Party membership has substantially changed in Western countries over the last decades. While the figures of party members have steadily declined in most countries (Biezen et al. 2012; Biezen and Poguntke 2014; van Haute and Gauja 2015; Whiteley 2011), we don’t have many studies indicating this is also the trend among party activists. This is explained mainly because of the lack of longitudinal data on party activism (Bale, Webb, and Poletti 2020; Whiteley 2011). Recent studies on British and Scandinavian parties point towards an increase in the number of members that consider themselves active (Bale et al. 2020:99; Demker, Heidar, and Kosiara-Pedersen 2020). In parallel to this, most parties have introduced new, more individualized and multiple forms of party attachment in order to face the general decline of traditional party members (i.e., the archetypical mass party ‘militant’) (Duverger 1954). These new forms of party attachment have been accompanied with the creation of a superbase of digital supporters and affiliates, and the establishment of digital mechanisms for communication between members and parties. Building on the multi-speed party model, these new trends of party attachments and activism have been recently theorized in the comparative party literature (Gerbaudo 2019; van Haute 2009; van Haute and Gauja 2015; Scarrow 2014).

Interestingly, while we know most members of political parties are not very active in the organization (Bale et al. 2020), activists are still key in their functioning. Active members can be a resource. They can help to mobilize members, campaign, run as candidates, hold party offices, or contribute to the daily life of the party organization. Political parties, on their part, are making efforts to engage activists in the decision-making procedures, reinforcing the idea of active members being relevant to sustaining the party organization. Members regularly attending their local party branches and their parties’ conferences have better chances to influence party decisions than those not having an active role in the party, even if parties also introduce new and more individualized forms of decision making (i.e., party leader and candidate selection, etc.) (Duverger 1954; Scarrow 1996; Ware 1996). In line with previous studies, we define activism as the activities party members engage in and are organized by or for the party (Demker et al. 2020:175). As such, party activists are those party members dedicating time to engage in these activities. Previous studies have also defined activists as those attending party conferences, who often represented the middle-level elite of the party and acted as the intermediary between the voters and rank-and-file members and the decision-makers in the party (van Haute and Gauja 2015; Whiteley et al. 1994:79).

The literature has explored who are the active members in political parties (van Haute and Gauja 2015), but not much attention has been devoted to variations within party activists (but see Demker et al. 2020; Heidar 1994). Overall, the focus has often been on the reasons of entry and exit, and the factors driving participation (see for instance the work of Whiteley and Seyd on the British case). Relevant questions linked to how homogeneous activists are, what kind of activities they engage in, and to what extent the existence of different profiles might be a challenge for internal party politics remain mainly unanswered. Our article explores some of these questions through the evidence provided by the Spanish case. In this regard, Spain is an excellent case study to understand party activism patterns due to the high variety of political parties that have achieved representation in the country’s multi-level institutional setting.

This paper has a double goal. First, it aims to categorize different types of party activists in Spain. More precisely, by applying cluster analysis we adopt an inductive approach to identify the different profiles of party activism based on the roles and activities performed instead of theoretically determining those patterns. Second, we analyse some factors that could help to explain the variance of these profiles among parties and individuals. In particular, we test the influence of personal traits, individual participation and incentives, combined with party characteristics such as ideology, organization, or government access, among others.

We use a dataset of Spanish party activists based on survey data from some of the most relevant political parties represented at national and regional level. The surveys were conducted between 2008 and 2017 following intensive fieldwork at several party conferences. Our findings point first to the existence of multiple patterns of party activism and the coexistence of both new and traditional modes of activism in Spanish political parties.

The article is organized as follows. We first introduce our overview of the literature and theoretical expectations. After explaining the data and methods, we then present the results of the cluster analysis and explain the variation across parties. We conclude with a summary of our findings and avenues for future research.

Party scholars studying party members and activists have often focused on three main aspects: who they are, why they are active, and what they do in the party (Bale et al. 2020; van Haute and Gauja 2015; Heidar 2006). To identify party activists, most studies first define party members and then select the subgroup considered as activists. Party members are generally defined by their formal attachment to the political party through the payment of a membership fee or by their behaviour, for instance, through voter loyalty or participation in party primaries (van Haute and Gauja 2015; Hazan and Rahat 2010; Heidar 1994; Panebianco 1988; Ponce and Scarrow 2014; Scarrow 2014). Party members are considered activists if they engage in the activities organized by and for the party (Demker et al. 2020; van Haute and Gauja 2015; Heidar 1994; Ponce and Scarrow 2014). The literature so far agrees that party membership is conformed predominantly by individuals who are more resourceful than the general population, while being at the same time quite a heterogeneous group (Heidar 2006). In contrast, results on party activists observe much less variety, suggesting they represent a more homogeneous group, and indicating they tend to be middle-aged individuals, male and with high levels of education (Dommett, Temple, and Seyd 2021; van Haute and Gauja 2015; Power and Dommett 2020)

To comprehend why members become active and, thus, become activists, party scholars have turned to models developed in order to understand why individuals join political parties in the first place. Previous research has applied and tested the explanatory capacity of models such as the resource model, the social-psychological model, the rational choice and the general incentives model, with the latter being the most used lately (Clark and Wilson 1961; Verba and Nie 1972; Verba, Schlozman, and Brady 1993, 1995; Whiteley et al. 1994; Whiteley and Seyd 2002). In that sense, most recent studies seem to indicate the general incentives model is the one with better explanatory capacity (Bale et al. 2020; Demker et al. 2020; Webb, Bale, and Poletti 2020). These studies show that the socioeconomic status of individuals and the presence of purposive incentives play a role in party activism. Additionally, other factors such as the party culture and solidary and material incentives also matter. Accordingly, these studies seem to suggest that both individual and party factors matter when explaining party activism (Heidar 1994, 2006; Seyd and Whiteley 2004; Whiteley and Seyd 1998, 2002).

Regarding what party activists do, studies mainly focus on both the activities and the intensity of their participation. These studies build on the hierarchical approach inspired by Duverger’s theory of concentric circles, which ranked members according to the extent and the quality of their involvement within the party (Duverger, 1954, 120). They assess both the intensity and quality of activism in parties and identify different types of members accordingly. Seyd and Whiteley (2002), in their study on Labour’s grassroots, distinguished three main areas of party activity and suggested key dimensions for activism, which helped to discern the participation of rank-and-file members from activists. For instance, attendance to party meetings was highlighted as a relevant indicator of activism, and the whole dimension of representation was also considered as a good indicator since it required high-intensity participation (Seyd and Whiteley 2002). Heidar (1994) distinguished between participating in internal activities which included attending branch meetings, seminars, etc., and external activities such as recruitment, delivering party leaflets, etc.

In the study of the Norwegian party membership, Heidar (1994) identified different patterns of participation and established different categories. While the names of these types have varied across time, some remain stable (Heidar 2014). In his 1994 study, Heidar identified the shop stewards, the veterans and the financiers. The shop stewards pattern corresponded to regular engagement in meetings and party debates, to some extent linked to the traditional mass-membership party activist idea. The veterans were those that had previously held office, and, finally, the financiers were those that only paid the party fee or gave donations (Heidar 1994). In subsequent studies, Heidar analysed changes between 1991 and 2009. Four patterns emerged; three of them had appeared in previous studies but with slightly different names: the party worker (formerly, the shop steward), the old guard (formerly, the veterans) and the credit card membership (formerly, the financiers). Overall, his work shows a picture of stability in the predominant types of party activism and, along with other research from the UK, provides empirical support for the polymorphic nature of party membership (Heidar 1994; Whiteley and Seyd 2002). Comparing Scandinavian countries, Demker et al. (2020) also show a similar picture of patterns of activism with the presence of the party worker, the veterans, and the party ambassador (or outreach category). The party worker was the most predominant category across parties, and has been considered the equivalent of the party activist (Demker et al. 2020:165-69). Demker’s study also highlighted the existence of new forms of online activism not always detected in previous research of party members, often due to lack of data (but see Heidar and Saglie 2003). Their findings are indeed connected to the emerging literature on this area illustrating how parties are now combining online and traditional forms of membership (Vittori 2020), and more particularly on Scarrow’s theory of multi-speed membership (Scarrow 2014). In her book, she classified different types of members and activists according to their level of attachment to the party and included their use of digital technologies too. Overall, these studies have identified different patterns of participation among party members and have often considered activists as one of them, but have not paid much attention to the potential variation within activists.

Building on this tradition, we argue that party activism will vary within parties to generate different profiles of activists. Scarrow (2014) pointed out that members can move from one level of endorsement and engagement to another (i.e., moving from member to activist to member again). This suggests that if members can change the intensity of their activism or the activities they focus on, the nature of party activism might also be polymorphic, meaning we might encounter more horizontal variation within activists than the literature has so far acknowledged. While the previous literature has considered office holders (current and past) and party workers as the party activist core (Bale et al. 2020; Demker et al. 2020), the theory of multi-speed party membership suggests the possible existence of different activists that can contribute differently to the party. Indeed, evolving party organizations will require other kind of members and activists to satisfy their needs, accomplish their goals and perform their functions. Previous studies suggest party members modify their intensity of participation following either their interests or the interests of their parties (Scarrow 1996, 2014). Other studies indicate that parties providing different costs and benefits will develop different levels of party activism (Kosiara-Pedersen, Scarrow, and van Haute 2017). Accordingly, we argue that political parties can have party activists with similar levels of high-intensity participation but focusing on distinct areas of party activity. We further expect individual factors, such as resources or incentives for participation, and party factors such as the organizational structure, to influence the coexistence of different types of party activists.

Building on the incentives model, the literature has already established that both socioeconomic factors and the reasons for joining political parties are also relevant to explaining different levels of activism (Heidar 1994, 2006; Seyd and Whiteley 2004; Whiteley and Seyd 1998, 2002). In addition, studies on political participation indicate that members with higher socioeconomic status and resources are better equipped to participate in politics (Brady, Verba, and Schlozman 1995; Verba et al. 1995; Weber 2020). Hence, we expect activists with different levels of resources to follow different activist profiles. Concretely, we expect activists with higher levels of resources to engage in areas that require more time and civic skills, while activists with lower resources should be active in less demanding activities.

Continuing with individual factors, we also expect the incentives for joining a party to influence the type of activities activists focus on, allowing for different profiles of activists. The literature has already shown the relevance of purposive incentives for activism indicating this type of incentive is widespread among activists (see for instance Bale et al. 2020). In that sense, while purposive incentives might be relevant for crossing the threshold to become active, accomplishing relevant policy goals should not necessarily be linked to one type of activity, since most types of activism will contribute, directly or indirectly, to this. As such, we do not expect activists engaging in different types of activism to be influenced by purposive incentives but rather by material or solidary incentives. In this regard, we expect activists joining for material incentives to follow patterns of activism that include both internal and external activities and require more integration within the party organization, which might facilitate their access to office. On the other hand, we expect activists joining because of solidary incentives to focus on activities that embed a social dimension and enhance camaraderie and social contacts.

Moving on to party-level factors, the presence of different types of activists might be beneficial for different kinds of organizations. For example, some parties might be better equipped to take advantage of multiples types of activists than others. This might even be true for the same party in different stages of its organizational evolution. Building on life-cycle approaches that expect parties to face different challenges, and pursue different organizational strategies at different organizational stages (Van Biezen 2005; Bolleyer and Bytzek 2013; Kölln 2016), we argue the profiles of party activists will differ for new and old parties. We expect new parties that need to build an organizational structure might have fewer resources and have fewer activists and, as such, to encourage a type of activist engaging in all areas of party activity, not necessarily specialized in either internal or external activities. On the other hand, older parties having more institutionalized organizations, maybe a higher number of activists, and a more diversified structure of resources can, as such, afford having more specialized activists. Similarly, political parties with access to government might benefit more from activists focused on both internal and external activities to sustain their position and the functioning of their organization, while political parties in opposition might benefit more from a higher focus on external activities that places them in a better position to win the next elections.

Finally, we also expect the multi-level nature of the party system to influence the types of activists. One of the characteristics of multi-level systems is the presence of statewide parties (SWPs) which develop a multi-level structure and compete in most of (or all) the regions and non-statewide parties (NSWPs) operating only in one or a few regions (Deschouwer 2003, 2006; Fabre and Swenden 2013; Swenden and Maddens 2009). One of the key challenges faced by SWPs is the need to adapt their organization to different political contexts (Fabre and Swenden 2013). While NSWPs might also face this challenge, they tend to operate in less heterogeneous contexts in comparison. As such, the organizational needs of these two types of political parties, as well as the patterns of interactions between their activists will differ, leading us to expect a different distribution of patterns of activists in statewide and non-statewide parties.

There are several strategies to identify party activists, i.e. party members with high-intensity activism. On the one hand, some studies employ indirect, subjective measures of party activism, based on general opinion polls or, alternatively, through surveys to party members. In opinion polls, individuals are asked to define themselves as party activists, as is the case of the European Social Survey. In the second case, members can be classified according to their degree of activism, captured through different variables (time devoted, specific activities, etc).

On the other hand, one can target a particular sub-population of members by observing those party areas with more intense involvement. In this respect, previous studies have usually selected those members attending party ordinary conferences as a specific object of research. These party delegates are not representative of the mean party member but their profile is, instead, close to the more active members of the organisation, especially at the territorial level. This is why scholars used to consider them as party ‘middle-level elites’ (Reif, Cayrol and Niedermayer, 1980; Pierre, 1986; Reif, Niedermayer and Schmitt, 1986). Although the population of party activists might be larger than the group of party delegates, this strategy allows us to observe a homogenous group to identify the horizontal diversity of patterns of party activism, something that is more difficult to analyse with the alternative strategies.

For instance, a survey among party members in the UK found that one third of the individuals did not devote one single hour during the 5/7 weeks of election campaign -the most intense period in party life (Bale, Webb, Poletti, 2020: 98). Besides, similar recent surveys of Danish and Swedish party members found that these individuals devoted around 4 hours per month on average to party work (Heidar and Kosiara-Pedersen 2019: 142-3). In contrast, the average number of hours per week devoted to party work by those attending party conferences in our sample is 14h, and only 7.5% of our sample spend less than an hour per week in ordinary periods. In that sense, members attending party conferences show an intense profile of internal activism. Furthermore, Spain presents an ideal case to test the horizontal variation within party activism. On the one hand, it has long-standing statewide parties such as the PSOE or the PP, but also new ones as Podemos or Ciudadanos that made their breakthrough at national level after the mid-2010s. On the other hand, Spain has a large variety of new and old non-statewide political parties represented both at the regional and national level and regularly celebrating party conferences. Hence, this allows for providing variance and identifying patterns through several features such as their left-right positions, their newness, their organizational dimension, or their territorial spread in a similar environment.

Party conferences are an important event in Spanish parties’ life. With the exception of some extraordinary conferences (which are not part of our target), these collective bodies decide on key aspects for political parties: delegates elect party officers, discuss and approve the party manifestos with the main ideas and proposals for the subsequent years, and set the framework for the strategic decisions until the next conference. They are usually organised in a representative basis, whereby delegates are elected by local branches in proportion to their affiliated members. However, some parties may decide to hold an ‘open conference’ or assembly (adopting formally a rate of 1 delegate per each member), where any affiliated member is able to participate. In our sample, that was the case of ERC and BNG. In these particular cases, the two sub-samples are interesting cases of self-selected membership surveys, with the possibility of a selection bias, as only those more active members in general actually participated in the event.

Gathering this kind of data is a very costly strategy faced with many difficulties (for instance, party organizations should be collaborative along the process). The collection of this kind of data entails three types of methodological challenges: dealing with the actual statistic representation of the respondents, the reliability of the data, and the comparison of the final data. In an attempt to curb these problems, in some cases we have been able to compare the distribution of basic features of the sample (gender, regional origin) with those provided by the organization for all the attendants, confirming very few significant deviations. Additionally, we provide the estimated sampling error, to give a perspective of the reliability of the data employed (Table A2). Notwithstanding these limitations, the empirical evidence on party delegates contributes significantly to a better understanding of the opinions of those individuals sustaining the ordinary life of political parties and their strategic decisions.

We build our analysis upon a large dataset gathered among party delegates attending party conferences in several Spanish political parties from 2008 to 2017. This is a unique collection of data on party activists’ attitudes and profiles, which allows us to have an overview of the main features of those members at the core of Spanish party organizations (local, regional and national). The dataset contains information for 23 conference surveys conducted to 15 different parties (see Table A1), resulting in 8,340 cases throughout a period of 9 years. In some parties, such as PP or PSOE, we have conducted more than one survey, and we have also gathered data from regional-branch conferences. On average, it means 556 cases per party and 362 per conference, although the real number of responses for each case has highly fluctuated. As shown in Table A2, response rates range from 10.1 to 67.6 per cent, with 29.3 per cent on average. In this respect, we may estimate a random sampling error in the results between 2 and 9 with a 95% confidence interval for each survey.

Our empirical analysis combines inductive and deductive approaches. As a first step, we have conducted a cluster analysis, based on the type of activities activists are doing within parties, seeking to observe patterns of party activists related to the activities performed. To measure varieties of activism and activists within political parties, two main techniques have been employed by previous studies. One approach used principal-component or factor analysis aiming to identify the main underlying correlates of activism among party members (Heidar & Kosiara-Pedersen 2019). It identified different patterns of activism within political parties. Alternatively, cluster analysis seeks to segregate individuals according to their prevalent patterns of activism, as it has been used by Gómez & Ramiro (2017), or Bale, Webb and Poletti (2020) for ideological types. Hence, while factor analysis groups types of activism (where one individual could be related to different types), cluster analysis identifies types of activists, with mutually exclusive categories.

Cluster analysis employs the Euclidean distances between individuals to identify homogenous groups of activists. After following the Ward’s Linkage method to produce hierarchical clustering, we adopt the Bayesian criterion (BIC) to identify the appropriate number of clusters. We finally selected four of them according to BIC scores (> 0.5), as including more clusters would have only marginally improved the BIC score. The outcome was three distinctive small clusters of activists, and a fourth larger one. In these categories, individuals have been clustered according to the combination of the types of activities and their specific intensity in each task. One individual may have intense dedication to one type of activity but lower in other types. Hence, the clusters do not segregate individuals on their degree of intensity of party work overall, but on the specific combination of activities and intensity deployed.

To conduct the cluster analysis, we have employed a battery of indicators from our dataset measuring how often delegates fulfil different tasks and activities within the party organization. According to the previous literature, we have distinguished three different types of activities: organizational involvement, electoral campaign, and social activities. Table 1 shows the different extent of involvement for each of the seven indicators. As we can see, party delegates are very active members of the organization, and participate in a different range of activities and tasks ranging from local party life to electioneering and social dissemination of their political ideas. Nevertheless, there is also an important fluctuation in the fulfilment of these activities. They tend to participate more in the local events and meetings, and are highly involved in electoral campaigns, as the plurality of the surveyed delegates do these activities very often. On the contrary, their politically-oriented social involvement reflects more variance. The distribution of the different items already suggests that we may find different patterns of combination of these activities, even if we are dealing with the most politically committed rank of the party membership.

In a second step, we have employed multivariable regression analyses to test the effect of those variables mentioned in the theoretical framework on the chances of becoming one type of activist. Since our dependent variable is a categorical variable, which uses as category of reference the fourth type identified in the clusters, we run a multinomial logistic regression. The main model includes two main sets of independent variables. On the one hand, we test individual factors related to the individual traits (gender, age, education, and professional status), incentives (material, purposive and solidary) and participation (hours devoted to intra party activities). On the other hand, we include variables capturing types of political parties to test our expectations about how party differences influence the diversity of horizontal party activism. For this purpose, we include four dummy indicators identifying left/right, non-statewide parties/statewide parties, old/new parties, and government or opposition status. We also control by parties with open conferences, where all members were entitled to attend.

The previous section has detailed the main indicators and the procedure followed to conduct the cluster analysis. This section will focus on describing the results of the cluster, portraying the key features of the four different groups derived from the Spanish party delegates dataset.

According to our general expectation, the results of the cluster analysis illustrate different modes of involvement amongst members that hold similar positions within the organization and have a similar level of participation, demonstrating the existence of horizontal variation in party activism (see Table 2). The key features of the four groups identified are mostly consistent with previous research highlighting the relationship between types of activists and the type of party organization (Duverger 1954; Scarrow 2014). The first group is the one that we have labelled as mass party activists (cluster 1), which represents one out of five members of the sample. They perform party tasks with high levels of intensity in both the electoral (external), the organizational and the socializing activities (internal) as they were reported by the classic literature on political parties during the first part of the 20th Century (Duverger 1954; Kriegel 1970; Michels 1949). The committed activists (cluster 2) share most of their key features with the mass party activists, but their level of intensity is lower and, more importantly, they are more reluctant to engage in socializing activities. They represent a first evolution of the mass party activist type still concerned with all the organizational and electoral face of party politics, but more disconnected with political parties’ links to civil society. For instance, while all of them usually meet with other party members to discuss politics (and 35% meet other members to do other activities), most of them rarely do the same with non-partisan people.

The canvasser activists (cluster 3) embody a second break with mass party politics. This group predominantly engages with electoral activities and is less committed to other regular party events. In particular, they are more clearly disconnected from socializing activities, which they perform very infrequently. This cluster accounts only for 10% of the sample. Finally, the cheering activists (cluster 4) is the larger group -more than half of the individuals- and the one that presents more singularities. In contrast with the mass party activist, they become involved in party activities with much less frequency. It might be understood as a further evolution of the canvasser activists as they share with them similar preferences for the electoral activity, but with less intensity. Nonetheless, they are more likely to socialize with other party members than canvasser activists, and tend to discuss politics with non-partisan members more frequently than committed activists. Our point is that this fourth cluster represents a distinctive category of party activists, appearing closer to a somewhat active rank-and-file member than the previous clusters of party activists. In that sense, they tend to participate with a lower level of responsibility in party conferences and they might be closely aligned with the party elite.

Overall, we have identified these four clusters of party activist according to their different involvement in the internal and external party activities. These clusters indicate distinct patterns of party activism in their nature and in their intensity. To check these differences, we can observe the extent of heterogeneity between the clusters with other indicators of political and social involvement. First, the four types of activists show different intensity in terms of time devoted to these activities, as shown in Figure 1. Although all the clusters include an important proportion of members with high-intensive participation (at least half of the members in each type spend more than five hours per week in party activities), they differ with the limits of this participation. While mass party activists (cluster 1) are those with more time-consuming involvement (four out of ten members spend more than ten hours per week), the other clusters reduce slightly their participation. Committed activists (cluster 2) spend around eight hours and a half, while canvasser activists (cluster 3) employ seven hours per week. The cheering activists (cluster 4) have a similar average of hours per week but with wider standard deviation, which suggests more internal variance of involvement among these individuals.

Another important difference between clusters is their level of responsibility in terms of offices (Figure 2). On the one hand, clusters 1, 2 and 3 have a similar proportion of party officers (seven out of ten), while cluster 4 activists are more divided in this issue with only 40% of them having formal party responsibilities. On the other hand, even if most of the party delegates of each cluster did not hold public offices at the time of the survey, cluster 1 and 2 have a higher proportion of public officials (almost half of them) in contrast with cluster 3 (one third) and cluster 4 (just one fifth). Again, this suggests more heterogeneity within cluster 4 amongst officers and non-officers.

Finally, we also found meaningful differences between clusters according to their extent of involvement in producing linkage with other party members and non-partisan individuals. We can better understand these differences through their level of participation in voluntary associations in different issues, as has already been positively tested among Scandinavian party members (Jupskås, Kosiara-Pedersen 2020). To observe their social linkage, we measured the number of voluntary associations where party activists declared current affiliation, among a list of 13 different types of policy organizations (youth, women, economic, unions, cultural, sports…). As we expected, the differences between clusters become narrower. All clusters have very high levels of social linkage, since nine out of ten members in each group belong to at least one or two associations. However, members in cluster 1 tend to get involved in more associations (2.6 on average) than the other clusters (2.3 on average) ¹ . In this respect, cluster 2 and 3 seem to be more different from cluster 1 and quite similar to cluster 4.

In the previous section, we have discussed the main features of each cluster. As indicated in our theoretical section, we expect these clusters to differ regarding different individual and party features. As stated in the methodological section, our dataset is composed of party surveys of unequal weight and features (see Table A2). Hence, we consider that the interaction between political parties and clusters needs to be accounted for. These differences could be the result of the variety of rules and composition of party conferences, the turnout of the party conferences and the surveys, or could also respond to other factors linked to how different types of parties might benefit from different types of activists. The next section will test the probability for party delegates to fall into one of the four clusters that emerged from the previous section.

The coefficients of the multinomial logistic regression model are presented in Table 3 (see also Table A.4 for the odds ratio obtained in logit models explaining each cluster). We have selected cluster 4 (the cheering activist) as the base category of reference because it was the most distinctive type resulting from the cluster analysis, as explained in the previous section. The general results of the model have been significantly tested. We have employed different specifications of the model (changing the category of reference, transforming the dependent variable to run logit analysis, and comparing the results of more simple models, dropping some of the predictors of the general model shown in Table 3) to check the robustness of the results.

Starting with the influence of individual traits, we find a positive effect of being female on the chances of belonging to clusters 1, 2 and 3 in comparison to cluster 4. This positive effect has not been clearly defined in previous works. Although recent findings suggest a distinctive effect of gender on party activism, the empirical evidence is inconclusive, since it may have weak (Bale, Webb, Poletti, 2020: 106) or different directions for different tasks (Jupskås, Kosiara-Pedersen 2020). Concretely, our estimates indicate that women tend to be more present in those clusters with more clear-cut levels of participation than men are, while the odds of being a cheering activist are reduced by 20 % compared to men.

Moving on to individual resources, previous studies have also shown contradictory results when checking the traditional assumption that more affluent individuals would be more available for party activities. Our model goes in the same inconclusive direction. Education seems to oppose that expectation (graduates have 58 % higher odds of being cheering activists and 37 % lower chances of being mass party activists). Similarly, our results show that unemployed activists (in comparison with employed ones) are more likely to be mass party activists. These findings would reject the resources hypothesis. On the contrary, the odds for students to be cheering activists increase by 65 % in contrast to clusters 2 and 3, which would be in line with the positive influence of individual resources on more demanding types of activism. In sum, far from being a homogeneous group, formed by middle-aged individuals, male and with high levels of education, clusters of party activists differ in some social traits but without offering clear evidence for the resources hypothesis.

More interestingly, the incentives for joining the party play a distinct role among party activists in contrast with what studies on general members have suggested so far. As we expected, purposive incentives related to ideology are far less important for explaining horizontal variation among high-intensive activists. While purposive incentives might be very important to explain the reasons for joining a political party or crossing the threshold of participation (Bale et al. 2020; Demker et al. 2020), when it comes to determining the type of activism activists engage in, material and solidary incentives have a greater impact. The lack of statistical significance of purposive incentives is robust across different specifications of the model. On the contrary, ambition and social interactions are more important for explaining differences, since those individuals in more demanding clusters are also more likely to express material and solidary incentives. In particular, material incentives increase by 20% the odds of being classified in clusters 1 and 2, while the odds for being a canvasser are 26% lower. Similarly, delegates indicating solidary incentives have around 20 % lower odds of being among canvasser and cheering activists.

Our third general expectation stressed the importance of political parties to explain variation among types of activism (see Tables A.1 and A.3 for descriptive information). In particular, contrary to our expectations of new parties including a lower degree of variation and benefit from activists engaging in all sorts of activities such as those from cluster 1 and 2, our results show new parties tend to promote a less engaged or discontinuous pattern of activism. At the same time, older parties seem to be the reservoir of old mass party activists. Concretely, in old parties the chances for delegates to be a type 1 increase by 60% while in new parties the odds of delegates being cheering activists are raised by the same proportion. In the same vein, parties in government (during the time of the conference) are more likely to have more traditionally engaged activists, as this factor increases the odds of being classified in cluster 1 (by 22%) and cluster 2 (by 54%), while it reduces by 40% the chances of being a cheering activist.

One clear pattern emerges when comparing the territorial dimension of political parties. Statewide parties have higher odds for clusters 1 (+56%) and 2 (+33%), while non-statewide parties more than double the odds of being cheering activists. Hence, we can observe how parties such as PSOE, PP or IU present a higher percentage of mass party activists and committed activists than the rest of the parties. This suggests the ties to traditional forms of activism are stronger for activists in statewide parties than in non-statewide parties. The only statewide party that follows a different pattern is Ciudadanos, which follows the same pattern as non-statewide parties. It is important to point out that in 2011 (when we surveyed the party) Ciudadanos was mainly operating in Catalonia and was then starting to increase its territorial coverage. Overall, the size of the party and the inclusiveness of the selection procedure for attending party conferences of statewide parties make it more difficult for activists less involved in party life to attend this type of events. On the contrary, in some non-statewide parties with less restrictive selection procedures for attending conferences and operating in more homogeneous political scenarios, the presence of cheering activists is quite significant, as they represent over 60% of their delegates (like the Catalans ERC and UDC).

Moreover, we have controlled for parties with more inclusive conferences, where all delegates could participate because the representative ratio was 1 (each party member could be a delegate). Accordingly, some of the parties with higher proportion of cluster 4 held open conferences where all the grassroots members could participate, for example BNG, ERC and UDC. A simple model regressing the clusters on this variable confirmed the positive effect. However, when we introduce the other explanatory variables in the equation, the effect of open conferences becomes negative, which indicates the chances of belonging to cluster 4 are mainly explained by other individual or party factors.

While previous studies on the drivers of party activism have often focused on the difference between different types of members following a hierarchical approach, our study goes beyond this and presents a more nuanced categorization of party activists’ profiles focusing on the horizontal variation within activists. Our paper goal was twofold, we aimed first to identify different types of party activists within political parties and, second, we aimed to explore the potential explanatory factors of those different types. The results of our cluster analysis, in the first part of our empirics, provides evidence of horizontal variation among party activists and sustains our thesis of the polymorphic nature of party activism. Concretely, our analysis shows the coexistence of four different types of party activists within Spanish parties: the mass party activist, the committed activist, the canvasser activist and the cheering activist. These four activist types vary in the intensity and the quality of their participation. While some are very active in organizational matters, others mainly focus their energies on electoral activities. Indeed, while the mass party activist is very active in all aspects of the party life (internal and external); the committed activist is very active when it comes to organizational and electoral activities but socializes less regularly around the party. The canvasser activists, on the other hand, dedicate most of their time to electoral activities, with social and organizational activities being secondary; and, finally, the cheering activists become involved in party activities with much less frequency but are more likely to socialize with other party members than other activists.

Furthermore, our multinomial regression model sustains previous findings in the literature on the relevance of individual and party factors to understand party activism (Heidar 1994, 2006; Seyd and Whiteley 2004; Whiteley and Seyd 1998, 2002). Concretely, our results show how individual resources such as professional status or individual traits such as gender are linked to the likelihood of becoming one type of activist or another. But they do not provide strong evidence for the resources hypothesis. Importantly, our findings suggest more attention should be paid to material and solidary incentives to understand horizontal variation among party activism. The latter is crucial for political parties searching to promote a particular type of activist over others, since focusing on more targeted activities could encourage canvassers or promote jack-of-all-trades activists engaging in both internal and external activities similar to the mass party activist. Similarly, the type of political party also matters in understanding the activist configuration enhanced by the party leadership. The age of the party, their access to the institutions or their territorial coverage matter for predicting the chances of developing different types of party activists within their organizations. In that sense, our findings are important for gaining a better understanding of party activism and party organization since we study those individuals sustaining the ordinary life of political parties and their strategic decisions. The role of members is still key to understanding how political parties behave and evolve. Learning about the different types of activists existing in political parties will provide a better indication of the type of organization leaders encourage or the evolution we could expect in the future.

Even though our findings draw on the Spanish case, we believe our results could be generalizable to other European countries. European parties have followed a similar evolution regarding party organization and party activism, as the party literature has widely explored (van Haute and Gauja 2015; Whiteley 2011). We should expect the four types of activists to be present in all political parties represented in European parliaments. The extent to which one profile is more predominant than the others will probably vary across multi-level and unitary systems. We expect multi-level systems to reproduce similar patterns as the one observed in the Spanish case between statewide and non-statewide parties. However, in unitary systems, most parties will be operative in the whole country and their party leaders will have similar incentives to foster organizational activism, as we have observed in Spanish statewide parties. Nonetheless, future research should test the extent to which our findings travel to other systems inside and outside Europe. Explaining the consequences of having or promoting more or less variation of party activists’ types within political parties was beyond the scope of this article. Future research should explore these in more detail in order to provide more insights on how parties cope with heterogeneity and the potential internal conflicts arising from it.