The political foundations of party organizational variance

The aim of this contribution is twofold: first, to verify empirically how and to what extent party organizations vary within countries, in time; second, to enhance the role of political factors in explaining organizational variance. While mainstream literature has generally overestimated cross-national party convergence, a renewed interest in the study of variance has recently gained ground. We thus focus on seven European countries, from 1990 to 2010, by combining party organizational data from the Party Organization Data Handbook and the Political Party Database Project, with domestic cultural, socioeconomic, technological data from the European Values Survey and the World Bank, as well as supranational economic data provided by the OECD. We are interested in verifying how much of the variance in party organizations can be explained by resorting to the parameters of the party systems vis-à-vis domestic and supranational extra-political factors. Our results show that the explanatory power of party systems’ parameters is stronger than the predictive ability of contextual variables.


Introduction
In the aftermath of WWII, since most liberal democracies were experiencing similar trends in cultural, socioeconomic and technological dynamics, convergence in their policy mix and representation models was considered the by-product of such tendencies (Bennett 1991;Howlett et al. 2009). The convergence hypothesis was further reinforced on the eve of the New Millennium, as a consequence of supranational processes such as globalization and Europeanization (Strange 1996;Knill 2006;Caramani 2010).
In this vein, also party scholars searched for party organizational convergence across liberal democracies. The research strategy based on ideal types-specifically party 'models' (Krouwel 2006;Katz 2017)-represents the most accomplished theoretical effort to bring together changes in contextual dynamics of a mainly extrapolitical nature and party organizational profiles. Regardless of the peculiarities of national political systems, ideal-typical organizational models were expected to be found in most liberal democracies. All in all, the established literature devoted to party organization has primarily focused on party change as an adaptive reaction to contextual stimuli of an extra-political nature, which were common to Western countries.
While the relevance of the 'environmentally-induced change' approach is undisputable (Panebianco 1988;Harmel and Janda 1994;Harmel 2006), there are some aspects over which it can be challenged, at both theoretical and empirical levels. The convergence hypothesis was first questioned by the literature committed to public policy (Howlett et al. 2009) and, in recent years, also by party scholars (Galvin 2016;Riedl 2016;Pizzimenti 2020). As for the latter, the search for new ideal types has been replaced by an orientation toward variance along different organizational dimensions ) and country-specific institutional settings (Poguntke et al. 2020). In particular, the dimensional approach promoted by the Political Party Database Project ) rests on the assumption that party change among party resources, representative strategies and structures does not necessarily follow the same dynamics nor the same timing. Moreover, domestic institutional and political peculiarities are considered relevant explanatory variables for the interpretation of party organizational change, whose patterns may significantly differ from country to country.
In line with this dimensional approach, this contribution aims to enhance the role of domestic factors of a political nature in influencing cross-countries party variance along different organizational dimensions . We aim at verifying whether cultural, socioeconomic and supranational factors are actually able to challenge political predictors-namely party systems' parameters-in explaining party organizational variance.
The paper is structured as follows: In "Convergence, contextualism and party models" and "A different perspective" sections, we introduce the main assumptions underlying our analytical framework. In the fourth section, we present our data and the methods adopted. In the following section, we discuss the empirical findings, while the Conclusive Remarks should help take stock of our analysis.

Convergence, contextualism and party models
The convergence hypothesis is a cross-cutting orientation in political science. Developed since the 1960s, it is based upon the basic idea that industrialization and modernization processes bring countries to experience similar patterns of political development (Bennett 1991), thus declassifying domestic political structures to a marginal role. The convergence hypothesis rests on an approach to politics that March and Olsen (1989; labelled as 'contextualism'-or 'societal functionalism' in Pierson's (2004) words. According to the contextualist approach, the broader socioeconomic context within which political institutions and organizations operate determines their profiles, preferences and choices, which adapt to external pressures. At the end of the twentieth century, the narrative centered on the homogenization effects produced by processes set at the supranational level (Caramani 2010)-such as economic globalization and political internationalization (Busch and Jörgens 2005)-further reinforced the contextualist approach.
In the party literature as well, the 'environmentally induced change' approach (Harmel and Janda 1994;Harmel 2006) traditionally represented the main lens through which party organizational change has been studied. In line with the convergence hypothesis, as Western European democracies were moving toward similar models of political representation-underpinned by analogous contextual processes (Dalton and Wattenberg 2002)-changes in party organizational profiles followed suit. As Katz and Mair (1990: 18) argue «[…] the ultimate source [of change] is in the party's environment, […] parties adapt to changes in their environment». The search for party models-a research strategy based on ideal types (Katz 2017)-was thus conceived to bring together contextual changesconsidered as independent variables-and party organizational change-the dependent variable. The literature on party models regards each new organizational model as the by-product of changes in the relations between parties, civil society and-since the 1990s-the State. This tradition can be traced back to Duverger (1954), who defined the mass party model as an organizational template embedded into a specific societal pillar (its classe gardée). It followed that any major change in the structure and composition of society was expected to produce a modification in the way parties organized their representative function (Bartolini and Mair 2001). The interpretation of party change as a consequence of contextual factors and the search for synchronous patterns of organizational modifications across liberal-democracies became predominant, laying the foundations of other fundamental contributions in the field. The progressive collapse of traditional social pillars and the weakening of segmental solidarities-a process observed since the 1960s-were considered conducive drivers for mass parties to converge toward the 'catch-all' model (Kirchheimer 1966;Epstein 1980). In this vein, Panebianco's (1988) 'electoral-professional' party; and the 'cartel party' model (Katz andMair 1995, 2009)-which broadened the determinants of party adaptation by including the role of the State and, in the re-statement of the original thesis, the impact of «external factors drawn from the world of international politics and economics»-represent the last accomplished theoretical effort based on ideal types and the contextual approach. Other new 'models' subsequently elaborated-e.g., the business-firm (Hopkin and Paolucci 1999) or entrepreneurial party model (Krouwel 2006), an organizational template that received academic attention in the late Nineties, in parallel with the emergence of a number of 'personal parties' (Calise 2000;2015); or Kitschelt's (2006) 'movement party'-were less successful.

A different perspective
The search for party models has brought to a proliferation of competing labels: Recently, doubts have been raised in terms of models' empirical reliability as well as in their actual ability to catch multiple combinations of party organizational dimensions into adequate categories Katz 2017). Moreover, the contextualist perspective underlying the environmental-induced approach considers political changes as mainly dependent on extra-political factors-whether set at the domestic or supranational level. In our opinion, on the contrary, political systems should be treated as autonomous organizational fields; and political parties-the main organizations competing to control political systems-should be considered less dependent on contextual factors than what is commonly hypothesized. We briefly expose the arguments underlying our theoretical perspective in the following paragraphs.

The autonomy of political systems
Political systems have functional precedence over all the organizational fields and related organizational populations of a polity (Olsen 2009). The autonomy of political systems is underpinned by the ultimate power, wielded by the State, to determine the institutional order of a polity Olsen 1989, 2008) through the monopolistic and legitimized use of coercion over a territory. A political system thus consists of a set of regulative structures that define how the political power is conceived, shared and exercised by different political units (public, semipublic and private organizations), which are legitimately entitled to formulate public coercive decisions that affect all the other organizational fields (Lanzalaco 1995). Given these premises, political systems should be considered more resistant to external pressures than what is commonly expected (Fioretos et al. 2016). Domestic contextual dynamics develop within the institutional framework set by the political institutions and organizations, whose actual 'adaptation' to external pressures should at least be investigated empirically rather than taken for granted. This assumption holds also when focusing on the relationship between national political systems and the broader international environment. The autonomy of national political systems at the times of the globalization process has been debated through different perspectives (Robertson and White 2007). However, both the mainstream Globalist (Strange 1996;Martell 2010) and Sceptical (Robinson 2007) approaches have proven empirically weak (Fjäder 2014). On the contrary, the middle-ground Transformalist approach to globalization-which remarks the major role still retained by national political systems, while recognizing the impacts of Globalization-constitutes a reliable analytical lens. In this view, domestic political institutions and actors actually constitute powerful filters for supranational pressures, by mediating the intensity through which the latter impact internal politics. Domestic political institutions and organizations show a tendency to persist indeed; and path dependence (Pierson 2000(Pierson , 2004 is the normal dynamic characterizing their evolution (Schreyögg and Sydow 2010) even in the presence of undeniable pressures to change.

The peculiarity of political parties
If political systems are autonomous and persistent, the prevailing contextualist perspective that considers parties as merely adaptive actors-both at the functional and organizational level-must be questioned. This interpretation led scholars to underestimate parties' primary and invariant function of institutionalization agencies (Pizzimenti 2020), that is specialized actors that perform the task of reproducing and reforming the political order. As Sartori (2005: 27) put it, «[…] parties are an instrument which shapes the society with some congruity to the needs and the overall properties of the political system». Parties are legitimate and specialized institutionalization agencies since they perform their systemic functions through the circuit of representation (participation-elections-parliament-government), which constitutes the basis of the institutional order of any liberal democracy. In other words, parties exist precisely for competing to access political institutions (von Beyme 1987) and to exercise power over the polity (Thies 2002). In particular, parties that enter the parliament-government arenas hold the legitimate power to shape the broader socioeconomic and cultural context. Clearly, political parties are not the only actors involved in the overall reproduction of the institutional order, as other collective entities (such as interest groups, professions, public opinion and confessional groups) contribute to circulating and renovating shared rationalized myths (Di Maggio and Powell 1991). However, parties are the only actors wielding the legitimate power of formulating coercive decisions. Moreover, political parties are also entitled to regulate their own sub-system, the party system. Parties set the rules that shape political competition; as well as the normative regimes governing their relationships with the State, by disciplining the requisites they need to possess to be formally recognized, to be admitted to the electoral competition, to access public subsidies etc. (Mair 1997). These normative regimes, in turn, have an impact on parties' further development and strategies, since laws and regulations may have unintended consequences even on their promoters.
Concerning party organization, an excessive emphasis on party adaptation to contextual stimuli clashes with organizations' conservative nature, which is a pillar of organizational theory (Selznick 2011;Scott 2013;Giger and Schumacher 2020;Pizzimenti 2020). Like any other organizational form, also parties must be able to respond to external pressures: However, this does not imply that party responses follow the pace of such solicitations, nor the same direction. If that happened-and this is another limit of the contextualist approach-party organizations would potentially be subject to constant changes: a condition that would compromise the stability of their patterns of interaction (which, on the contrary, are assumed by the literature-Mair 1997; Siaroff 2019), as well as their function of institutionalization agencies. Finally, since pressures to adaptation come from different (and often conflicting) sources, parties would find it extremely difficult to select the 'most relevant' ones to conform with.

Party systems' parameters and party organizational variance
Within the political system, the main interactions among parties develop at the party system level (Wolinetz 2006). By privileging the contextualist approach, to date, the literature has paid scarce attention to the political roots of party organizational profiles. In particular, the relationship between the features of the party system and party organizational tendencies deserves to be investigated more in-depth. We maintain that some specific features of the party system-its fragmentation and the patterns of competition among parties-are significantly related to party organizational variance. In fact, fragmented party systems are expected to be characterized by tendencies toward party organizational allomorphism (Leblebici et al. 1991), as parties are interested in keeping/modelling their organizational profile to stand out from each other. At the same time, different patterns of competition-namely the actual contestability of the party system-affect party organizational strategies by creating incentives and/or disincentives toward organizational variance and/or convergence.
Given these premises, we opt for an approach that combines the functional criterion set by Bardi and Mair (2008) for the analysis of the party systems, with the dimensional approach elaborated by Scarrow et al. (2017) for the study of party organizational change. As for the former, party systems are split into three different arenas (electoral, parliamentary and governmental), which work according to different logics. Concerning the dimensional approach, its main focus is on party organizational variance, rather than convergence. Party organizational development follows different timing and dynamics depending on parties. Three clusters of organizational dimensions and related sub-dimensions are considered: party resources (human and financial), representative strategies (parties' prevailing orientation toward an integrated identity or a consumer choice profile) and structures (the subdimensions concerning leadership autonomy to act on behalf of the party as well as the organizational rights and powers accorded to the representative of the party in public office). Comparative empirical analysis is privileged instead of archetypical party models, to identify inter-party similarities and divergences along different organizational dimensions.

Research design and methods
Party system parameters affect both the way parties organize along different dimensions and the degree to which parties vary, in each country. Depending on the country, then, different parameters of the party system are expected to be associated with different organizational responses, at different times. To test our theoretical premises, we want to answer the following research questions: (1) Do parties vary along different dimensions, in each country, in time? (2) To what extent does the party organizational profile vary within countries, in time? (3) To what extent party system-related parameters are able to explain party variance vis-à-vis extra-political factors?
To our aims, we include in our sample seven established Western European liberal-democracies-Austria, Belgium, Germany, the Netherlands, Norway, Sweden and the UK. Our observation points are set at two time markers (approximately the beginning of the 1990s and the 2010s). We select countries according to their electoral system, which is a distinctive feature of any political system: We thus include three countries with a party list proportional representation system (Austria, Belgium and the Netherlands); 1 country with a mixed-member proportional (Germany); 2 countries with a two-tiered party-list proportional representation system (Norway, Sweden); 1 country with a single-member plurality system (the UK).
As for parties, we focus on a total of 38 cases for which comparable organizational data exist at our time markers, by combining and codifying the data provided by the Political Party Database Project  with those published in the Party Organizations Data Handbook (Katz and Mair 1992). In line with the PPDB analytical framework, our coding scheme includes organizational variables pertaining to three dimensions: party Resources (2 variables), Representative Strategies (16 variables) and Structures (8 variables), for a total of 1948 Observations (excluding missing values)-See Appendix.
First, in each country, we measure the level of party organizational variance along the three organizational dimensions mentioned above. For each organizational dimension, we calculate the standard deviation of each variable in the cluster, for each of our party populations (i.e., the parties included in a country), at both our time markers. Then, we calculate three dimensional indexes as a simple mean between the standard deviations of the variables in each cluster, as follows: where i stands for the variables included in each of the dimensions and represents the standard deviation.
Second, we calculate a general index of party organizational variance (POV), per country and period, as a simple mean of the three dimensional indexes, ranging from 0 (no variance) to 1 (complete variance). The POV index is calculated as follows: The analysis will be integrated with detailed information about the trajectories of the values registered by parties along the sub-dimensions analyzed.
Finally, we attempt to shed some light on the possible relationship between party-system related parameters (Siaroff 2019) and party organizations. According to our theoretical premises, for each arena of the party system, we focus on variables that measure the level of fragmentation, as well as the contestability of the party system. Concerning the former, we include electoral fragmentation an indicator that weighs parties by size; seat ratio first to second party, which compares the number of seats of the two largest parties; and the number of minister in cabinet, that indicates the magnitude of the executives. As for contestability, we include the number of parties with 15% of the votes; disproportionality, an indicator that measures the difference between vote shares and seat shares; and  Table 1. We thus compare the predictive capacity of party-system parameters in explaining party organizational variance with that of cultural, socioeconomic, and international factors. We also include a dummy variable 'time' that allows us to use a single regression model to analyze our countries at two different time markers. Cultural factors are derived from the European Value Survey, and are selected based on previous research on the matter (Inglehart 1997;Hofstede et al. 2010): We include variables measuring the level of Interest in Politics (a key marker of Uncertainty Avoidance, Hofstede et al. 2010), the attitudes toward Materialism and the importance assigned to Freedom and Equality (both used, among others, by Inglehart 1997). As for socioeconomic factors, we include GDP pro capita, the average years of schooling and the percentage of workers employed in services: these variables are generally used to analyze modifications in social structures, which are considered among the main predictors of party organizational change. Finally, supranational factors (being a country member of the EU, terms of trade and trade in goods and services) have been considered potentially relevant by Katz and Mair (2009) to explain patterns of party organizational change: Still, there is a lack of systematic empirical analysis on this supposed relationship. The selected variables are those for which data exist at our time markers. To our aim, we run several OLS regressions taking each country's POV as our dependent variable to assess the possible impact of all these factors.  (2019) Party system parameters Relation with POV Electoral fragmentation (EFRG) H1-The higher the fragmentation of the electoral arena, the higher the expected organizational variance Number of parties with 15% of the votes (P15V) H2-The higher the number of parties with 15% of the vote, the higher the expected organizational variance Disproportionality (DISP) H3-The higher the disproportionality of the electoral system, the lower the expected organizational variance Seat ratio first to second party (SR1:2) H4-The higher the seat ratio of the first to the second party, the higher the expected organizational variance

Number of ministers in the government (MINIS-TERS)
H5-The higher the number of ministers in the government, the higher the expected organizational variance Two-party government share (2PGS) H6-The higher the share of government participation of the two biggest parties, the lower the expected organizational variance

Empirical findings
In time, a tendency toward lower variance along each dimension can be observed for Belgian, German, Norwegian and Swedish parties; on the contrary, variance increases concerning Dutch and British parties, while Austrian parties become more dissimilar from each other along two dimensions out of three (Fig. 1). Variance in party Structures shows the highest value in all countries (but the Netherlands and Norway, in 2010), with a mean value of 0.23 compared to 0.12 for Representative Strategies and 0.09 for Resources. However, the trajectories of change differ from case to case. While in the passage from 1990 to 2010 variance significantly decreases among German parties (− 0.18), for Austrian and British parties the increased variability of party Structures is striking (+ 0.27 and + 0.14, respectively). Also, Dutch parties look more divergent along this dimension in 2010: However, values only slightly differ comparing the two time markers (+ 0.02). Concerning Representative Strategies, Belgian and German parties show practically identical values in both 1990 and 2010: The former are also the most stable of the sample in their Resources, a dimension along which Norwegian parties register the sharpest shift in the period (− 0.09). A multifaceted picture emerges when focusing on individual organizational variables. As for resources, almost no variance can be observed among British and German parties concerning the M/V ratio, in time; on the contrary, different tendencies characterize Austrian parties (the most dissimilar population in the sample) and Swedish parties (whose variance dramatically decreases, from 0.17 to 0.02). A clearer tendency toward decreasing party variance concerns the impact of public funds on total party income, except for the Netherlands (where direct subsidies were introduced in 1999) and the UK (where the underlying logic of the legislation tends to reward opposition parties). Coming to party representative strategies, the ways in which parties vary along the 16 analyzed variables are manifold. We registered no variance in both periods for all the parties only among three variables (the recognition of membership as a formal category; membership sponsored by current members; and the exclusion of non-members/supporters in selecting/deciding on candidates). In all the other cases, to cite a few, parties vary to a different extent. Parties show almost no variance concerning the formal provision of membership dues and a limited variance with regard to the (marginal) role of the national secretary in selecting/deciding candidates and the (frequent) inclusion of affiliated organizations in candidate selection procedures. Conversely, high variance can be registered in relation to the possibility for individual members to belong also to other national parties (in particular among Dutch and Swedish parties) and to the (limited) role played by local level organizations in candidate selection procedures. Finally, the highest variance along the variables pertaining to party structures concerns the ex officio presence of the leader of the party group in the lower house; variance is consistent also with regard to the ex officio presence of the prime minister/chancellor in the national executive organ: In both cases, Austrian and British parties record the highest within-country variance.
We now want to assess to what extent party overall organizational profiles vary within countries, in time. As Fig. 2 shows, variance in party organizational profiles follows different trajectories, at different depths. At the beginning of the 1990s, in the Netherlands we find the most convergent organizational population, being on the contrary Belgian parties the most divergent. At this time marker, the overall variance across countries is quite limited, as the mean value of the POV index for the whole sample is 0.15. This value does not change, in time: But while some party populations become more convergent, others show increasing variance. This is particularly true for Austrian parties (with the POV index moving from 0.16 to 0.25), which form the least convergent population in our sample in the 2010s; but a similar tendency can be observed also for Dutch (+ 0.03) and British (+ 0.07) parties. All in all, in the whole period Norwegian parties constitute the least divergent population (POV mean value 0.09); on the contrary, the Belgian and Austrian party populations show the highest mean value of the index (0.23 and 0.21, respectively).
Having ascertained that, depending on the country, party populations show different tendencies along organizational dimensions as well as in their overall organizational profile, we now turn our attention to the possible determinants of such variance. According to our assumptions, we emphasize the role played by the parameters of the party system vis-à-vis other domestic and supranational drivers of an extrapolitical nature, which have been considered potentially relevant by the literature on party organizational change. We run four different OLS models including the POV index as our dependent variable, and party systems related parameters, cultural factors, socioeconomic factors, and supranational factors as independent variables, as well as the dummy variable 'time' 1 (Table 2).
In Model 1, we include the parameters of the party system as independent variables plus the dummy variable 'time.' This model explains 55% of the variability of the POV index once adjusted on the number of regressors and the size of the sample-with a high statistical significance (p value 0.08). In this respect, it is remarkable to notice that the values registered by Model 2 including cultural predictors show a higher capacity to explain the variability of the index (adjusted R 2 = 0.66; p value 0.01). However, the analysis of the relationships between individual party system parameters and the POV index shows that political variables have a stronger predictive power on average, even if our hypotheses are only partially confirmed. In fact, higher party organizational variance is explained by both higher values of electoral fragmentation and by a higher number of parties with at least 15% of votes (with coefficients of 0.07 and 0.05, respectively), in line with hypotheses H1 and H2. A positive relationship between the POV index and the number of ministers in the government is observed in the model, thus confirming also H5. On the contrary, still concerning the governmental arena, higher organizational variance corresponds to a higher share of government participation of the 2 biggest parties, rejecting our sixth hypothesis. Moreover, none of the parameters related to the parliamentary arena (DISP and SR1:2) has a significant capacity to predict POV (p value = 0.14 and 0.91, respectively). All in all, party organizational variance seems mainly related to the fragmentation and the contestability of the electoral arena: It seems thus plausible that in a 'crowded' and competitive electoral arena parties tend to adopt different organizational profiles to distinguish themselves. Interestingly, not even in settings where the two largest parties have a high share of participation in government-thus Table 2 The predictive capacity of party system parameters All continuous predictors are mean-centered and scaled by 1 standard deviation. ***p < 0.001; **p < 0.01; *p < 0.05 suggesting possible patterns of cartelization-organizational convergence prevails. This finding is rather counterintuitive with respect to the consolidated literature, as it suggests that in those countries where the two main parties join more frequently governmental coalitions, organizational variance is higher. Also, the positive relationship between the number of ministers and the POV index seems to substantiate this finding.
Turning to the other models, the positive relationship between Interest in Politics and the POV index suggests that higher citizens' political engagement brings parties to search for alternative organizational profiles: This may constitute a promising avenue for further research, since the link between citizens' interest in politics and organizational analysis has not been explored in-depth yet. Also, the spread of postmaterialist values is positively associated with organizational variance: This finding is in line with the literature dedicated to the emergence of 'new parties' (Harmel 1985;Hug 2001). On the contrary, according to our third model, none of the most frequently used socioeconomic markers shows a significant capacity to predict party organizational variance (p value 0.34), with the only exception of the intercept value: this shortcoming can be probably explained by the limited number of observations included in our sample. The same holds also for our fourth model, for which we can register only a significant positive value of the intercept (p value < 0.001).

Conclusive remarks
The aim of this contribution was twofold: We wanted to verify how and to what extent party organizations vary within countries, in time, as well as to enhance the role of political factors in explaining organizational variance. By adopting a dimensional approach and a diachronic perspective, our empirical analysis focused on 7 European democracies, covering 38 parties. Our findings partly support our theoretical premises.
First, we have shown how party organizational dimensions vary depending on the analyzed party population and in time as well. This finding is consistent with our assumptions, and it helps to corroborate the reliability of the dimensional approach, whose analytical precision is higher compared to party models. Also, the extent to which the overall party organizational profile varies within countries, in time, reveals significant differences in the trajectories followed by each party population.
Second, our findings show that party organizational variance is mainly dependent on the parameters of the party systems related to the electoral and to the governmental arenas, while we are not able to draw conclusions about the predictive capacity of the parliamentary arena. In general, our findings show that to higher fragmentation and party system contestability corresponds a higher organizational variance. However, also cultural factors are able to explain a large share of the POV index, thus suggesting that an integrated perspective combining party system parameters and political culture variables may represent a reliable analytical approach to frame party organizational dynamics. At the same time, socioeconomic and supranational factors included in our analysis have proved not satisfactory to explain party organizational variance. In this regard, further research is needed, by increasing the number of possible predictors-albeit difficulties in raising comparable data should not be underestimated.
To conclude, it is our opinion that the specialized literature has overestimated the relevance of extra-political factors for interpreting and explaining party organizational dynamics, which have mainly been framed as a simple adaptive process. On the contrary, in line with Sartori (2005), we maintain that political phenomena seek for political explanations. We are persuaded that this contribution may induce scholars to avoid simply assuming party organizational convergence, by bringing the relevance of variance and domestic party systems' peculiarities back in.

Political variables (Siaroff 2019 and authors' own elaborations)
EFRG. Electoral fragmentation. P15V. Number of parties with 15% of the votes. DISP. Disproportionality. SR1:2. Seat ratio first to second party. MINISTERS. Number of ministers in the government. 2PGS-Two-party government share.

Supranational variables (OECD and authors' own elaborations)
Trade in goods and services; Terms of trade; Number of years since the Country has become member of the EU.

Socioeconomic and demographic variables (world bank data and de la fuente and domenech 2012)
GDP per capita (2010 US$); Employment in services (%of total employment); Average years of schooling. Party organizational variance: values per country, period and dimension. Funding Open access funding provided by UniversitÃ di Pisa within the CRUI-CARE Agreement.
Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article's Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article's Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http:// creat iveco mmons. org/ licen ses/ by/4. 0/.