2. The role of transnational party alliances

On first glance, the account that parties copy foreign incumbents as a heuristic to cope with uncertainty seems reasonable. But there is a large pool of foreign incumbents from which political parties could potentially learn, which suggests that there is considerable information to process in order for copying from foreign incumbents to occur. Surprisingly, recent studies do not find that domestic parties learn from and emulate foreign incumbent parties of the same party family. Peculiarly, this (non-)finding is at odds with findings in the literature on policy diffusion suggesting that governments are influenced by domestic and foreign governments—particularly from similar ideological camps (see e.g., Shipan and Volden, Reference Shipan and Volden2008; Gilardi, Reference Gilardi2010). Furthermore, it goes against the anecdotal evidence that actually motivates current research on party policy diffusion.

We consider influential research about political parties in the EU (Marsh and Norris, Reference Marsh and Norris1997; Hix, Reference Hix2002) to theorize that party policy diffusion takes place through transnational party alliances at the EU level which are composed of representatives from national parties and provide an arena for inter-party exchange. Transnational party alliances have a special role in facilitating contact between national party representatives. They organize congresses and conferences which bring together prime ministers and party leaders to agree political strategies and resolutions.

Most centrally, transnational party representation in Europe takes place through the party groups in the EP which have huge potential to facilitate party policy diffusion (Wolkenstein et al., Reference Wolkenstein, Senninger and Bischof2020). Since the EP is one of the EU's co-legislators (together with the Council of the EU) a key task of party groups in the EP is to build and coordinate political majorities on legislation. To this end, groups generally convene during “Group week” in Brussels where they prepare the upcoming plenary agenda. In addition, they also meet in Strasbourg during plenary week to (de-)brief parliamentary sittings. While the purpose of these meetings clearly serves the goal of having smooth and informed parliamentary processes, they are also used for, “[…] development of Group positions on major political issues or debates or broader political strategy, and for receiving visiting delegations or leaders of national parties or other personalities” [emph. added] (Corbett et al., Reference Corbett, Jacobs and Shackleton2011, 117).

Hence, meetings of transnational EP party groups go way beyond the work in the EP. They organize their own political activities which form an important channel of communication between national parties, as groups frequently welcome representatives from national parties, including ministers and front-bench parliamentarians. In addition, they often send delegations to national parties, organize seminars and conferences with national parties, and they publish brochures, studies, and newsletters aimed in part at national parties. A striking recent example is a meeting of the European Liberal Forum in April 2018 which brought together top-level representatives from liberal parties belonging to the EP party group ALDE “with the aim to equip politicians and staff members from liberal parties across Europe with concrete arguments and strategies on how to counter populistic and nationalistic tendencies” [emph. added] (European Liberal Forum, 2018).

Representatives to the EP participating in events and meetings organized by the transnational EP party groups have regular contact to the national party and contribute to the position-taking of the national party just like parties’ national parliamentary delegations and other factions (Gauja, Reference Gauja2013; Greene and O'Brien, Reference Greene and O'Brien2016; Senninger and Bischof, Reference Senninger and Bischof2018; Ceron, Reference Ceron2019).

EP party groups are not the only form of transnational party cooperation at the EU level. There are also European Parties (“Europarties”) that are organized outside of the EP, but these are weak and depend almost entirely on the EP groups for funding and organizational staff (Kreppel, Reference Kreppel2002). As a result, we suggest that transnational party policy diffusion is channeled through EP party groups and not Europarties. Belonging to transnational party groups in the EP also differs from belonging to party families because relations between parties in EP party groups are highly institutionalized and more active, and membership does not perfectly overlap, i.e., parties in the same party family do not automatically belong to the same EP party group. Therefore, EU membership, success in European elections, and strategic decisions about membership make transnational party alliances in the EP a much more important category for understanding party policy diffusion than party family.

In sum, institutionalized transnational party alliances in the EP provide a channel through which parties observe, learn, and evaluate the successful election strategies of foreign incumbents:

3. Research design

To test our hypothesis, we use time-series cross-sectional data from Böhmelt et al. (Reference Böhmelt, Ezrow, Lehrer and Ward2016) that include 215 political parties from 26 Western European democracies for the time period from 1977 to 2010.Footnote ¹ The unit of analysis is a party-year.

To measure party's left-right position, we rely on the Comparative Manifesto Project data (MARPOR) (Volkens et al., Reference Volkens, Lehmann, Merz, Regel and Werner2014), which provide a measure of each party's general left-right placement (RILE). The original MARPOR variable ranges from -100 (extreme left) to +100 (extreme right). We use a linearly rescaled version of the response variable that ranges from 1 (extreme left) to 10 (extreme right). As information from MARPOR codes only election years, values from years between elections are missing. Therefore, each party's position between election years is assigned its value at the last election.Footnote ²

Using party's left-right placement as our dependent variable we estimate spatio-temporal autoregressive models of the following form:

(1)$$y_{it} = \phi y_{it-1} + {\boldsymbol \beta} {\bf x}_{it-1} + {\boldsymbol \rho} {{\bf W}y_{ie-1}} + \lambda_i + \tau_t + \epsilon_{it}$$

where y _it is the dependent variable (RILE Party Position) for party i at time t, y _it−1 signifies the (one year) temporally lagged dependent variable (Lagged RILE Party Position), x_it−1 is a vector of temporally lagged control variables as described in Böhmelt et al. (Reference Böhmelt, Ezrow, Lehrer and Ward2016), λ _i and τ _t are party and year-fixed effects respectively, and ɛ is the error term.

Wy _ie−1 is then the spatial lag variable that is the product of a spatial weights matrix W (NxNxT) that captures the relative connectivity of units w _ik and a matrix (NxT) of the time-lagged dependent variable of other units k. The time lag is constructed such that it considers positions of parties in the year before the last election held in their country.Footnote ³ The approach considers that drafting manifestos is a “time-consuming process [...] which typically takes place over a two-three year period during which party-affiliated research departments and committees draft sections of this manuscript, which are then circulated for revisions and approval upward to party elites and downward to activists” (Adams and Somer-Topcu, Reference Adams and Somer-Topcu2009: 832).

As described above, our decision to use a time-lagged spatial lag variable is theoretically motivated. However, under certain assumptions, the approach also helps to avoid simultaneity bias of spatial ordinary least squares (S-OLS) (Beck et al., Reference Beck, Gleditsch and Beardsley2006; Franzese and Hays, Reference Franzese and Hays2008b; Franzese et al., Reference Franzese, Hays and Cook2016). A central assumption is that interdependence does not occur instantaneously which in our study is probable due to the time-consuming process of drafting manifestos. In addition, parties’ actions do not depend on foreign parties’ expected actions. Instead parties are backward-looking (i.e., they look to what foreign incumbents have done in the past), again implying non-instantaneous interdependence.Footnote ⁴

Even though the tools to empirically analyze interdependence between units have significantly developed in recent years (Anselin, Reference Anselin1988; Franzese and Hays, Reference Franzese, Hays, Box-Steffensmeier, Brady and Collier2008a; Darmofal, Reference Darmofal2015), challenges for empirical researchers still exist. First, one needs to distinguish interdependence from correlated responses to common shocks. Second, interdependence must be disentangled from temporal dynamics (Franzese and Hays, Reference Franzese and Hays2006, Reference Franzese and Hays2008b; Plümper and Neumayer, Reference Plümper and Neumayer2010).

We take different measures to reduce the possibility that what appears to be interdependence among political parties is actually the result of common exposure, and temporal dynamics (common trends) are addressed by adding a lagged dependent variable and year fixed effects. Further, party fixed effects hold unobserved variables constant. In addition to the above considerations, we account for public opinion and globalization. Including such relevant non-interdependent (domestic and unit-level, exogenous-external, and context-conditional) explanations is important.Footnote ⁵ Failure to include the variables described above in the model leads to overestimating of the effects of interdependence.

4. Empirical results

Model 1 (without controls) and Model 2 replicate earlier studies that conclude that foreign incumbent parties promote party policy diffusion (Böhmelt et al., Reference Böhmelt, Ezrow, Lehrer and Ward2016). The estimates on the spatial lag variable (ρ Foreign Incumbent) in our models are similarly positive and statistically significant, which is consistent with these previous findings. The remaining models turn to our hypothesis about the influence of transnational party groups. The spatial lag variable linking parties that belong to the same EP party group (ρ Foreign Incumbent: Same party group) is positive and statistically significant. The spatial lag variable linking parties that are not in the same EP party group but represented at the EU level (ρ Foreign Incumbent: Different party group) is positive, smaller in size, and not statistically significant. The spatial weights matrices are not row-standardized (Williams, Reference Williams2015; Williams et al., Reference Williams, Seki and Whitten2016), and, consequently, to estimate the short-term effects (i.e., first post-diffusion period effects), the coefficients of the spatial lags are multiplied by the average number of neighbors (Plümper and Neumayer, Reference Plümper and Neumayer2010).Footnote ⁶

Each political party in the data has 22.29 neighbors, on average, that are foreign incumbents, and only 1.97 neighbors that are foreign incumbents and belong to the same party group. Also, each party has 7.82 neighbors that are foreign incumbents but belong to different party groups. Focusing on our first hypothesis, the short-term effect of foreign incumbents in the same party group is 0.003 (SE = 0.001). It is calculated from its rounded coefficient of 0.002 (Model 4) multiplied by the average number of neighbors. The coefficient size is comparable to findings in other research on party policy diffusion (Böhmelt et al., Reference Böhmelt, Ezrow, Lehrer and Ward2016).

Substantively, a party's left-right policy position would be 0.003 points farther to the right in the short run, if the two foreign incumbents (on average) in the same transnational party group shift one unit to the right, compared to the year before. For the second spatial lag variable (Foreign Incumbent: Different party group), if all foreign incumbents in different transnational party groups (i.e., eight on average) move to the right, the statistically insignificant effect would be 0.001 (SE = 0.002). The difference in the average number of foreign incumbents between the same transnational party group and other transnational party groups is quite large, and the implication of this difference is non-trivial. The fact that the short-term effect of foreign incumbents in the same party group outperforms the short-term effect of foreign incumbents in different party groups underscores the substantive importance of belonging to the same transnational party group: there are fewer parties leading to a greater overall effect on the focal party position.Footnote ⁷

Since our model includes a temporally lagged variable (Lagged RILE Party Position), the coefficient estimates of the spatial lags only reflect short-term effects in a current year (Whitten et al., Reference Whitten, Williams and Wimpy2019). Therefore, we estimate asymptotic long-term effects for our spatial lag variables by considering the coefficient of our temporally lagged dependent variable (Plümper et al., Reference Plümper, Troeger and Manow2005; Plümper and Neumayer, Reference Plümper and Neumayer2010). The asymptotic long-term effects are shown in Figure 1 (left panel).

Figure 1. Temporal short-term and asymptotic long-term effects if foreign incumbents in the same party group and in different party groups shift their left-righ position by one.

Note: PES, Social Democrats; EPP, European People's Party; ALDE, Liberals; UEN, Union for Europe of the Nations.The horizontal bars are 95 percent confidence intervals and the vertical dashed line represents a spatial effect of 0. Estimates in the left panel are based on Models 4 and 5 in Table 1. Estimates in the right panel are based on the models in Table S2.

Table 1. Single and multiple spatial lag regression models (S-OLS and MSTAR)

Note: *p < 0.1; **p < 0.05; ***p < 0.01

If the two foreign incumbents in the same party group (on average) switch their left-right position by one, the long-term effect on a party is 0.013 (SE = 0.005) (0.29 units on the original MARPOR scale). If all foreign incumbents in different party groups switch their left-right position by one, the effect on a party is 0.007 (SE = 0.009), and this effect is insignificant. Also, Model 6 includes both spatial lag variables in a multiple spatial lag regression model, and it suggests that belonging to the same party group is a significant predictor of a party's policy position while belonging to a different party group is not. Model 7 shows results from a model in which the data are updated (1977–2016).

Finally, it might be that how party policy diffusion in the EP occurs varies across party groups. Specifically, Social Democrats should be more inclined to adapt their policies to successful incumbent parties of their group as their efforts to exchange internationally is well known in the scholarly debate (Eley, Reference Eley2002). To estimate heterogeneity across groups, we show long-term effects from four models in which we disaggregate the weight matrices so that only parties within each EP group are connected with each other (Section A in the Supplementary Information). We do find a positive effect of EP group belonging across all groups other than the Liberals, which is consistent with the central findings of this study. Interestingly and in line with our priors discussed here, diffusion is strongest amongst parties belonging to the Social Democratic group.Footnote ⁸ In contrast, results for models in which we disaggregated the weight matrix so that all parties from the Social Democratic party family are connected with each other (see Table S3) provide no evidence for party policy diffusion. This underscores the important role of institutionalized party cooperation in the EP.Footnote ⁹ Furthermore, the case of the UEN provides important evidence showing that foreign incumbents can influence the other parties from the same EP group even when the group has no long-standing history of cooperation beyond the EP. The finding is in line with recent studies on radical right and populist parties that emphasize that these parties increasingly collaborate through EP party groups (McDonnell and Werner, Reference McDonnell and Werner2019).