Multilevel Party Competition and Legislative Resources

This section presents a theoretical model of preference formation within multilevel settings to explain how European parties’ policy positions are formed. We conceptualize position formation within Europarties as a two-level bargaining process in which national parties compete with each other in an attempt to upload their own policy positions to their Europarty. We expect that national parties’ ability to shape the policy preference of their Europarty is determined by their legislative resources; that is, the share of seats they control in the EP.

Parties are conceptualized as rational, goal-oriented and purposeful collective actors. Following De Swaan's party behaviour model,Footnote ²² we assume that parties are policy-seeking actors and that they are motivated by political goals that guide their behaviour. We also assume that they intrinsically value policy outcomes and sincerely believe in the policy goals they seek to achieve. Political parties therefore strive to maximize their influence on policy-making to achieve a policy outcome that is as close as possible to their own ideal position. A party's ability to pursue its policy objectives depends on its capacity to shape the political decision-making process. Political parties therefore also implicitly strive for office as a means of influencing policy.Footnote ²³

EP elections have often been described as ‘second-order’ elections that are generally viewed as less important as there is ‘less at stake’ than in national elections, which determine the composition of the domestic government.Footnote ²⁴ As a result, public salience and turnout is considerably lower than in national elections. However the powers of both the EU and the EP have increased considerably over time, and successive treaty changes have turned the EP into a powerful co-legislator. Policy-seeking national parties therefore have strong incentives to influence European decision-making through the EP, as a large number of political decisions is taken at Brussels.

While national-level political parties can independently choose their policy goals and single-handedly pursue their policy objectives in the domestic arena, the situation on the European level is much more complex: national parties come together in Europarties, which are transnational federations of national parties from a wide variety of member states. Even though national parties join these Europarties based on their ideological orientation and policy proximity,Footnote ²⁵ there is a considerable variation in terms of the ideal policy positions of national parties within the same European party. For example, the British Labour Party, the French Socialists and the German Social Democratic Party are all members of the Party of European Socialists, but they vary extensively in their ideology (see Figure 1). Despite these ideological differences, national parties within the same Europarty have to settle on a common policy position in order to speak with one voice in Brussels and effectively pursue their goals on the European level. Europarties have accordingly adopted their own election manifestos ever since the first direct EP election in 1979. These party manifestos guide the legislative activities of the Europarties and their associated party groups in the EP. MEPs regularly receive voting instructions based on the policy guidelines of these manifestos, and they can be punished for not adhering to them.Footnote ²⁶ New member parties are also required to officially agree to follow the manifestos’ policies.Footnote ²⁷

Given that national parties within the same Europarty differ considerably in their policy goals, the crucial question is how these very diverse national parties arrive at a common policy position at the European level. In other words, how do national parties with diverse policy preferences agree on a common position as expressed in the election manifesto of their Europarty? In order to explain Europarties’ policy position choices, their preference formation is conceptualized as a two-level party competition in which the success of national parties depends on their share of EP seats.Footnote ²⁸ As discussed earlier, we conceptualize political parties as policy-seeking actors that pursue a particular policy objective. While national parties can independently choose and pursue their own policy goals in the domestic arena, they cannot single-handedly do so on the European level. By contrast, they have to come to an agreement with other members of their Europarty concerning the policy positions of their common European party federation. National parties therefore seek to influence position formation within their Europarty in order to upload their own policy preferences to their European party. The importance that national parties attach to the election manifestos of their Europarties is reflected by the seniority of the representatives that national parties send to negotiate the Europarty manifestos. For instance, the British Labour Party sent its foreign minister to lead the working group that was set up in 1998 to draft the PES election manifesto for the EP election in 1999.Footnote ²⁹ The French, German and Italian member parties also sent high-ranking officials to these working group meetings.Footnote ³⁰

National parties that are members of the same European party therefore compete with each other in an attempt to upload their own individual policy preferences to the European party. Drawing on insights from the literature on coalition governments, we argue that a national party's share of parliamentary seats determines its success. Coalition theorists have argued that the payoffs that political parties gain from coalition governments crucially depend on the resources a party brings to the coalition.Footnote ³¹ GamsonFootnote ³² reasoned that ‘any participant will expect others to demand from a coalition a share of the payoff proportional to the amount of resources which they contribute to a coalition’, which has been famously denoted as ‘Gamson's law’.Footnote ³³ Even though Europarties do not constitute traditional government coalitions, which are the focus of previous coalition research, they are alliances of national parties that share a variety of features with government coalitions. National parties come together in Europarties to achieve common legislative objectives, just as national parties come together in a coalition government in order to govern together. Similarly, national parties demand payoffs for their participation in Europarties, just as national parties demand payoffs from their participation in coalition governments. Europarties can therefore be conceptualized as transnational coalitions of national parties that demand payoffs for their participation. Gamson's law therefore offers important insights to help understand payoff allocation in Europarties.

Payoffs are goods that are distributed among coalition partners and used by them to advance their individual objectives.Footnote ³⁴ In line with the policy-seeking party model, we assume that the crucial payoffs for national parties on the European level consist of policy prerogatives.Footnote ³⁵ National parties pursue their own policy objectives and try to upload their individual policy goals to their Europarty. They seek to adopt a common Europarty position that is as close as possible to their own preferred position. In a unidimensional policy space, the utility for national parties declines with the distance between the Europarty position and their own policy stance. This utility function is formulated as follows:

$$ {{U}_j}\: = \:{\rm{ - (}}{{P}_{NATj}}{\rm{ - }}{{P}_{EUR}}{{{\rm{)}}}^{\rm{2}}}, $$

where U_j is the utility for a national party j. P_NATj is the position of national party j on the unidimensional policy scale and P_EUR is the location of its Europarty on the ideological scale. Thus the utility rises as the policy distance between a national party and its Europarty decreases, and reaches its maximum at zero – indicating absolute congruence between the Europarty and the preferred position of the national party.

We furthermore assume that the most important resource that a national member party provides to its Europarty is its share of seats in the EP.Footnote ³⁶ Their seat share reflects the degree of legislative support that national member parties can deliver to their Europarty to achieve common policy objective.Footnote ³⁷ The strength of a Europarty, or more precisely its associated party group, crucially depends on the number of seats it controls in the EP. As decisions in the EP are taken by majority vote, the ability of Europarties to achieve their policy goals in this forum depends on their seat share. The EP's powers have considerably increased over time; it is now a veritable co-legislator on the European level. Policy-seeking parties can therefore influence the EU legislative process through the EP in order to achieve a policy outcome that is as close as possible to their preferred policy positions. The ability to shape EU policy-making through the EP is determined by the legislative resources of national parties. The higher a national party's seat share, the larger its impact on the EP's position on policy initiatives and – ultimately – on policy outcomes. National member parties that won a large number of parliamentary seats therefore provide their Europarty with considerable legislative resources that are necessary to achieve the common policy objective. As a result, the number of seats a national party can bring to the table directly determines its level of influence in negotiations about defining common policy position for their European party. As Euromanifestos are drafted in the run-up to EP elections, the number of seats a national party currently controls is decisive. National parties that gained a large number of seats in the last election have demonstrated their ability to win a large number of parliamentary mandates for their Europarty. Since their prospects of wining a similarly large number of seats in the upcoming election are promising, they therefore wield greater power in drafting the Europarty manifesto. Hence the larger the share of parliamentary seats a national party currently controls, the higher the policy payoffs it can gain from negotiations with other members of the same Europarty.Footnote ³⁸

$$ {{P}_{EUR}}\: = \:\mathop{\sum}\limits_{j\:{\rm{ = }}\:{\rm{1}}}^J {{{P}_{NATj}} \,\ast \,{{\alpha }_j}}, \; \; with\,\mathop{\sum}\limits_{J\:{\rm{ = }}\:{\rm{1}}}^J {{{\alpha }_j}\:{\rm{ = }}\:{\rm{1}}} $$

Formula 2 illustrates our theoretical model.Footnote ³⁹ We expect that the policy position P_EUR that a Europarty adopts is the result of multilevel negotiations in which J national member parties attempt to upload their positions to their Europarty. The policy position choice of a Europarty P_EUR corresponds to the sum of the weighted preferences of national member parties. The preferences of national member parties P_NATj are weighted by their legislative resources a_j. The legislative resources of a national party correspond to its share of the aggregated number of EP seats that all national member parties bring to their Europarty. The policy preferences of national parties with a high degree of legislative resources a_j have a larger impact on the policy position of the Europarty P_EUR. Therefore the policy position of a Europarty does not correspond to the average position of its national parties. National member parties do not have an equal weight in multilevel negotiations; their influence depends on the number of EP seats they control relative to the other member parties. Policy position choices of Europarties are therefore skewed towards national parties that gain a large number of seats. Hence the distance between the ideal point of a national party and the policy position of its Europarty is negatively associated with its legislative resources. National member parties with a high seat share are located closer to the ideal point of their Europarty than national parties that won only a small number of parliamentary seats.

Research Design

In this section we describe how we constructed the dataset that was used to empirically test our theoretical expectations. We first explain how we measured the policy positions of European and national parties, as well as the distance between these position estimates. Afterwards we will discuss how the explanatory and control variables were operationalized.

Measuring Policy Positions of Parties and their Distances

To measure the policy positions of European and national parties, we rely on the Euromanifesto dataset.Footnote ⁴⁰ European parties, and most of their national member parties, have adopted election manifestos for all EP elections since its first direct election in 1979. Party manifestos have been used by a wide variety of scholars to estimate the policy positions of political parties, since the parties clearly spell out their ideological stances on a variety of policy issues.Footnote ⁴¹ The most widely used dataset on policy positions is provided by the Comparative Manifesto Project (CMP), which analyses national party manifestos in a wide variety of countries from 1945 until today by means of manual hand coding.Footnote ⁴² The CMP developed a classification scheme with fifty-six categories grouped into seven policy domains. Where possible, directly opposing pro and contra categories were specified. The Euromanifesto project (EMP) manually codes the election manifestos issued by European and national parties for EP elections using a coding scheme similar to the one developed by the CMP. Human coders first divided the election programmes into ‘quasi-sentences’ that contain an argument understood as ‘the verbal expression of one political idea or issue’.Footnote ⁴³ These quasi-sentences were then allocated to the appropriate categories, as specified in the EMP coding scheme.

In order to compute policy position estimates, we drew on the widely used ‘RILE’ procedure. As discussed above, the Euromanifesto project divided all EP election manifestos into quasi-sentences and allocated them to predefined policy categories. Some of these categories are considered to indicate left-winged policy positions and some are classified as right-winged issues. The ‘RILE’ scaling technique yields policy position estimates from these coded manifestos based on the following procedure. First, the percentages of left and right categories of the total number of coded quasi-sentences are computed. Then the percentage of left sentences is subtracted from the percentage of right sentences. Negative scores represent left positions and positive scores represent right positions. At the extreme, a party devoting its entire program to left-wing issues would score −100; similarly a totally right-winged program would receive a score of +100. For example, if a party manifesto contained 200 quasi-sentences, of which 100 (50 per cent) are allocated to left categories and 40 (20 per cent) to right categories, it would receive a score of −30 (that is, 20–50). Concerning the European dimension, we used the ‘pro-anti European integration scale’ constructed by subtracting the percentage of anti-EU sentences from the percentage of pro-EU sentences. In this case, positive scores represent a pro-integration attitude, while negative values represent an anti-integration position.Footnote ⁴⁴

There are also alternative approaches to measuring policy positions; expert surveys are the most prominent.Footnote ⁴⁵ There is a vibrant discussion in the literature about the quality of manifesto data and expert surveys for the measurement of policy positions, as each approach is associated with certain advantages and disadvantages.Footnote ⁴⁶ We decided to use Euromanifesto data for theoretical and empirical reasons. First, from a theoretical point of view, this study analyses Europarties’ preference formation with regard to drafting common Europarty election manifestos. We conceptualized preference formation as a two-level party competition in which national parties compete to influence the policy position that the Europarty adopts in its election manifesto. From a theoretical perspective, the election manifestos adopted by the Europarties and their national member parties are therefore the ideal data source for our analysis. Secondly, from an empirical point of view, it is crucial that the estimates measure policy positions at the time of EP elections to have simultaneous position estimates for when national parties negotiate their Europarty manifestos. As the expert surveys are only available for a limited number of time points, and are not available simultaneously for national parties and their Europarties, we decided to rely on the Euromanifesto data, which provides us with simultaneous policy position estimates for all Europarties and all their national member parties for all EP elections in 1979, 1984, 1989, 1994 and 2004. We performed the following check, based on the Chapel Hill expert survey data, to test the robustness of our results.Footnote ⁴⁷ First, we computed the Europarty position based on the average positions of its national member parties, as estimated by the expert survey. Secondly, we computed the Europarty position based on the average positions of its national member parties, weighted by their relative seat share, in line with our theoretical model summarized in Formula 2. Finally, we compared these estimates to the Europarty position estimates derived from the Euromanifesto data. The results indicated that the average expert survey position of national parties – weighted by their relative seat share, as suggested in our theoretical model – better predicts the Euromanifesto positions than the simple average expert survey positions of national parties. The expert survey data therefore supports our findings and the validity of the Euromanifesto position estimates.

In order to study policy position formation within European parties, we concentrated on the four major parties on the European level, as they have dominated politics in the EU:Footnote ⁴⁸ the EPP, the PES, the ELDR and the EGP. The EPP is the largest political group in the EP and currently controls 271 of its 736 seats: it incorporates forty-eight Christian Democratic and Conservative parties from the twenty-seven EU member states. The PES brings together the EU's Socialist, Social Democratic and Labour Parties; it is currently represented by 184 MEPs and it is composed of thirty-three national parties. The third-largest political force in the EP is the ELDR, which has seventy-two of the 736 seats and brings together fifty-four liberal parties from various European countries. The EGP, which comprises thirty-eight green parties from various European countries, currently has the smallest number of representatives in the EP (forty-seven). All together the dataset includes 286 observations, which constitute national member parties that aim to shape the election manifestos adopted by the four major Europarties across the six EP elections between 1979 and 2004.

To measure the absolute distance between the ideal points of national member parties and the policy position adopted by their European party (P_EUR − P_NATj) we computed the Euclidean distance between these two positions. We analysed the distance between European parties and their national members on the left-right scale for two reasons. First, several scholars have empirically evaluated the policy space at the European level and concluded that it is characterized by a unidimensional left-right space.Footnote ⁴⁹ Secondly, although some other scholars disagree and suggest that the European political space is characterized by two dimensions (left-right and pro-anti European integration dimensions),Footnote ⁵⁰ there is hardly any variation on the pro-anti European integration dimension.Footnote ⁵¹ This is particularly the case, since we focus on the four major European parties that largely agree on the scope of European integration, as compared to Eurosceptic parties. Figure 2 illustrates the absolute ideological distance between national parties and their Europarties on the left-right scale. To check the robustness of our results across a different conceptualization of the dimensionality of the European political space, we repeated our analysis for the two-dimensional space (left-right/pro-anti integration). We measured the distance between national parties and their Europarties on the left-right and pro-anti European integration based on the two-dimensional Euclidean distance.

Fig. 2

Absolute distance between national parties and their Europarty on the left-right scale, 1979–2004 Note: EGP = European Green Party, PES = Party of European Socialists, ELDR = European Liberal Democrat and Reform Party, EPP = European People's Party.

Measuring Independent Variables

We measured a national party's legislative resources by the number of EP seats it won in the last EP election relative to other member parties of the same Europarty. We therefore used the lagged legislative resources of national parties to predict their influence on Europarties’ policy position choices.Footnote ⁵² The number of parliamentary seats simultaneously captures the overall weight of a national party's member state in the EP and its relative importance in its home country, as a party's seat share is determined by its domestic vote share and the number of seats allocated to its home country in the EP. The vote share of national parties in their home country or the absolute number of seats are, by contrast, insufficient to measure the supply of legislative resources to their Europarty. For instance, a national party from Malta could win 100 per cent of the national votes, but would still provide fewer seats to its Europarty than a German party winning only 10 per cent of the national votes, since Malta only has six seats, while Germany has ninety-nine seats in the EP. Similarly, as the EPP overall currently controls 271 seats while the EGF controls fifty-eight, a national member party of the EPP that won twenty seats is less important to its Europarty than a Green party that only won ten seats, since the Green party provides about 17 per cent of its Europarty seats while the Conservative party only provides 7 per cent of all the EPP seats.

To obtain the number of seats relative to other member parties of the same Europarty a_j of national party j, we divided the number of seats x_j obtained by party j by the total number of seats gained by all members of the same Europarty. To ease interpretation of the regression coefficients, we then multiplied the estimate by 100 so that the measure of legislative resources ranges from 0 to 100.

$$ {{\alpha }_j}\: = \:\frac{{{{x}_j}}}{{\sum _{{j\: = \:1}}^{J} {{x}_j}}}{\rm{\ast100}} $$

As the proximity model of voting predicts that European parties adopt policy positions based on the distribution of their supporters’ policy preferences, we controlled for the distance between the ideal points of national parties and the median policy position of the potential European electorate.Footnote ⁵³ We measured the location of the median ideal point using survey data gathered from the European Election Studies (EES) for EP elections between 1989 and 2004. For the 1979 and 1984 elections we relied on Eurobarometers 12 and 22.Footnote ⁵⁴ All these surveys included several questions about electoral participation and voting behaviour in EP elections. One of the questions refers to the classical left-right individual self-placement. This variable ranges from 0 (extreme left) to 10 (extreme right) and is available for all EP elections. The EES furthermore comprises a question that measures citizens’ attitudes towards European integration. This variable ranges from 1 (unification has already gone too far) to 10 (integration should be pushed further).Footnote ⁵⁵ Unfortunately this question was only included in the last two EES waves (1999 and 2004), which causes a significant reduction in the number of cases. We transformed these two scales so that they correspond to the ideal point scales of political parties.Footnote ⁵⁶

The median ideological position of the potential electorate of a European party was calculated as the median policy position of citizens that voted for the party.Footnote ⁵⁷ Thus for each Europarty we calculated the median ideal point of its potential electorate on the left-right scale for each election, and on the pro-anti European integration dimension for the 1999 and 2004 elections. In order to measure the distance between a national party and the median position of the potential European electorate of its Europarty, we then computed the Euclidean distance on the left-right scale and in the two-dimensional policy space.

We also included several additional control variables in the analysis. To account for the potential learning effects of prior experience in negotiating the policy positions of European parties, we controlled for the duration of EU membership of a national party's home country, measured by the number of years between its accession and the date of the election. We used the logged number of years, as it is plausible to assume that the size of the learning effect decreases over time. We also controlled for the salience of EP elections in different member states of the EU. If EP elections are particularly salient in a member state, national parties from this country might, on average, have stronger incentives to influence the position formation of their Europarty, as their domestic constituents more closely monitor the election campaign than in countries where EP elections raise little public interest. We would therefore expect that the policy distance to Europarties decreases, on average, with the salience of EP elections in the home countries of national member parties. It has to be noted, however, that national parties from the same country might place different emphasis on EP elections, despite the high salience of EP elections to their electorate. We measured the salience of EP elections as the difference in turnout between European and national elections.Footnote ⁵⁸ European elections are generally regarded as ‘second-order’ elections that, just like many local and regional elections, lack salience; therefore turnout is generally low.Footnote ⁵⁹ There is seen to be ‘less at stake’ in second-order elections than in first-order elections (typically national parliamentary elections). As a consequence, there are fewer incentives for people to vote in these contests, so turnout is generally lower than in national elections.Footnote ⁶⁰ We also took into account whether national parties were in government at the national level by including a dummy variable that was coded based on data gathered from the Political Data Yearbook published by the European Journal of Political Research. National parties that form the government in their home country are represented through their MEPs as well as their ministers in the Council of the EU and their heads of government in the European Council. As members of the Council at both levels, national government parties can crucially affect the EU's agenda and the outcomes of legislative processes. They can therefore determine the policy issues on which Europarties have to position themselves, and due to their voting power in the Council, they can also offer legislative support to Europarties with regard to upcoming policy initiatives. Since they can also rely on an extensive network of preparatory bodies, they enjoy important information advantages over national opposition parties. It is therefore crucial to control for government participation. Table 1 provides summary statistics of all variables in the model.

Table 1

Summary Statistics of Variables

Data Analysis

The special structure of the data has to be taken into account in order to test our theoretical expectations. National parties are members of European parties and seek to shape the policy positions of their European party in the run-up to EP elections. More specifically, every national party is a member of one particular European party and competes with other members of the same European party to shape the election manifesto adopted by the Europarty for a specific EP election. For instance, since the German Social Democratic Party and the Spanish Socialist Workers’ Party are both members of the PES, they compete to shape the policy positions that the PES adopts in its election manifesto. European parties also compete every five years in European-wide elections to the EP. For each of these elections, all four major European parties adopt a new election manifesto, which their national member parties seek to influence. Position formation in the run-up to EP elections might vary considerably, with contextual characteristics specific to each particular election. For instance, the ability of national member parties to influence policy position choices within European parties might be affected by the number of EU member states or the issues on the political agenda during a specific election campaign. The election-specific context might therefore facilitate (or hamper) national parties’ ability to shape the election programme of their European party.

National parties are therefore clustered within the four major European parties and each specific election year so that the observations are not completely independent, as assumed by ordinary regression. In order to take this clustering of data into account, we estimated ordinary least square (OLS) regression with clustered robust standard errors.Footnote ⁶¹ As national parties are clustered simultaneously in Europarties and election years, we created a cluster variable that takes this twofold clustering into account. The data structure leads to twenty-four clusters based on four Europarties (EPP, ELDR, PES, EGP) and six elections (1979, 1984, 1989, 1994, 1999, 2004). In addition to the clustering, the dataset is also characterized by a time component, as six succeeding elections are analysed. To control for potential autocorrelation induced by the time-series structure of the data, we included the lagged dependent variable.Footnote ⁶² However, as incorporating lagged dependent variables is also associated with a number of problems, we have also estimated the model without the lagged dependent variable.Footnote ⁶³ As the results are substantially the same, we only present the full model, which includes the lagged dependent variable.Footnote ⁶⁴

Unfortunately, we can only compute the ideological distance between the potential European electorate and the national parties on the left-right scale for all elections to the EP. As mentioned previously, the self-placement of voters on the pro-anti European integration dimension is only available for the 1999 and 2004 elections. Including the ideological distance on the pro-anti European integration dimension thus considerably decreases the number of observations. We therefore also estimated the OLS regression excluding the ideological distance between national parties and voters on the European dimension to test whether the effects detected in the smaller sample also hold across a larger number of cases. As the results are substantially the same, we only present the complete models including all variables.

Table 2 presents the results of the regression analysis. The first model reports the results of the analysis of the distance between European and national parties on the left-right scale. The second column contains the results of the analysis based on the distance in the two-dimensional left-right and pro-anti EU space. As theoretically expected, the legislative resources of national parties are negatively associated with the distance between their ideal point and the policy position of their Europarty. Across both models, legislative resources have a statistically significant negative effect on the ideological distance between a national party and its Europarty. More precisely, if legislative resources increase by one unit, the distance between a national party and its Europarty decreases by 0.161 units in the left-right and by 0.331 units in the two-dimensional space. National parties with a large number of parliamentary seats are therefore particularly successful in pulling their Europarty position towards their ideal points: the higher the legislative resources of a national party, the stronger its ability to determine position formation within its Europarty. By contrast, the distance between the policy position of a national party and the median position of the potential European electorate does not have a statistically significant effect in any of the model specifications. Similarly, the duration of EU membership, the salience of EP elections and being in government at the domestic level do not have a systematic effect on the ability of national parties to shape position formation within their Europarty. The lagged dependent variable has a statistically significant positive effect, which indicates that national parties that fail to shape the election manifesto of their Europarty at t ₀ will most likely also not be successful in determining its policy position choice at t ₁.

Table 2

OLS Regression with Clustered Standard Errors

***p ≤ 0.01, **p ≤ 0.05, *p ≤ 0.10; standard errors in parentheses.

To illustrate the effect of legislative resources on the ideological distance between national parties and their Europarty, we simulated predicted values as suggested by King, Tomz and Wittenberg.Footnote ⁶⁵Figure 3 displays the simulated predicted distance in the (a) left-right and (b) two-dimensional spaces as legislative resources change while holding all other variables constant. The point estimates of the predicted values are indicated by the solid lines and the 95 per cent confidence intervals are illustrated by the dashed lines. The predicted distance between Europarties and their national member parties in the left-right and two-dimensional space constantly decreases with an increase in national parties’ legislative resources while other variables are held constant. Thus legislative resources have a steady positive effect on the ability of national parties to influence policy position choice within their European party.Footnote ⁶⁶

Fig. 3

The effect of legislative resources

To test the robustness of the findings, we estimated three further model specifications (see Table 3). First, we included a dummy variable for party family to check whether the effect of legislative resources on the policy position choice of Europarties might differ across different party families. The results confirm the previous analysis, as legislative resources also have a statistically significant effect on policy position choice if we control for party family. Secondly, one might also argue that the size of the country could be driving the effect of legislative resources, in the sense that parties from large member states dominate position formation. However country size is already implicitly included in the model, as the number of EP seats a national party can win depends on the size of the country. The allocation of EP seats to member states is more or less proportional to their population. National parties from large member states such as Germany (99 seats) can therefore win more seats than parties from small member states such as Malta (six seats). The legislative resources of national parties therefore already implicitly take country size into account. However, in order to test the robustness of the findings, we estimated an additional model specification that explicitly controlled for country size measured by the population in hundreds of thousands of inhabitants. Even when additionally controlling for population size, the direction (and, by and large, also the size) of the effect of legislative resources remains constant. Thus the effect of legislative resources is not driven by the size of the country, but instead has an independent effect on position formation. It has to be noted, however, that the effect of legislative resources is only statistically significant in the left-right space and not in the two-dimensional space, which might be explained the smaller number of cases (N = 94) and the relatively high correlation between legislative resources and population size in that sample (r = 0.59). Thirdly, as some national party manifestos may be drafted only after the Europarty has already adopted its election manifesto, we reran the analysis using national parties’ policy position estimates obtained from their party manifestos from the previous EP election. We essentially arrived at the same findings whether we used the election manifestos of national parties drafted for the same or the previous EP election.Footnote ⁶⁷

Table 3

Alternative Model Specifications

***p ≤ 0.01, **p ≤ 0.05, *p ≤ 0.10

Note: The reference category for the party family dummies is green parties. Standard errors in parentheses.