Who's at the helm? When party organization matters for party strategy

1 The included countries are Austria, Belgium, Denmark, Finland, France, Germany, Ireland, the Netherlands, Sweden, and the United Kingdom.

2 The mean and median voter position are used interchangeably here. For their measurement, see below.

3 By contrast, several studies found little empirical evidence for the notion that parties respond to past election results (see Adams et al. Citation2004; Ezrow et al. Citation2011).

4 It is important to distinguish this conceptualization of factionalism from the way in which it has been defined by Budge, Ezrow, and McDonald (Citation2010), among others. While both accounts depart from the assumption that parties are unitary actors, rid of internal division, the latter perspective views parties as caucuses of ideological factions competing for dominance. Based on that assumption, the authors explain why, after suffering electoral losses, it may lead a party to reverse its policy platform. That is, an opposing faction could seize control of the party and implement a strategy in line with its own, alternative ideological goals. While equally valid, the focus in this paper is on the alleged divide between party leaders and activists, not the internal competition between different ideological camps.

5 Reduced political choice in the context of party system convergence has even been found to explain the decline of class voting (Evans and Tilley Citation2012a, Citation2012b).

6 It is important to note that not all party members are activists per se, as the degree to which members are active within the party is best described on a continuum (Panebianco Citation1988). However, given the stated importance of ideological congruence and the simultaneous decline in party activism and membership (Kölln and Polk Citation2017; Pedersen et al. Citation2004; Whiteley Citation2011; Whiteley and Seyd Citation1998), it is highly likely that activist exit does occur (see, for instance, Heidar and Saglie Citation2003).

7 The decision to use the ‘mean’ instead of the ‘median’ voter position is a methodological one; the Eurobarometer indicator employed here is a discrete variable, which results in a lack of variation in the position of the median voter.

8 To test the assumption that a party's organizational structure is, indeed, relatively stable, I compare the information from Laver and Hunt (Citation1992) to a more recent measure by Rohrschneider and Whitefield (Citation2012). Not only is the correlation between these two indicators fairly strong (0.7), a party's internal balance of power in the early 1990s is a statistically significant predictor of its score in the 2000s. Furthermore, three of the five parties with the highest and lowest party organization scores are the same across both surveys – VB, PLP (both Belgium), and the Conservative Party (United Kingdom) are among the most leadership-dominated parties; Agalev, Ecolo (both Belgium), and Grüne (Germany) are among the most activist-dominated parties. Studies by Schumacher, De Vries, and Vis (Citation2013) and Schumacher and Giger (Citation2017b) have also found similar levels of cross-measure reliability.

9 These robustness checks include, among others, a logit transformation of the MARPOR positional data (Lowe et al. Citation2011), a measure of a party's electoral performance based on seat share instead of vote share, and alternative party system indicators, such as Dalton's polarization index (Citation2008) and degree of party system fragmentation.

10 Table A1 presents the correlation matrix for all included variables (see Appendix A).

11 Activist and leadership-dominated parties are defined by the 10th and 90th percentile of the party organization variable (10, 25), respectively.

12 The correlation between the change in vote and seat share is strong (0.81). The only exceptions are France (0.68) and the United Kingdom (0.75), which have more disproportional, plurality-based electoral systems. The correlation between the two indicators for the other eight countries in my dataset is 0.95. The results shown here are based on the latter subset of countries.