So who really does the donkey work in ‘multi-speed membership parties’? Comparing the election campaign activity of party members and party supporters Academic research paper on "Political Science"

Accepted Manuscript

So who really does the donkey work in 'multi-speed membership parties'? Comparing the election campaign activity of party members and party supporters

Paul Webb, Monica Poletti, Tim Bale

PII: S0261-3794(16)30405-X

DOI: 10.1016/j.electstud.2017.02.002

Reference: JELS 1826

Electoral Studies

To appear in: Electoral Studies

Received Date: 11 October 2016 Revised Date: 20 January 2017 Accepted Date: 12 February 2017

Please cite this article as: Webb, P., Poletti, M., Bale, T., So who really does the donkey work in 'multispeed membership parties'? Comparing the election campaign activity of party members and party supporters, Electoral Studies (2017), doi: 10.1016/j.electstud.2017.02.002.

This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.

So who really does the donkey work in 'multi-speed membership parties'? Comparing the election campaign activity of party members and party supporters

Paul Webb, Monica Poletti and Tim Bale

So who really does the donkey work in 'multi-speed membership parties'? Comparing the election campaign activity of party members

and party supporters

ABSTRACT

One of the traditional functions of party members is to campaign on behalf of their party at general elections. However, they are not the only people who volunteer for the job. In the context of the growing literature on 'multi-speed membership' parties, it is important to ask what non-members do for parties they support. This paper examines how different actors contributed to the electoral campaigns of six parties at the 2015 UK General Election, using survey data covering not only members of the Labour, Conservative, Liberal Democrat, United Kingdom Independence, Scottish Nationalist, and Green parties, but also voters who identified themselves as being close to one of those parties but did not formally belong to them. As well as exploring how much work they do during campaigns, we ask whether the two groups choose different activities and are differently motivated. We find that, at the individual level, party members do more than non-member supporters, and that this is especially true of more intensive forms of activity. We also find that constituency context and political attitudes influence levels of activity in similar ways for members and supporters. However, we find no consistent impact from demographic factors or ideological incongruence. At the aggregate level, we estimate that the campaign work done by supporters may match or even exceed that done by party members.

Introduction

At elections, a good ground game may not be everything but it still means something. In a tight race, it may even mean the difference between a party winning and losing, especially in the UK, a country carved up into single-member constituencies operating under the plurality rule (Johnston and Pattie 2003; Karp, Banducci and Bowler 2008; Fisher and Denver 2009; André and Depauw 2015; Johnston, Pattie, Scully and Cutts 2016). But if 'boots on the ground' are at least potentially important, who is it who wears them and why? And what exactly do they do once they've donned them?

The answer to the first question has traditionally been obvious: grassroots members of political parties. The answer to the second no less so: delivering leaflets, putting up posters, holding meetings and canvassing voters, and then getting them to the polling stations. But what if all this no longer holds? Given (a) the almost ubiquitous decline in the number of people joining political parties in Western European countries (van Biezen, Mair and Poguntke 2012), and (b) the simultaneous rise of new communication technologies, there is good reason to suppose things might have changed. Developments in both the demand- and supply-side of party politics have arguably led to the emergence of 'multi-speed membership parties' (Scarrow 2015) in which different types of party adherents and followers play roles of varying natures and intensities. As a consequence, traditional party members are no longer the only significant contributors to parties and, by extension, to election campaign efforts.

The notion of political parties comprising not only members but also different types of affiliates is not new. Already in the mid-1960s, for instance, Duverger (1964) was distinguishing between electors, supporters, members and militants. A looser conception of affiliation in the American context, closer to identification than membership, was also inherent in Key's (1958) tripartite model of the party in the electorate, the party organisation and the party in office. More recently, however, growing attention has been devoted to the role played by party supporters and affiliates in political parties in general and in election campaigns in particular (Gauja 2013, 2015, Faucher 2015, Fisher et al. 2014). In view of this, drawing on specially commissioned survey data, our aim in this paper is, firstly, to investigate the differences and similarities in campaign activity in the UK General Election of 2015 of party members and supporters - by which we mean those who strongly identify with a party but decide not to formally join it. Specifically, deriving

our expectations from existing literature, we compare the scope, frequency and intensity of the campaigning activities of these two groups. We then examine whether their socio-demographic profile, political attitudes and constituency contexts affect the intensity of their campaign activities in similar or different ways. In broad terms, we find that at the individual level party members are much more likely to campaign for their preferred parties than non-member supporters, and that this is especially true of the more intensive forms of activity. We also find that constituency context (i.e. the marginality of a seat being contested at the election) and the political attitudes of respondents (especially social liberalism, postmaterialism and strong feelings about the EU) influence levels of activity of both groups in similar ways. Demographic factors and subjective ideological distance from their party (ie, 'ideological incongruence'), however, seem to have no consistent impact on either group. For all this, however, the fact that there are so many more supporters than members in the country as a whole means that the sum of campaign activity undertaken by supporters may match or even outweigh that of party members at the aggregate level.

Multispeed membership parties and electoral activity: What we know

In a seminal contribution to the literature on political parties, Susan Scarrow (2015) has argued that parties in parliamentary democracies have often adapted to the widely recognised reality of membership decline by seeking to engage citizens beyond their members with a range of alternative affiliation options. This has resulted in the emergence of 'multi-speed membership parties' in which these new options are typically centralised (because such supporters affiliate directly to the central rather than their local party), digital (in that they are facilitated by electronic communications technology), and accessible (since they are easy to exercise at low cost). By offering such options, parties are in effect seeking to reduce the financial, procedural and/or reputational costs of membership for citizens who might be sympathetic to them. Scarrow describes six different classes of party adherent, any combination of which might be found empirically in a multi-speed party: full members, 'light' members, 'cyber' members, 'financial sustainers', 'followers and friends' and 'party sympathisers'.

One implication of the multi-speed model for active political engagement by citizens is that it may facilitate the development of party links with single-issue campaigns and/or local forms of civic action. Cross and Gauja (2014) point, for instance, to issue-specific

groups affiliated to parties in Australia (Rainbow Labor and Labor for Refugees), and to community organising within the Australian Labor Party. Under Ed Miliband's leadership, the British Labour Party also briefly experimented with community organising, although the initiative was quietly dropped prior to the 2015 election (Bale 2015: 140, 199, 249, 263). This mobilization strategy seeks to emulate the American practice of recruiting supporters directly to become active in various types of neighbourhood volunteer work (not necessarily of an overtly political nature) without signing up as party members. However, a key aim is to generate support for the party, which might eventually translate into electoral or political activity.

An obvious source of these new types of affiliates are individuals who are not already members but who strongly identify with the party. Some of these party supporters may already engage with the party beyond the act of voting in various ways and have been shown to undertake campaign activities traditionally associated with party members. A recent British study by Fisher et al (2014), using data collected in 2010 from parties' election agents, found, for instance, that 'although members clearly still matter, they are not the only source of voluntary activity, especially in election campaigns' - so much so that '[o]ver three quarters of constituency (district level) campaigns in Britain recruited supporters in 2010 and on average, supporters engaged in around two thirds of the activities of members' - a level of input that was judged to be 'non-trivial' (Fisher, Fieldhouse and Cutts 2014: 91-2; see also Scarrow 2015: 103-109 and, on Australia, Gauja and Jackson 2016: 9-12).

Data and Hypotheses: Comparing the election campaign activity of members and supporters

The comparison between party supporters and party members in terms of their involvement in traditional forms of election campaign activity in the UK is also the particular focus of this paper. Using data gathered at the individual level, we explore the campaign activities of the party members and party supporters at the 2015 election. Immediately after the general election of May 2015, we surveyed the members of six British political parties - the first time they have ever been surveyed simultaneously and so soon after an election.1 One of the reasons for conducting the survey when we did was to find out in detail (and hopefully with a greater degree of accuracy than would have been the case had we asked them long after the event) what they did for their

parties during the campaign. At the same time as conducting those surveys, we also surveyed these parties' strongest non-member supporters - that is, people who felt a strong sense of partisan identification but who were not themselves members. Again, the simultaneous implementation of parallel surveys of members and supporters is a first for Britain, and gives us a unique dataset with which to work.

The members survey was conducted in May 2015 (Conservative: n=1192; Labour n=1180; Lib Dem n=730; UKIP n=784; Green n=895; SNP n=963). YouGov recruited the survey respondents from a panel of around 300,000 volunteers who are paid a fee of 50p for completing a survey. Upon joining the YouGov panel volunteers complete a survey asking a broad range of demographic questions which are subsequently used to recruit respondents matching desired demographic quotas for surveys. Potential respondents for the party member survey were identified from questions asking individuals if they were members of any of a list of large membership organisations, including the political parties. At the beginning of the fieldwork period some 8840 YouGov panellists who were party members were eligible to take part in the poll, of whom 5696 respondents subsequently took part in the survey, effectively a response rate of 64.4 percent. Results reported in this article are not weighted in any way since there are no known official population parameters for the various party memberships. However, YouGov party membership surveys using unweighted data have generated predictions for party leadership contests that come very close (that is within 1%) to the final official outcome, which gives us confidence in the quality of the data (https://yougov.co.uk/news/2016/09/24/labour-members-exit-poll-corbyn-wins-all-except-yo/). Further validation was provided by comparing demographics of our UKIP sample with those generated by a far larger UKIP survey (n=13,568) conducted by Paul Whiteley and Matthew Goodwin using a mailback method. We are grateful to Professors Whiteley and Goodwin for facilitating this. The non-member supporters survey was conducted by YouGov in June and July 2015, with the sample drawn in a similar fashion from the British Election Study panel. Specifically, some 6378 respondents who identified themselves as 'very strong' identifiers with the Conservatives (n=1142), Labour (n=1136), the Liberal Democrats (n=1004), the SNP (n=996), the Greens (n=1029) or UKIP (n=1071) constituted our sample. The data are drawn from a national British Election Study (BES) sample that is weighted by age, gender, social class, and region for each group of identifiers and the strength of identification. Henceforth in this

article references to 'supporters' should be understood to refer to these very strong partisan identifiers who are not party members.

Although party members and party supporters are different, in that the former have gone a step further than the latter by formally joining a political party, the substantive distinction between the two can be harder to pin down. Duverger (1964: 62), for instance, pointed out that whereas it is the case that a supporter 'remains outside the organisation and the community it forms', the difference with party members 'blurs and at times disappears'. One reason for this is that members are often inactive, and are now rarely driven to join parties to obtain material benefits (van Haute 2011, van Haute and Gauja 2015). Moreover, comparing members and supporters of the Green Party in Australia, Gauja and Jackson (2016) found that although active members are distinct in several respects from supporters, there is relatively little difference between inactive members and active supporters. This point is underlined in Ponce and Scarrow's evidence that 'behaviourally defined' understandings of members are able to identify 'a relatively homogenous group of active partisans' which includes supporters who might not be active within their party, but whose general political engagement implies they can play a significant 'ambassadorial' role in 'spreading the party message' (2016: 684-685).

Despite the similarities, however, on the whole, the number of activities carried out by an average member will tend to be higher than the number carried out by an average party supporter. Beyond this, party members and supporters might tend to participate in different activities (Duverger 1964: 116). Thus, party members can be expected to participate more in activities most obviously associated with formal membership, such as party meetings or intra-party decision making. Moreover, as Gauja and Jackson (2016), building on the research of Seyd and Whiteley (1992, 2002) and Whiteley at al. (1994, 2006) suggest, if we distinguish between low- and high-intensity activity, we might expect the two groups to participate in different forms of activities at election time: supporters will be more likely to undertake expressive low-intensity activities (such as displaying posters or, nowadays, putting in a good word for the party via social media), whereas members will be more likely to undertake high-intensity activities, such as door-to-door and phone canvassing (Duverger 1964: 101). More specifically in the UK context, previous research by Fisher and his colleagues found that:

...supporters were quite likely to staff polling stations relative to members, and were perhaps surprisingly likely to be involved at the campaign headquarters, despite not being formal members. However, in respect of other activities where voters were contacted either on the doorstep or by telephone, supporters were less likely to be involved than members (2014: 83).

Based on this literature, we can already identify three hypotheses worth testing:

H1: Individuals who are members of political parties will, on average, do more for those parties during election campaigns than will individuals who strongly support those same parties but are not actually members.

H2: Notwithstanding the greater propensity of party members to undertake campaign activity than non-member supporters at the individual level, the aggregate-level input of the latter group will match or exceed that of party members because of their greater numbers in the adult population.

H3: The more 'high-intensity' the campaign activities, the more pronounced the

difference between members and supporters, to the advantage of the former.

After testing these expectations regarding the levels and type of participation at election time of the two different groups, we explore the demographic and attitudinal correlates of their campaign activism. In the light of previous research on political participation and party membership (Verba and Nie 1972; Seyd and Whiteley 2002: 44-47), we should expect that, among both members and supporters, those most active during the campaign will be those who are more highly-educated, male, more middle-class, and middle aged (after which activism tends to decline again). Yet previous research in the UK (Fisher et al., 2014) has found that, at the constituency-level, there is an inconsistent relationship between demographic factors and activism across the two groups, and across the various parties. For the Conservatives, the proportions of graduates and owner occupiers in a constituency predicted the campaign input of members, while the proportion of manual workers influenced the activity of supporters. For Labour, an ethnically diverse population was a significant predictor of member activity, but

population density correlated with supporter activity. Several features of the constituency's demographic profile (an ethnically diverse population and proportions of graduates, owner-occupiers and council or housing association tenants) impacted on the campaign activity of Liberal Democrat supporters - but none of these mattered when it came to the party's members (ibid., 86-89). The same research also found that the presence of Black and Minority Ethnic (BAME) candidates boosted Conservative supporters' activity but not members' activity, while having female candidates correlated positively with Labour members' activism, but not with that of their supporters. Based on these findings, then, we put forward a fourth hypothesis:

H4: There will be no consistent link between socio-demographic characteristics and campaign activism, whether we are talking about party members or supporters.

As for political attitudes, May's (1973) 'law' of curvilinear disparity holds that party members, on the whole, are more ideologically zealous than are party elites or voters (but see e.g. Norris, 1995). Given this, Gauja and Jackson (2016) suggest that party supporters might occupy a mid-way position between party members and voters. They find, however, only mixed support for that idea. Interestingly, other scholars have taken a different approach, focusing on party members who identify themselves as ideologically at odds with the party (e.g. Van Haute and Carty 2011), as they perceive an ideological distance or a gap between their party and themselves. It has been suggested, for instance, that this 'ideological incongruence' could lead party members to vote for a different party at elections or, if the gap is unbridgeable, even to leave the party altogether (Kölln & Polk, 2015). In the case of campaign activity, we might therefore expect that higher levels of ideological incongruence could also lead party members to be less engaged during the electoral campaign, and that this ideological gap might have similar effects on party supporters. We do not expect, in other words, ideological disagreement with the party to have different implications for campaigning efforts for the two groups.

Nor have we any prima facie reason to expect differences between the two groups when it comes to other political attitudes that we suspect may affect participation in campaigning, namely, social liberalism, postmaterialism, and (given its salience in the UK context at the time) attitudes to the EU. Since liberalism places greater emphasis on

democratic engagement as a civic right and a preferred value that maximises liberty and enhances political knowledge on the part of citizens (Howarth 2007), we expect both party members and supporters who are more socially liberal to engage more in campaigning. Similarly, as post-materialist values emphasise self-actualization and self-expression through democratic participation (Inglehart 1977; Klingemann & Fuchs 1995: 16-17), we would also expect to find a positive relationship between postmaterialist orientation and campaign activism among members and supporters. In addition, we would expect citizens with a decided opinion on the question of the UK's membership of the EU to regard it as highly salient and therefore an important reason to campaign, either in the hope of winning support for a pro-referendum or antireferendum party or candidate. Individuals who showed less determinate views on EU membership at the time of the election would not be expected to be driven to participate by this issue, however. All this produces the following hypothesis:

H5: Ideological incongruence will to lead to lower participation, whereas being socially liberal, postmaterialist, or having decided views on the Brexit referendum debate would each lead to higher participation in the election campaign. These expectations regarding political attitudes can be expected to hold equally for members and supporters.

Finally, constituency contextual factors might also influence the decision to participate in electoral campaigns for both members and supporters. Fisher and colleagues found, for instance, that the prospect of victory or possible defeat in a seat prompted more activity by both members and supporters (Fisher, Fieldhouse and Cutts, 2014: 86). By the same token, we expect that the marginality of a seat will lead both groups to participate more, in part because constituencies which are likely to produce closer races are where parties tend to focus their mobilization efforts, and in part because individuals will calculate that the marginal impact of their campaign efforts on the election outcome in the constituency is likely to be greater.

H6: Party members and strong supporters of political parties will be more likely to involve themselves in campaign activities when they live in constituencies which are expected to produce closer electoral races.

Patterns of campaign activism among party members and supporters

Before engaging directly with tests of these hypotheses, we start by briefly surveying some basic descriptive statistics relating to the demographic and political profile of our samples of party members and party supporters. Table 1 reveals few differences across these two sets of respondents in terms of their age profiles. It is also apparent that party members are more likely to be male than their counterparts among supporters; indeed, the differences are quite sharp in all cases. Party members are, on average, much more likely to be educated to graduate level than non-member supporters, with Labour, Liberal Democrat and Greens being the most highly educated in these terms, and Ukippers the least. Party members are also generally more likely to be from non-manual occupational grades than supporters; Liberal Democrats and Conservatives are the most middle-class in this sense, and once again, Ukippers are the least. In terms of subjective self-location on the left-right scale, it is striking but perhaps not too surprising (in view of May's Law) that party members are, without exception, more radical (in the sense of being closer to one end of the ideological spectrum or the other) than their supporter counterparts: that is, Labour, Green, SNP and even Liberal Democrat members all regard themselves as more left-wing than do supporters of their parties, while Conservative and UKIP members are both further to the right. The relative ordering of mean scores for the parties is identical within each of samples: from left to right it runs from Green to Labour, SNP, Liberal Democrat, UKIP and Conservative.

TABLE 1 ABOUT HERE

Moreover, as suggested previously, parties have some kind of general mobilizational advantage in having members and strong partisan identifiers. We find for instance that the stronger a respondent professes his or her partisan identity to be, the more likely they are to vote loyally for 'their' party in the actual election (Table I in Appendix), and that party members are - with the exception of the SNP - even more loyal to their party than non-member partisans (Table II in Appendix).

What are the campaign activities that each of these different groups of actors undertakes? Table 2 addresses this question by revealing the range and number of campaign activities that they performed in the run-up to the election in May 2015. The most striking feature of this table is that, as we expected, full members of political parties are far more active on average than non-member supporters. This is shown by the

summary campaign activism index scores, constructed at the individual-level by adding 1 for each campaign activity (out of the 7 asked about in the survey) declared by respondents. A score of 0 indicates no activity whereas a score of 7 indicates maximum campaign effort. The overall activism index mean for party members is 2.47, almost five times higher than that registered by party supporters, which comes out at 0.51. This gap between members and supporters is similar across parties, ranging from 1.64 for Greens to 2.22 for the SNP. The size of the gap is largely due to the fact that, on average, about 70% of party supporters were completely inactive, compared to about 20% of party members. Even looking at each campaign activity individually, in all cases and across all parties, party members participate far more than party supporters. Thus, we can confidently conclude that, on the whole, party members remain vitally important campaign resources for political parties, who are much more readily mobilized on behalf of their candidates during elections. This clearly confirms H1 (that members will be more active than supporters).

TABLE 2 ABOUT HERE

However, while this is undeniably true when we compare the relative rates of activity undertaken by these two groups, it must be borne in mind that the story may be quite different when we consider the overall impact of that activity on election campaigns at the aggregate level. In other words, we have to take into account the fact that in Britain, as is the case throughout Europe (see Hooghe and Kern, 2015: 953), parties attract far more supporters than members. Consequently, their overall contribution to campaign activity might rival that of party members. A few simple calculations can provide us with estimates of the overall input of each group in the 2015 general election.

TABLE 3 ABOUT HERE

The first section of Table 3 estimates the numbers of each of these groups that the parties were able to draw on as potential campaigners in the run-up to the election. The number of members that each party had in May 2015 (or as close as possible to that time) is taken from Keen (2015). The number of non-member supporters each party might have been able to call upon is estimated by taking the percentages of the overall BES post-election sample who are 'very strong' partisans for each party, and then

calculating what this would amount to in terms of the UK's 45,325,078 registered electors in 2015. Given that there were more than 8.8 million very strong partisans in the electorate (just under a fifth), compared to just 600,000 party members, it is obvious that the parties might have benefited hugely from former, notwithstanding their much lower rates of campaign activity at the individual level.

Just how much this is the case is shown in the third section of the same table, which multiplies the mean score on the campaign activism index (shown in Table 2 and in the second section of Table 3) by the number of people in each group. The figures can be interpreted as estimates of the minimum total number of campaign activities performed by party members and non-member partisans respectively on behalf of 'their' parties; 'minimum' because, of course, our respondents were only asked whether or not they had performed each of these activities - but not how often they had done so. Even allowing for this caveat, it is apparent that non-member supporters may well have contributed more activity overall during the 2015 election campaign, having performed a minimum total of 4.5 million activities altogether. By contrast, party members only managed a minimum of just under 1.5 million activities. Labour in particular seems to have benefited from the input of non-member supporters. This confirms H2 and reinforces one of the major claims made by Fisher and his colleagues (2014) concerning the importance of non-members for campaigning activity.

TABLE 4 ABOUT HERE

Of course, none of this takes account of the specific nature of different types of campaign activity: close inspection of the pattern of the findings reported in Table 2 reveals that party members are at a particular premium when it comes to the more intensive forms of campaign activity such as canvassing and delivering leaflets. For instance, whereas party members were only 2.5 times more likely than non-member supporters to have 'liked' something for their party on Facebook, they were 12 times more likely to have canvassed and 15 times more likely to have delivered campaign leaflets. In order to look at this more systematically, in Table 4 we have distinguished between low-intensity, medium-intensity and high-intensity activities, summarizing participation in these terms. In terms of time and effort, our classification of low intensity activities includes 'liking something by party/candidate on Facebook',

'Tweeting/re-tweeting something by party or candidate' and 'displaying election poster in window' (i.e., all things that can be done without actually leaving one's home); our medium-intensity activity category includes 'delivering leaflets' and 'attending public meetings or hustings' (campaign actions which require one to physically leave the home); while our high intensity category includes 'canvassing face to face or by phone' and 'standing as a (parliamentary or local) candidate' (i.e., things that require direct personal interaction with voters as ambassadors for the party).

The general impression does not change. For instance, whereas party members were about two-and-a-half times more likely to have participated in low intensity activity than party supporters, they were six times more likely to have participated in medium intensity activity, and twelve times more likely to have participated in high intensity activity. On this latter point, however, we should note that high intensity activities are often effectively reserved for members, so it is hardly a surprise that supporters score very low on these activities. Clearly, then, the higher the 'cost' to the individual in terms of time and effort, the more likely that party members rather than non-members will deliver campaign activity on behalf of their parties. This offers broad confirmation of H3, namely that the gap between members and supporters will grow with the intensity of campaign activity.

In brief, while this analysis confirms the multi-speed membership idea that adherents who are more loosely connected to parties than full members may be valuable as human resources in election campaigns, it also points to the continuing centrality of the formal membership for core activities which are vital to electoral mobilization.

Modelling activism: Are members and non-members motivated by the same factors?

We now turn to the second major aim of this paper - to find out what factors influence campaign activism and whether they differ between members and supporters? In order to do this we create models of activism that enable us to compare and contrast the factors that motivate these two sets of actors to campaign during elections. Specifically, we look at socio-demographic (H4), ideological (H5) and constituency (H6) factors, while controlling for party effects.

The socio-demographic characteristics we measure include sex, education, age, age squared4, and social grade. Attitudinal determinants of activism are measured using various scales which tap ideological dimensions widely recognised as salient features of contemporary British politics:

(a) Ideological incongruence is measured by taking the absolute difference between respondents' left-right self-position and that which they ascribe to their national party (both measured on a 0=left to 10=right scale). The measure of ideological incongruence runs from zero (minimum ideological incongruence; i.e. no gap between respondent and party) to 10 (maximum ideological incongruence; i.e. maximum distance between respondent and party). We assume that those who perceive a higher ideological distance between themselves and their party will be less willing to campaign on behalf of it.

(b) Social liberalism-authoritarianism is measured by a standard additive index running from zero (representing the liberal end of the scale) to 10 (representing the authoritarian end) (Heath et al 1993). This is highly reliable.5 Our assumption here is that respondents who are more socially liberal will be more likely to participate in campaigns.

(c) Post-materialism is measured by an index that uses a classic Inglehartian measurement based on four policy objectives about which respondents are invited to express their preferences6. Respondents selecting the two materialist options as their first and second priorities are designated materialists (coded as 1), while those selecting the two post-materialist options are designated as post-materialists (coded as 3), and everyone else is deemed to be attitudinally 'mixed' on this dimension of belief (coded as 2). We would expect to find a positive relationship between postmaterialist orientation and campaign activism.

(d) Attitude towards Britain's relationship with Europe is measured by a question about the referendum on UK membership of the EU: 'If there were a referendum on EU membership prior to the next general election, how would you vote?' If respondents indicated a preference for leaving or staying in regardless of any renegotiated terms of membership that the government might achieve, they were coded as 1; if they

indicated that their decision would depend on the outcome of negotiations they were coded as 0. We would expect those who had definite views on Brexit to be more likely to participate in campaign efforts.

As regards H6, in order to determine whether a race is close we look at marginality, measured using the winning majority of the local MP in 2010, with an index that runs from zero (0% majority) to 10 (100% majority).

Our main dependent variable is the summary campaign activism index score. However, given the extremely wide difference between members and supporters in the high-intensity activities, essentially because those activities are largely reserved for members, in order to make members and supporters more comparable this time we employ a version of the index that only takes into account the low- and medium- intensity campaign activities (running from zero to five and based on the number of different activities carried out), which aims to capture the breadth of activism at the 2015 general election. We treat it as a count variable and run negative binomial regression analysis on it, the most appropriate statistical analysis tool for an over-dispersed count outcome (Long 1997).9 We also

use clustered standard errors at the constituency level to take into account the different level of measurement of the marginality variable.

In Table 5 we run equivalent models for members and partisan supporters that incorporate these demographics (H4), ideological factors (H5), and constituency marginality (H6) as independent variables. In each of these models, using the SNP as the reference category (the party with the highest mean scores on the activism scale for both members and supporters), we include dummy variables for the party to which respondents belong (for party members) or strongly identify with (for party supporters), in order to control for party effects.10 In the model for party supporters we also employ a dummy variable in order to control for party supporters who were previously party members, since we expect these to be less likely to participate in the campaign given their decision to abandon their parties. Table 5 reports the Incident-Rate Ratio coefficients (IRRs), which are the key parameter estimates for negative binomial regression. 11 To get a sense of the substantive meaning of the effects, in the Appendix (Table III) we also show the predicted number of activities for members and supporters

with specific socio-demographic, attitudinal and contextual characteristics for each political party.

TABLE 5 ABOUT HERE

We can see that for party members (Model 1) all demographic characteristics significantly influence campaign activism. Except for age, however, they do so in unexpected ways. As predicted, on average, and controlling for all other variables in the model, the older a member, the wider the range of campaign activities he or she undertakes. This is true, however, until members reach their forties, when their participation in campaign activities starts to decline. However, surprisingly, the party members who put the most effort into campaign activities tend to be female, less educated and of lower social grade. For party supporters (Model 2) it is only gender and age that make a difference - but also in unexpected directions: again it is females rather than men, and young people rather than older ones who tend to participate more. What explains all this is hard to pin down. It is perhaps the case that the usual demographic effects do not apply to party members as a group, since these are people who are already highly politically interested. Indeed, it may be an indication of their unusually high political motivation that females, lower social grade and younger people who join parties are prepared to do so even though they are not demographically typical of party members in general. That is, they join parties and participate in spite of their demographic profile rather than because of it, which suggest a high level of motivation. Whatever the reason, what we can say is that the inconsistency we observe is similar to that reported in the constituency-level analysis of Fisher et al (2014), and therefore essentially confirms H4. We might also observe that ex-members are significantly less likely to be active than other party supporters, all other things being held constant; this should not surprise us, since the very fact that they have left their parties indicates a degree of disappointment or disillusionment with party politics.

When we look at ideology (H5), we can see that both groups do more campaigning when they have decided views on the EU and when they hold more socially liberal values. Being more post-materialist than materialist is also a significant and positive

driver of campaign participation for party members and party supporters. Thus far, H5 is confirmed. However, while the ideological incongruence between the respondent's left-right self-placement and 'their' party (at a national level) matters both for party members and party supporters, the effect runs in different directions, which is a surprise: the greater the distance, the lower the participation of party members in the campaign, which makes intuitive sense and is as we expected; somewhat paradoxically, however, greater perceived ideological distance actually seems to spur non-member supporters to a wider range of participatory activities.

On closer examination this seems to reflect the fact that, among the ranks of the most active supporters we find a substantial group of people who are ideologically quite radical yet see themselves as distant from the party leadership. This sense of distance might well be a reason for their not joining the party in the first place, of course (although we have no direct evidence of this), but it is interesting that they are nevertheless prepared to work for it. When all is said and done, such ideologues might still feel that the party they campaign for is the best option available, notwithstanding its perceived shortcomings. This is best illustrated by reference to Labour's supporters. Examination of the available evidence shows two things that may be relevant here (see Table 6). First, apart from the very small number of people who scored 4 on the activism scale, the more active these non-member supporters are, the more left wing they feel themselves to be. Second, the more active that supporters are, the greater their ideological incongruence with the party (i.e., the greater the left-right gap they perceive between themselves and the party). Thus, it would seem that greater left-right distance is

associated with higher levels of activism among the party's non-members supporters

largely because of the presence of ideological radicals among their ranks. TABLE 6 ABOUT HERE

Overall, we would argue that the findings pertaining to attitudinal influences largely confirm H5, in that their influence is broadly similar across party members and supporters (especially in respect of social liberalism, post-materialism and European integration), with the exception of ideological incongruence. When we turn to the final

predictor in our models, the marginality of the constituency that respondents live in, we see that the safer a seat is, the less inclined both party members and party supporters are to participate in the campaign. This is straightforward and clear confirmation of H6.

By way of summarizing our findings pertaining to hypotheses 4 to 6, we can say that all of our hypothetical expectations are borne out by the data bar the inverse findings relating to perceived left-right distance from the national party of members and supporters. In short, while demography and left-right ideological incongruence work in various and inconsistent ways, being more socially liberal, more post-materialist and having clear views on the EU all impel both members and supporters to involve themselves in campaign activities, as does living in a marginal constituency.

Conclusion

Party supporters are increasingly the target of multi-speed membership parties. Since individuals have long appeared to be less willing to commit to formal membership, party supporters must be seen as a potential alternative source of popular legitimacy and organisational significance for parties. When considering this broader pool of grassroots activists, however, it is important to take into account the differences between the two groups as well as the similarities. Thanks to our contemporaneous surveys of party members and supporters straight after the 2015 General Election, we have been able to investigate similarities and differences in the campaign activity of the two groups in the British context, as well examining the demographic, attitudinal and contextual drivers of participation.

To summarise our major findings, we have seen that, at the individual level, members of political parties are far more likely to engage in campaign activity on behalf of their party than are people who support that party but who do not formally belong to it. The size of this 'campaign activism gap' varies according to the intensity of the form of activity: the more intensive an activity is in terms of time and effort involved, the greater the input of party members relative to non-member supporters. This notwithstanding, at the aggregate level the overall impact of campaign work undertaken by non-member

supporters might very well be as great as (or even greater than) that of party members, given the far larger number of people who simply support parties rather than actually belong to them. In trying to understand what influences people in each of these groups to involve themselves in election campaigns, we have found that the constituency context is likely to be important: for both party members and supporters, the closer the constituency race is expected to be, the more likely that they will be active. We have also seen that political attitudes are generally influential: social liberalism, post-materialism and clear-cut feelings about the EU all incline people to be more active in election campaigns; perceived left-right congruence with the national party also inclines party members to participate more, although not supporters. Demographic influences on campaign activity are not obvious, bar the finding that that women are the most active at elections in the case of both members and supporters.

The differences in demographic drivers of participation may be pointing in the direction of party supporters being different types of people to party members, at least in some respects. To this extent such supporters might be a little more representative of the general population. But the inconsistency of findings regarding demographic correlates is countered by the similarity of attitudinal drivers of campaign activism; only general subjective ideological congruence differentiates members and supporters in this regard. The similarities of political attitude that supporters share with members, along with their greater numbers and aggregate campaign impact, suggests that the former might indeed be a very significant pool from which parties can draw. This means that non-member supporters can potentially be mobilized as boots on the ground on similar political attitudinal grounds as party members. Contextual characteristics such as constituency marginality likewise imply a similar mobilizational potential for both groups.

In other words, although party members remain vital for party campaigning, since they are much more readily mobilized into high-intensity activities, our analysis confirms that party supporters may be a significant organisational and human resource when most needed by parties. Looking solely at party members restricts us to only half of the picture and may provide an exaggerated impression of parties losing their grip on society. Members, then - to borrow from Fisher and his colleagues (2014) - 'are not the

only fruit', but comparing members with supporters is not akin to comparing the proverbial apples and oranges. Both groups are important and similarly motivated resources for twenty-first century multi-speed membership parties. Moreover, this study seems to confirm the logic of parties opening themselves up to categories of supporters who are not formally defined as full members, but who could have a significant role to play in the parties' mobilizational efforts. This provides a cue both for scholars who study the supply-side of political participation and the demand-side of party organizational developments and strategies.

Tables

Table 1: Social and political characteristics of British political party

members, 2015.

Attribute Con Lab LD UKIP Green SNP Total

Mean age M 54 51 51 58 42 49 51

S 57 52 50 56 41 52 52

% male M 71.2 61.6 68.5 75.9 57.5 56.4 65.0

S 51.6 49.6 43.2 55.9 35.9 51.5 48.1

% graduates M 37.9 56.3 55.8 23.1 56.4 41.7 45.4

S 25.3 29.8 44.6 13.3 48.5 28.4 31.3

% ABC1 M 74.6 69.7 76.0 59.9 65.2 61.9 68.2

S 68.6 51.9 67.5 43.0 55.9 46.9 55.7

Mean left-right M 7.76 2.39 4.10 7.34 1.90 2.96 4.44

S 7.53 2.97 4.40 6.77 2.40 3.65 4.66

N M 1193 1180 730 785 845 963 5696

S 1142 1136 1004 1071 1029 996 6378

Notes: M: Members, S: Supporters. Left-right (self-location) = mean seZf-placement on a scale running from 0 (left-wing) to 10 (right-wing).

Table 2: Which of the following things did you do for the party during the 2015 election campaign?

Activity Con Lab LD UKIP Green SNP Total

Activism Index - M 2.15 2.56 2.38 2.28 2.43 3.02 2.47

Activism Index - S 0.25 0.48 0.42 0.37 0.79 0.80 0.51

'Liked' something by M 39.6 51.1 47.4 44.2 67.6 72.7 534

party/cand. on FB S 10.2 18.8 14.8 17.2 35.5 31.2 21.0

Tweeted/re-tweeted M 26. 0 3 6.9 31.1 22.9 45.7 48.6 35.2

something by party S 3.4 7.8 7.8 4.4 18.5 10.4 8.6

Displayed election M 29 .6 51.2 37 .8 42.9 45.1 67.7 45.7

poster in window S 2.5 10.7 6.0 8.1 9.5 21.1 95

Delivered leaflets M 43.5 42.5 45.9 38 .3 28.8 35.4 394

S 2.4 3.0 3.8 1.6 2.4 2.6 2.6

Attended public M 31.3 31.4 28.2 40.5 27.3 49.0 34.6

meeting or hustings S 4.2 4.8 6.8 3.4 8.7 12.6 6.6

Canvassed face to face M 36.5 35.7 32.6 26.1 19.1 28.2 30.4

or by phone S 1.6 2.4 2.2 2.1 4.0 2.4 2.4

Stood as candidate M 9.1 7.0 15.1 13.0 10.2 0.2 8.6

(parliamentary or S 0.6 0.1 0.6 0.2 0.3 0.0 0.3

local)

Number - M 1193 1180 730 785 845 963 5696

- S 1142 1136 1004 1070 1029 996 6337

Note: M= members; S=supporters. Campaign activism index is based on an additive scale that runs from 0 (no activity during the election campaign) to 7 (maximal activity during the campaign, excluding "other"). All activities figures are percentages.

Table 3: Estimates of mean number of campaign activities by party members and very strong partisans, 2015.

Con Lab LD UKIP Green SNP Total

1. Estimates of national totals, 2015

Members 150000 188000 51000 42000 61000 110000 602000

Supporters 3061993 3883464 446623 636577 165192 659054 8852903

2. Activism index (0-7)

Members 2.15 2.56 2.38 2.28 2.43 3.02 2.47

Supporters 0.25 0.48 0.42 0.37 0.79 0.80 0.51

3. Mean total number of campaign activities

Members 322500 481280 121380 95760 148230 332200 1486940

Supporters 765498 1864063 187582 235533 130502 527243 4514981

Note: 'Members' = number of political party members at time of May 2015 general elections, or as near as possible thereof. 'Supporters' = projected numbers of non-members who are 'very strong' partisan identifiers for each party, based on BES 2015 Internet Panel (post-election), Wave 6.0; the number of party members is then subtracted from this figure in order to avoid double-counting, given that most party members are also highly likely to designate themselves 'very strong' partisan identifiers.1 In the third part of the table, each figure is the mean number of campaign activities reported by each group (as indicated by the campaign activism index in part 2 of the table), multiplied by the estimated number of people the group (as reported in part 1 of the table). The figures may therefore be interpreted as the minimum overall number of campaign activities conducted by each group.

1 Calculations of number of non-member supporters are as follows: Based on 45,238,492 registered electors in May 2015

Percentage of very strong Conservative partisan identifiers = 7.1% of BES sample (3,211,933) -150,000 party members = 3,061,933

Percentage of very strong Labour partisan identifiers = 9.0% of BES sample (4.071,464) -188,000 party members = 3,883,464

Percentage of very strong Liberal Democrat partisan identifiers = 1.1% of BES sample (497,623) - 51,000 party members = 446,623

Percentage of very strong UKIP partisan identifiers = 1.5% of BES sample (678,577) - 42,000 party members = 636,577

Percentage of very strong Green partisan identifiers = 0.5% of BES sample (226,192) - 61,000 party members = 165,192

Percentage of very strong SNP partisan identifiers = 1.7% of BES sample (769,054) - 110,000 party members = 659,054.

Table 4. Intensity of campaign activities: Low, Medium and High Intensity of Activity (%)

Intensity of Campaign Con Lab LD UKIP Green SNP Total

Activities

Low-Intensity M 57.4 75.3 67.4 65.0 80.6 87.6 72.0

S 13.5 26.5 21.9 23.1 43.1 41.9 28.0

Ratio M/S 4.3 2.8 3.1 2.8 1.9 2.1 2.6

Medium-Intens. M 52.5 52.2 52.8 51.1 38.5 56.2 51.0

S 6.3 7.0 9.4 4.3 10.3 14.3 8.5

Ratio M/S 8.3 7.5 5.6 11.9 3.7 3.9 6

High-Intensity M 38.0 37.0 36.0 30.0 23.7 28.2 32.6

S 2.0 2.4 2.8 2.3 4.1 2.4 2.7

Ratio M/S 19.0 15.4 12.9 13.0 5.8 11.8 12.1

N M 1193 1180 730 785 845 963 5696

S 1142 1136 1004 1070 1029 996 6337

Note: M= members; S=supporters. Low intensity activity includes liking a party or candidate on Facebook, Tweeting something positive for or about the party or candidate, or displaying a poster. Medium intensity activity includes delivering party leaflets and attending public meetings. High intensity activity includes canvassing voters and standing as a candidate.

Table 5. Explanatory models for breadth of political participation during election campaign

Dependent Variable: (1) Party Members (2) Party Supporters

Campaign Activism IRR s. e. IRR s. e.

Demographics

Female (ref: male) 1.043** (0.020) 1.176*** (0.056)

Education (1 lowest - 7 highest) 0.986** (0.007) 1.019 (0.015)

ABC1 (vs. C2DE) 0.949** (0.046) 0.992 (0.050)

Age 1.249*** (0.046) 0.855* (0.073)

Age squared 0.972*** (0.004) 1.004 (0.009)

Previous party member 0.53 3*** (0.030)

Ideology

Ideological Incongruence (0-10) 0.979** (0.009) 1.037** (0.019)

Liberty-Authority Index (0-10) 0.969*** (0.006) 0.971** (0.013)

Post-Materialism Index (1-3) 1.105*** (0.018) 1.203*** (0.049)

Clear views on EU (ref: unclear 1.121*** (0.034) 1.167** (0.066)

views)

Marginality

Majority % in 2010 (0-10) 0.985* (0.009) 0.965* (0.019)

Political Party

Conservative 0.783* (0.031) 0.333*** (0.032)

Labour 0.780*** (0.024) 0.566*** (0.045)

Lib Dem 0.691*** (0.027) 0.463*** (0.039)

UKIP 0.782*** (0.032) 0.519*** (0.044)

Greens 0.717*** (0.025) 0.728*** (0.052)

Constant 1.940*** (0.214) 2.792 (0.789)

Alpha+ 0.010 (0.012) 0.523*** (0.065)

N (Individuals, Constituencies) 5071, 622 4763, 627

Log Likelihood -8938.06 -4362.28

AIC 3.532 1.840

BIC -25232.51 -31441.16

Notes: Incident-Rate Ratio (IRR) are displayed, s.e. in parentheses; clustered standard errors at constituency level are calculated. *** p<0.01, ** p<0.05, * p<0.1.

+ Alpha is the over-dispersion parameter. If significant, the data is over-dispersed and negative binomial is more appropriate than Poisson model.

Table 6: Labour supporters' self-location on left-right scale and perceived left-right distance from the party, by activism level.

Labour supporters

Campaign Mean left- Left-right

activism right distance

scale score location from party

0 3.15 (702) 1.35 (670)

1 2.85 (185) 1.51 (183)

2 2.24 (97) 1.75 (97)

3 1.75(28) 2.36 (28)

4 3.67 (6) 2.83 (6)

5 - -Total 2.97 (1018) 1.46 (984)

Note: Figures in parenthesis = number of cases.

Appendix

Table I: Partisan identification by vote, 2015

Party Id. Very strong Quite Strong Not very Strong Number

Conservative 95.6 87.9 52.4 6446

Labour 94.6 83.7 53.4 6658

Liberal Democrat 66.7 54.4 41.9 1101

UKIP 78.4 49.6 20.8 1450

Green 72.3 54.7 17.2 587

SNP 92.3 68.6 29.2 1746

Note: Each cell reports the percentage of partisan identifiers voting loyally for 'their' party at the 2015 general election. Data source: BES panel study, (post-election) wave 6.

Table II: Voting preference by party membership and partisan

identity, 2015

Con Lab LD UKIP Green SNP Total

Members 9 0.5 90 . 7 82.7 88.0 79.1 91.6 87.7

Supporters 8 5.5 86. 0 66.9 85.7 66.2 97.0 81.5

Note: Members = percentage of party members voting loyally for 'their' party in 2015; Supporters = percentage of non-member party supporters (ie, very strong identifiers) voting loyally for 'their' party in 2015. Data Sources: UK Party Membership Survey 2015.

Table III: Number of estimated campaign activities per party

Party Members

Party Supporters

Conservative

Labour

Lib. Dem.

Greens

A (low) 1.3

3.3 3.3

3.0 3.3

B (high)

A (low)

1.4 1.1

1.0 1.7

B (high)

1.2 1.1

1.3 1.8

1.1 1.6

Note: 'A' is the predicted number of campaign activities (on a scale from 1 to 5) undertaken by an individual that has the accumulated characteristics of someone we would expect to exhibit low levels of campaign activity - that is, a man (either party member or party supporter), who has very high ideological incongruence with the party (10), is very authoritarian (10), materialist (1), has no clear ideas on the EU (0), and lives in a very safe constituency (6). B is the predicted number of campaign activities undertaken by an individual who has the characteristics of a highly active campaigner - that is, a woman, who has very low ideological incongruence with the party (0), is very socially liberal (0), post-materialist (3), has clear ideas about the EU (1), and lives in a constituency with very low majority (0). We hold demographic characteristics constant for both A and B-type individuals, so that they have A levels qualifications, C2DE social grade and are 30 years old. In the case of party supporters, A is the predicted number of campaign activities for somebody who was not previously a party member, whereas B is the predicted number of campaign activities of an individual who was previously a party member.

References

Aldrich, John, Gibson, Rachel, Cantijoch, Marta Cunhill, and Konitzer, Tobias (2015).'Getting Out the Vote in the Social Media Era: Are Digital Tools Changing the Extent, Nature and Impact of Party Contacting in Elections?' Party Politics, 21: 1-14.

Bale, Tim (2015) Five Year Mission: The Labour Party under Ed Miliband. Oxford: Oxford University Press.

Cross, William, and Gauja, Anika. (2014). Evolving Membership Strategies in Australian Political Parties. Australian Journal of Political Science , 49: 611-625.

Denver, David, Hands, Gordon and Fisher, Justin (2001) Constituency Campaigning in the 2001 General Election. ESRC Award R000239396.

Duverger, Maurice, (1964). Political Parties: Their Organisation and Activity in the Modern State, University Paperbacks Edition. London: Methuen.

Faucher, Florence (2015) 'New forms of political participation. Changing demands or changing opportunities to participate in political parties?', Comparative European Politics, 13: 405-429.

Fisher, Justin and Denver, David (2009) 'Evaluating the Electoral Effects of Traditional and Modern Modes of Constituency Campaigning in Britain 1992-2005', Parliamentary Affairs, 62: 196-2010

Fisher, Justin, Fieldhouse, Ed, and Cutts, David (2014) 'Members are not the only fruit: Volunteer activity in British political parties at the 2010 general election', British Journal of Politics and International Relations, 16: 75-95.

Gauja, Anika (2013). The Politics of Party Policy: From Members to Legislators. Houndmills: Palgrave Macmillan.

Gauja, Anika (2015) 'The construction of party membership', European Journal of Political Research, 54: 232-248.

Gauja, Anika and Jackson, Stewart (2016) 'Australian Greens party members and supporters: their profiles and activities', Environmental Politics, 25: 359-379.

Gibson, Rachel K. (2015) 'Party change, social media and the rise of "citizen-initiated" campaigning.' Party Politics, 21: 183-197.

Heath, Anthony, Evans, Geoffrey and Martin, Jean (1993) 'The measurement of core beliefs and values: The development of balanced socialist/laissez faire and libertarian/authoritarian scales', British Journal of Political Science, 24: 115-58.

Hooghe, Mark and Kern, Anna (2015) 'Party membership and closeness and the development of trust in political institutions: An analysis of the European Social Survey, 2002-2010', Party Politics, 21: 944-956.

Inglehart, Ronald J. (1977) The Silent Revolution: Changing Values and Political Styles Among Western Publics. Princeton, NJ: Princeton University Press).

Johnston, Ron and Pattie, Charles (2003) 'Do canvassing and campaigning work? Evidence from the 2001 general election in England', British Elections and Parties Yearbook, 13: 248-273.

Johnston, Ron, Pattie, Charles, Scully, Roger, and Cutts, David (2016) 'Constituency Campaigning and Canvassing Support at the 2011 National Assembly of Wales Election', Politics, 36: 49-62.

Karp, Jeffrey, Banducci, Susan and Bowler, Shaun (2008) 'Getting Out the Vote: Party Mobilization in a Comparative Perspective', British Journal of Political Science, 38: 91112.

Keen, Richard (2015) Membership of UK Political Parties: House of Commons Library Briefing Paper SN5125 (London: House of Commons).

Key, V.O. (1958) Politics, Parties and Pressure Groups. 4th ed. New York: Crowell.

Klingemann, Hans-Dieter, and Fuchs, Dieter (1995) Citizens and the State (Oxford: Oxford University Press.

Kolln, Ann-Kristin and Polk, Jonathan (2015) 'Is Everyone Enjoying the Party? Examining Ideological Incongruence among 10,000 Swedish Party Members', Paper prepared for presentation at the Val: Opinion: Demokrati Workshop Series, Department of Political Science, University of Gothenburg, 22 September 2015.

Long, S., and Freese, J. (2006) Regression Models for Categorical Dependent Variables Using Stata. College Station, TX: Stata Press.

May, J.D., (1973). Opinion structure of political parties: the special law of curvilinear disparity. Political Studies, 21: 135-151.

Milbrath, Lester W. (1965). Political Participation: How and Why People Get Involved in Politics. Chicago: Rand Mc Nally.

Norris, Pippa. (1995). 'May's Law of Curvilinear Disparity Revisited' Party Politics, 1: 29-47.

Ponce, Aldo F. and Scarrow, Susan E. (2016) 'Which members? Using cross-national surveys to study party membership, Party Politics, 22: 679-690.

Scarrow, Susan E. (2015) Beyond Party Members: Changing Approaches to Political Mobilization (Oxford: Oxford University Press).

Seyd, Patrick and Whiteley, Paul (1992) Labour's Grassroots: The Politics of Party Membership. Oxford: Clarendon Press.

Seyd, Patrick and Whiteley, Paul (2002) New Labour's Grassroots: The Transformation of the Labour Party Membership. London: Palgrave.

van Biezen, Ingrid, Peter Mair, and Thomas Poguntke (2012) "Going, going,... gone? The decline of party membership in contemporary Europe." European Journal of Political Research 51: 24-56.

Van Haute, Emilie and Gauja, Anika (2015) Party Members and Activists. London: Routledge.

Van Haute, Emilie (2011) Party Membership in Europe: Exploration into the Anthills of Party Politics. Brussels: Editions de l'Universite de Bruxelles.

Van Haute, Emilie and Carty, R Kenneth (2011). 'Ideological misfits: A distinctive class of party members', Party Politics, 17: 1-11

Verba, Sidney and Norman H. Nie (1972) Participation in America: Social Equality and Political Democracy. New York: Harper & Row.

Verba, Sidney, Schlozman, Kay L., Brady, Henry, and Nie, Norman (1993). 'Citizen activity. Who participates? What do they say?' American Political Science Review, 87: 303-318.

Whiteley, Paul, Seyd, Patrick, and Billinghurst, A. (2006) Third Force Politics: Liberal Democrats at the Grassroots. Oxford: Oxford University Press.

Whiteley, Paul, Seyd, Patrick, and Richardson, J. (1994) True Blues: The Politics of Conservative Party Membership. Oxford: Clarendon Press.

1 This research was made possible by the support of the Economic and Social Research Council's grant ES/M007537/1. We gratefully acknowledge this support.

2 Note that, contrary to Fisher et al (2014) and Gauja and Jackson (2016), we prefer a tri-fold distinction of campaign activities into low-, medium- and high-intensity. This is because activities like displaying a poster or supporting the party on the web (low-intensity activity) clearly require less effort than leafletting or attending a party meeting (medium-intensity activity).

3 These items are coded as follows: Gender: 1 (male), 2 (female); Education: 1 (No qualifications), 2 (Junior vocational qualifications), 3 (CSE), 4 (GCSE, O levels, etc.), 5 (A levels/ Scottish highers), 6 (Higher vocational qualifications), 7 (Graduate); Social grade: 1 (C2DE - ie, manual employee), 2 (ABC1 - ie, non-manual employee); Age: Respondent's age in years, divided by 10 to ease interpretation (change for a 10-year difference shown), so that it now runs from 0 to 10; Age squared.

4 There is evidence that older individuals up to their forties or fifties tend to participate more in political actions, after which age activism tends to decline again (Milbrath, 1965: 135).

5 The individual items from which the Liberty-Authority scale is constructed as follows: "Young people today don't have enough respect for traditional values"; "People who break the law should be given stiffer sentences"; "For some crimes the death penalty is the most appropriate sentence"; "Schools should teach children to obey authority"; "Censorship of films and magazines is necessary to uphold moral standards". Responses are coded so that 1 is the most socially liberal option and 5 the most socially authoritarian option. The resulting scores are normalized and multiplied by 10 so that the final scale runs from 0 to 10. Cronbach's alpha for these five items in the party membeship data = .846, and in the non-member supporters data it = .837, which indicates consistently high levels of scale reliability across the two datasets.

6 Two items (maintaining order in the nation, and fighting unemployment) constitute materialist preferences and two items (giving people more say in important government decisions, and protecting freedom of speech) constitute post-materialist preferences.

7 The winning majority has been divided by 10 to make a change of 1 unit more readily interpretable compared to other independent variables in the models (i.e. coefficients indicate the effect of a change of 10 percentage point in the winning majority)

The excluded high-intensity activities are canvassing and standing as candidates; by definition, the latter can only apply to those who are full members, while parties are often loth to include non-members as canvassers for fear that they might be 'loose cannons' who are not fully cognisant of accepted canvassing procedure or party policies. The five activities that count towards the dependent variable are therefore: liking a party or candidate on Facebook, Tweeting something positive for or about the party or candidate, displaying a poster, delivering party leaflets and attending public meetings. Note that we considered whether a weighted dependent variable taking into account different costs of activities would be appropriate. However, since results were extremely similar to those shown and it was not completely intuitively clear what it would mean to apply weights to the number of different activities carried out in campaign, we present results for the unweighted variable. Factor analysis run for each party in the party members and supporters datasets confirms that a single latent factor underlies the five activities.

9 Count variables can be modelled using either Poisson regression or negative binomial regression. They both have the same mean structure, but negative binomial also has an extra parameter (alpha) to model over-dispersion (i.e. when the conditional variance exceeds the conditional mean). In our case, negative binomial regression was deemed more appropriate due to a significant overdispersion in the campaign participation index of party supporters (The Likelihood Ratio test (121.64, 1 df, p<0.001) comparing the full negative binomial model to a Poisson model shows that Alpha is non-zero (Long & Freese, 2006)).

10 All else being equal, models confirms that members and non-members of every party participated in significantly fewer activities than members and supporters of the SNP. The gap with SNP participation is particularly evident for Liberal Democrat and Conservative party supporters, and, to a lesser extent, for Labour and UKIP supporters.

11 As with odds ratios (OR) in binomial logistic regression, IRRs are the exponentials of the Poisson and negative binomial regression coefficients (exp (b)) and they are often used for ease of interpretation (a coefficient higher than 1 corresponds to a positive relation, whereas a coefficient lower than 1 corresponds to a negative one). The Poisson and negative binomial regression coefficients have an additive effect in the log(y) scale, while the incident-rate ratio coefficients (IRR) have a multiplicative effect in the y scale.

Similar findings apply to Conservative non-member supporters, albeit to a less pronounced degree. Details available from authors on request.

Acknowledgement

This work was supported by the Economic and Social Research Council (grant number ES/M007537/1).