. Our dependent variable, Opposition, reflects the proportion of a party’s statements in a given year that contains opposition, consequently ranging from 0 to 1 (for a graph showing the proportion of opposition statements, by party and year in each of the six countries studied, see Figure A1 in the Online Appendix, and for additional information on the coding process, see Section D in the Online Appendix). Table A1 in the Online Appendix reports the descriptive statistics of all variables used in our analysis and shows that among the parties in our sample, across all years included, the mean share of opposition is 37%.

Table 1. Regression analysis of the effect of party distance to the government on the left–right dimension and EU dimension, respectively, on oppositional behaviour.

	(1)	(2)	(3)	(4)	(5)
Lagged opposition	0.50* **	0.42***	0.31***	0.28***	0.25**
	(0.10)	(0.10)	(0.10)	(0.10)	(0.10)
Government distance (LR)	0.04***(0.01)	0.02*** (0.01)	0.00(0.01)	−0.00 (0.01)	−0.01 (0.01)
Government distance (EU)		0.02*** (0.01)	0.02*** (0.01)	0.02*** (0.01)	0.03*** (0.01)
Seat share			0.00	0.00	0.00
			(0.00)	(0.00)	(0.00)
Government party			−0.15***	−0.16***	−0.18***
			(0.03)	(0.03)	(0.04)
EU saliency				0.01	0.00
				(0.01)	(0.01)
GAL–TAN				−0.00	−0.01
				(0.00)	(0.01)
Year				0.01**	0.01**
				(0.00)	(0.00)
Constant	0.11***	0.13***	0.27***	−17.85**	−14.63*
	(0.03)	(0.03)	(0.05)	(8.09)	(7.61)
Country FE	No	No	No	No	Yes
Adj. R2	0.48	0.52	0.56	0.58	0.59
Observations	250	250	250	250	250

Note: Panel-corrected standard errors in parentheses. *p < 0.10, **p < 0.05, ***p < 0.01.

We use data from the CHES survey (Jolly et al. 2022) to measure the distances between a party and other actors within the party system. For the time period under investigation, CHES expert estimates of party positions are available for 2004, 2010, 2014 and 2016. We calculated a yearly measure of party positions by using linear interpolation in STATA. This method is often used when time data points are missing in analyses based on time-series data (cf. Hobolt and de Vries 2015). Party positions on the left–right dimension range from 0 (extreme left) to 10 (extreme right). Party positions on the EU dimension reflect the overall orientation of a party’s leadership towards European integration and are recoded as ranging from 1 (strongly opposed) to 10 (strongly in favour). Admittedly, the average position of parties can mask a great deal of heterogeneity. Higher levels of intra-party divergence will naturally result in more imprecise estimates of a party’s position. This, however, is a problem of equal concern for all studies seeking to estimate parties’ positions on a given issue dimension, whether they rely on expert surveys or manifesto data (c.f. Ecker et al. 2022).

The Government distance variables measure a party’s distance to the government’s position on the left–right and EU dimension, respectively. The Mean party distance variables measure a party’s distance to the average party position in parliament on each dimension. To capture the average position of parties in the government and in parliament, respectively, we follow Gross and Sigelman (1984) by identifying the centre of political gravity (CPG) in each body. The CPG is calculated using the following general formula: $$\mathit{\text{CPG}}=\sum _{i=1}^1{T_{i}C}_{i\mathrm{ }}$$ where T_i is the seat share of party i in the body in question, and C_i refers to the position that party i takes on the left–right or EU dimension. The variables Government distance (EU) and Government distance (LR) measure the absolute distance between a given party and the government’s CPG on the EU scale and left–right scale, respectively. Correspondingly, the variables Mean party distance (EU) and Mean party distance (LR) reflect the absolute distance between a given party and the CPG in parliament, on each dimension.

Our remaining four independent variables reflect the distances between a party and the general voter, on the one hand, and the party’s supporters, on the other hand. These measures are calculated by combining expert evaluations of party positions from CHES with survey data on voter and supporter positions from the ESS, as is common practice in the research field (cf. De Vries 2007; Hobolt and De Vries 2015). Since the ESS is a biannual survey, we used linear interpolation to generate yearly data. In the ESS, the respondents’ left–right positions are measured on a self-placement scale ranging from 0 (left) to 10 (right), and the corresponding EU scale ranges from 0 (Unification has already gone too far) to 10 (Unification should go further). The variables Mean voter distance (LR) and Mean voter distance (EU) reflect a party’s absolute distance to the mean voter on the left–right scale and EU scale, respectively. Our mean voter distance variables are calculated as a party’s distance to the average position among all potential voters in the next election, representing individuals aged 15 and above in each country, aligning with the overall delimitation of the ESS sample. Finally, the variables Mean supporter distance (LR) and Mean supporter distance (EU) reflect a party’s absolute distance to its mean supporter on these two scales. We identify a party’s supporters using an ESS item asking the respondents if there is a particular political party they ‘feel closer to’. For each of the 42 parties in our sample, we calculate the mean supporter’s left–right and EU positions in each year. For small parties, the mean supporter positions are calculated based on relatively few ESS responses but never less than 30 per party and year (for pairwise correlations between all our eight distance variables, see Table A2 in the Online Appendix).

Table 2. Regression analysis of the effect of party distance to the legislative centre of political gravity on the left–right dimension and EU dimension, respectively, on oppositional behaviour.

	(1)	(2)	(3)	(4)	(5)
Lagged opposition	0.56***	0.45***	0.28***	0.25***	0.23**
	(0.09)	(0.10)	(0.10)	(0.10)	(0.10)
Mean party distance (LR)	0.04*** (0.01)	0.01 (0.01)	0.01 (0.01)	0.01 (0.01)	0.00 (0.01)
Mean party distance (EU)		0.03*** (0.01)	0.03*** (0.01)	0.03*** (0.01)	0.04*** (0.01)
Seat share			0.00	0.00	0.00
			(0.00)	(0.00)	(0.00)
Government party			−0.15***	−0.16***	−0.17***
			(0.03)	(0.03)	(0.03)
EU saliency				0.00	−0.00
				(0.01)	(0.01)
GAL–TAN				−0.01	−0.01
				(0.01)	(0.01)
Year				0.01**	0.01*
				(0.00)	(0.00)
Constant	0.08***	0.11***	0.25***	−17.52**	−14.21*
	(0.03)	(0.03)	(0.04)	(7.82)	(7.46)
Country FE	No	No	No	No	Yes
R2	0.46	0.51	0.58	0.59	0.61
Observations	250	250	250	250	250

Note: Panel-corrected standard errors in parentheses. *p < 0.10, **p < 0.05, ***p < 0.01.

Our dataset has a party–year structure and contains 294 observations. To test our hypotheses, we estimate several dynamic panel models using ordinary least squares regression (OLS). The oppositional behaviour of a party in any given year is likely to be strongly correlated with its oppositional behaviour the year before. To account for serial correlation, all models include the lagged dependent variable, Lagged opposition, and to account for panel-specific heteroscedasticity, panel-corrected standard errors are calculated (Beck and Katz 1995). Using party fixed effects to account for unit-specific heterogeneity would lead to biased coefficients (Nickell 1981: 1418) because our dataset only covers 12 years. Hence, party fixed effects are not included in our models. To control for linear time trends, we include the variable Year. Moreover, the final specification of each model includes Country fixed effects that absorb all cross-country differences in the average levels of opposition, to rule out the possibility that such differences can explain our findings. In order to ensure the robustness of our results, we conducted additional analyses using fractional logit instead of OLS and curvilinear distance measures as independent variables. The detailed results of these robustness checks are presented in the Online Appendix (see Tables B1–B4 for results based on curvilinear distance measures and Tables B5–B8 for fractional logit results).

A number of control variables are included in the models. First, we expect status as a government party to be linked to less oppositional behaviour and lower values on several distance variables. Thus, we include the dummy Government party, which is based on information from the ParlGov dataset (Döring and Manow 2020) and coded 1 if a party is in government for six or more months in a given year and 0 otherwise. Second, the ParlGov variable Seat share is included because large and small parties may employ different strategies, hence differing both in terms of their oppositional behaviour and their distance to other actors. Finally, two further CHES variables are added as controls because they tap into values that may influence a party’s oppositional behaviour as well as its distance to other actors on the EU and left–right scales. A party’s GAL–TAN position captures its views on social and cultural values on a scale from 0 (libertarian/postmaterialist) to 10 (traditional/authoritarian). Some would argue that GAL/TAN should be seen as a third dimension that structures the European political space (e.g. Bakker et al. 2012), alongside the left-right and the EU dimension. As European integration is one of the key issues defining GAL/TAN (Kriesi et al. 2008), we believe it makes little sense to add a third model of oppositional behaviour to our investigation. Instead, GAL/TAN is included as a control variable. The variable EU saliency, finally, measures the importance a party publicly attributes to the issue of European integration, on a scale from 0 (of no importance, never mentioned) to 10 (the most important issue).

Results

Our first hypothesis (H1a) suggests that parties that are further from the government’s position on the EU dimension are more likely to express opposition in EU affairs. A contrasting hypothesis (H1b) proposes that parties that are further from the government’s position on the left–right scale are more likely to engage in oppositional behaviour. In Table 1, we assess these hypotheses using linear regression with panel-corrected standard errors (Beck and Katz 1995). Columns 1–3 show that a party’s distance to the government on the left–right dimension has no statistically significant relationship with its oppositional behaviour after we control for its government distance on the EU dimension, government status and seat share.

In contrast, a party’s propensity to engage in oppositional behaviour clearly increases with growing distance from the government on the EU dimension, which is a pattern that lends support to the EU contestation model. Our control variables largely behave as expected: government parties are less prone to voice opposition than parties not part of the government. Oppositional behaviour does not covary with seat share. Finally, the coefficient of the lagged dependent variable (lagged opposition) is large and statistically significant, thus confirming that it should be included in the specification to avoid omitted variable bias.

In column 4, we add controls for the importance that a party ascribes to the EU issue (EU saliency) and its position on the GAL–TAN scale, as well as linear trends over time (Year), without observing any important changes in the coefficients. Our final model in column 5 includes country fixed effects to account for cross-country differences in the average levels of opposition. Here, each step further away from the government on the EU scale (0–10) is associated with a 3% point increase in oppositional behaviour (p < 0.01). This effect is arguably quite considerable.

Next, Table 2 evaluates the hypotheses about the importance of a party’s distance to the mean party in parliament, for its propensity to express opposition. H2a suggests that parties further from the mean party on the EU dimension are more likely to express opposition. In contrast to this, H2b posits that parties that are further from the mean party on the left–right scale are more likely to express opposition in EU affairs.

As shown in Table 2, the EU contestation model finds significant support as we evaluate these alternative hypotheses. A party’s oppositional behaviour intensifies with increasing distance to the mean party in parliament on the EU dimension. In a pattern similar to that in the previous analysis, columns 1–3 show that a party’s distance to the mean party on the left–right dimension lacks a statistically significant relationship with the proportion of opposition that it expresses, when we control for the corresponding distance on the EU dimension, government status and seat share. Column 4 adds three control variables – EU saliency, GAL–TAN position and Year – without altering the results in any substantial way. The final model in column 5 also applies country fixed effects, and the EU contestation model still finds firm support. With each step of increasing distance to the mean party on the EU scale, there is an associated 4% point increase in a party’s average propensity to engage in oppositional behaviour (p < 0.01).

In order to examine whether European integration or left-right related party–voter dynamics affect oppositional behaviour, we developed two further sets of hypotheses. In line with the EU contestation model, H3a suggests that parties that are further from the mean voter position on the EU dimension are less likely to express opposition in EU affairs. Hypothesis (H3b) instead holds that parties that are further from the mean voter position on the left–right scale are more likely to express opposition in EU affairs.

Table 3 contains the results that evaluate H3a and H3b together. So far, our analysis has shown that an increasing distance to the government and the mean party on the EU dimension is associated with more oppositional behaviour among our sample of parties, whereas no such relationship exists with increasing distances along the left–right scale. In contrast, the results in Table 3 indicate that a party’s distance to the mean voter on the EU dimension makes no difference for its oppositional behaviour. This pattern is consistent across all five models. Meanwhile, each step of increasing distance between a party and the mean voter on the left–right scale is linked to a 5% point increase in the share of voiced opposition (p < 0.01).

Table 3. Regression analysis of the effect of party distance to the mean voter on the left–right dimension and EU dimension, respectively, on oppositional behaviour.

	(1)	(2)	(3)	(4)	(5)
Lagged opposition	0.53***	0.53***	0.34***	0.31***	0.27***
	(0.09)	(0.09)	(0.10)	(0.10)	(0.10)
Mean voter distance (LR)	0.05*** (0.02)	0.06*** (0.02)	0.05*** (0.01)	0.05*** (0.01)	0.05*** (0.01)
Mean voter distance (EU)		−0.01 (0.01)	−0.00 (0.01)	−0.01 (0.01)	0.00 (0.01)
Seat share			−0.00	−0.00	−0.00
			(0.00)	(0.00)	(0.00)
Government party			−0.16***	−0.17***	−0.18***
			(0.03)	(0.03)	(0.03)
EU saliency				0.01	0.01
				(0.01)	(0.01)
GAL–TAN				−0.00	−0.00
				(0.00)	(0.00)
Year				0.01**	0.01**
				(0.00)	(0.00)
Constant	0.07***	0.09**	0.24***	−18.51**	−16.68**
	(0.02)	(0.04)	(0.05)	(8.19)	(8.11)
Country FE	No	No	No	No	Yes
R^2	0.47	0.47	0.55	0.57	0.59
Observations	250	250	250	250	250

Note: Panel-corrected standard errors in parentheses. *p < 0.10, **p < 0.05, ***p < 0.01.

This pattern suggests that parties employ different strategies in relation to the voters, on the one hand, and other elite actors, on the other. In line with the predictions of the EU contestation model, a greater distance from other elite actors (the government or other parties) on the EU dimension is correlated with more oppositional behaviour in EU affairs. As the left-right contestation model predicts, on the other hand, a greater distance to the mean voter on the left–right scale is linked to more intensified efforts to voice opposition in EU affairs.

We now turn to examine whether this pattern also holds for the dynamics between political parties and the specific group of voters that is the closest to them: party supporters. H4a states that parties further from the mean supporter position on the EU dimension are less likely to express opposition, whereas H4b suggests that parties further from the mean supporter position on the left–right scale are more likely to express opposition.

The results in Table 4 suggest that a party’s oppositional behaviour is unrelated to its distance from the mean party supporter on the EU dimension. This pattern appears in all five models, including that in column 5, which includes country fixed effects. At the same time, an increasing distance between a party and its mean supporter on the left–right scale is linked to increasing oppositional behaviour. On average, every extra step of distance between a party and the mean party supporter on our 0–10 left–right scale is linked to a 7% point increase in the share of opposition that it expresses (p < 0.01). This is arguably a rather substantial effect.

Table 4. Regression analysis of the effect of a party distance to its mean supporter on the left–right dimension and EU dimension, respectively, on oppositional behaviour.

	(1)	(2)	(3)	(4)	(5)
Lagged opposition	0.59***	0.59***	0.36***	0.33***	0.29***
	(0.09)	(0.09)	(0.10)	(0.10)	(0.10)
Mean supporter distance (LR)	0.05*** (0.02)	0.04** (0.02)	0.05*** (0.02)	0.06*** (0.02)	0.07*** (0.02)
Mean supporter distance (EU)		0.00 (0.01)	0.01 (0.01)	0.01 (0.01)	0.01 (0.01)
Seat share			−0.00	−0.00	−0.00
			(0.00)	(0.00)	(0.00)
Government party			−0.16***	−0.17***	−0.18***
			(0.03)	(0.03)	(0.03)
EU saliency				0.00	0.00
				(0.01)	(0.01)
GAL–TAN				−0.01	−0.01
				(0.01)	(0.01)
Year				0.01**	0.01**
				(0.00)	(0.00)
Constant	0.13***	0.11***	0.28***	−20.10**	−17.64**
	(0.03)	(0.03)	(0.05)	(8.24)	(8.06)
Country FE	No	No	No	No	Yes
R^2	0.44	0.43	0.53	0.55	0.56
Observations	243	242	242	242	242

Note: Panel-corrected standard errors in parentheses. *p < 0.10, **p < 0.05, ***p < 0.01.

Taken together, our results suggest that we need both the EU contestation model and the left-right contestation model in order to understand the oppositional behaviour of political parties. It appears that these models contribute different pieces to the puzzle. The EU contestation model offers a way to understand the link between parties’ oppositional behaviour and their strategic positions in relation to other elite actors: the government and other parties in parliament. To this, the left-right contestation model adds a way to understand how a party’s oppositional behaviour is simultaneously conditioned on its position in relation to the general voter and party supporters, which are two other factors that political parties must be constantly aware of.

Conclusions

Democracy is, as Ian Shapiro reminds us, ‘an ideology of opposition as much as it is one of government’ (1996: 51). The way oppositional behaviour reflects the dimensionality of the political space in contemporary Europe thus speaks to the state of our democracies. The results from our attempt to uncover how key dimensions of political contestation relate to each other when parties engage in oppositional behaviour in EU affairs highlight three important findings.

First, whereas previous studies have shown that the distance between an opposition party and the government (Tuttnauer 2018) matters for the extent to which it will oppose the government, we have shown that the distance to the government, to other parties, to voters and to supporters all impact a party’s oppositional behaviour in EU affairs. Understanding strategic party behaviour, like the choice to express opposition, requires that we include several key actors in the political space in our spatial models of party competition.

Second, our results show that identifying the drivers of oppositional behaviour in EU affairs cannot be reduced to a matter of the distance between a party and other actors on either the EU dimension or the left-right dimension. Instead, we find that the importance of both dimensions is partially corroborated by the data: two of the four hypotheses derived from each model are supported, whereas the other two are rejected. Thus, the results of our study indicate that oppositional behaviour in EU affairs should be understood in terms of both European integration and left-right contestation. Viewed through the lens of party competition, our results refute the idea that parties refrain from competing on the EU dimension (Mair 2007) and that there exists an ‘opposition deficit’ (Rauh and de Wilde 2018) in EU politics. Our findings rather support the view that parties do compete on European integration, and that they do so in a predictable manner: the greater the distance between a party’s position and the position of the government and other parties on the EU dimension, the more likely the party will engage in oppositional behaviour.

So where does this take us? If our main finding simply was that everything matters when we seek to understand the dimensionality of oppositional behaviour in EU affairs, then we would not have accomplished much. A third important result of the study, however, is that a distinct pattern emerges from the data in terms of how the two models come into play. What we find is that the hypotheses derived from the EU contestation model find support when we probe the distance between a party and the government, and a party and other parties in parliament. In contrast, the hypotheses derived from the left-right contestation model find support when we examine the distance between a party and the voters, and a party and its supporters. Thus, oppositional behaviour in EU affairs is associated with two parallel dynamics. On the one hand, parties compete with other elite actors on the EU dimension. The greater the distance between a party and other political actors on the EU dimension, the more likely it is to engage in oppositional behaviour. On the other hand, parties also take into account their position in relation to voters and supporters on the left-right dimension. The greater the distance between a party and the voters or its supporters on the left-right dimension, the more likely it is to engage in oppositional behaviour in EU affairs. A most likely candidate for expressing opposition in EU affairs would thus be a party that is far from the government and the other parties in parliament on European integration, while at the same time being far from the voters and its supporters on the left–right dimension.

We believe our study makes a significant contribution to existing scholarship by demonstrating how the EU dimension and the left-right dimension relate to each other when parties engage in oppositional behaviour in EU affairs. Our results confirm that EU affairs is indeed politicised in the domestic arena (Hooghe and Marks 2009; Hutter and Grande 2014), but the politicisation observed is an elite phenomenon. Regarding the connection between political parties and the public our findings instead align with previous studies (De Vries and Edwards 2009; Rovny and Edwards 2012), which suggest that parties primarily compete along the traditional left-right dimension. This way our findings also speak to the literature on mass-elite linkages (Ray 2003; Steenbergen et al. 2007) and party competition (De Vries and Hobolt 2020; Hobolt and De Vries 2015). We find no evidence to suggest that parties are receptive to the issue stands of non-elite actors (voters or party supporters) on European integration, when engaging in oppositional behaviour in EU affairs. Hence, the study lends no support to the presence of a bottom-up connection (Carrubba 2001) on the EU dimension when parties decide to express opposition in parliament. This means our results also say something important about the connection between issue dimensionality and the strategic action of parties to further their main objectives. What we find is that parties navigate along the left-right dimension when using oppositional behaviour to promote their vote-seeking and office-seeking goals. Moreover, the fact that party positions relative to the government and other parties on the EU dimension also drive opposition suggests that parties primarily operate along this axis to further their policy-seeking objectives.

The results of our study open up interesting avenues for future research. We end by highlighting two lines of future investigation that could improve our understanding of the dimensionality of oppositional behaviour in EU affairs by addressing some of the limitations of the current study. A first important line of inquiry would be to include more cases to allow a rigorous control for systemic features that may impact our results. Previous work on the dimensionality of the political space in Western Europe has demonstrated substantial cross-national variation when it comes to how dimensions of political contestation relate to each other (Bakker et al. 2012). To check if the results of the present study are robust we would need to extend our study beyond our six parliaments. Only then would we be able to determine whether and to what extent systemic features impact on the dimensionality of disagreement when parties engage in oppositional behaviour. Second, the fact that our study only includes data from one of the key arenas in parliament, the committee, while leaving plenary debates out of the picture, means that we need to be cautious about concluding that we have provided the full picture of the dimensionality of oppositional behaviour in EU affairs. A thorough understanding of how the EU dimension and the left-right dimension relate to each other when parties express opposition in national parliaments would require that we examine if and how oppositional behaviour differs between EAC deliberations and the more public plenary debates.

Acknowledgements

The authors would like to thank three anonymous reviewers and the editors of WEP for their helpful comments. We are also grateful to Frederike Albrecht and Timothy Bergfalk for excellent research assistance.

Disclosure statement

No potential conflict of interest was reported by the author(s).

Additional information

Funding

This work was supported by Marianne and Marcus Wallenberg Foundation [grant: MMW 2014.0011], and the Swedish Research Council [grant: 2017:02931].

Notes on contributors

Christer Karlsson

Christer Karlsson is a Professor of Political Science at the Department of Government, Uppsala University. His research has focused on political opposition, constitutional change and international climate change politics. His latest work has appeared in journals such as Environmental Politics, Journal of European Public Policy, Journal of Common Market Studies, Parliamentary Affairs and Party Politics. [[email protected]]

Moa Mårtensson

Moa Mårtensson is an Associate Professor of Political Science at the Department of Government, Uppsala University. Her research interests include political opposition, political and social equality, as well as climate policy. Her most recent work is published or forthcoming in British Journal of Political Science, Journal of European Public Policy, and Electoral Studies amongst others. [[email protected]]

Thomas Persson

Thomas Persson is an Associate Professor of Political Science at the Department of Government, Uppsala University. His research interests include European Union politics, political behaviour and political parties. His articles have appeared in journals such as British Journal of Political Science, Journal of European Public Policy, Party Politics and Journal of Common Market Studies. [[email protected]]