Introduction

In social movements, changes in individual-level perceptions and attitudes can have far-reaching consequences; irrespective of the success of a movement in relation to the political system, people’s experiences from participation may likely impact the activists’ future life and political engagement (McAdam, 1988, 1989). Through movement participation in activist networks, the individual accumulates a history of activism that combines learning of skills and moral socialisation. In this process, the activist learns the cultural codes, styles, habits of action and ways of thinking, which influence future participation (della Porta, 2018; Eliasoph, 1998; Eliasoph & Lichterman, 2003; Lichterman & Eliasoph, 2014; van Stekelenburg, 2017). Furthermore, participation may change the activist’s self-understanding through processes of changes in worldview such as emotional liberation and collective identity formation. The changes to the activist’s worldview alter how the activist perceive and approach different situations in the future and may ultimately lead to changed patterns of action (Jasper, 2018; McAdam, 1999a; Melucci, 1989; Passy & Monsch, 2020). Hence, individual-level outcomes of altered perceptions, attitudes and values are one reason why even movements with no notable institutional impact and prefigurative politics may still be of significance to society’s values and moral order.

Studies of such individual-level outcomes of activism usually depart from how participation impacts the participants (Bosi et al., 2016; Carlsen et al., 2020b; Giugni, 1998; Giugni et al., 1999; Toubøl, 2019). Not disputing the importance of the participation process itself, this leaves aside the question of how predispositions—that is, attitudes, tastes, habits, values, principles, etc., formed prior to engagement with the movement—influence not only participation but also the outcomes of movement activism. While the question of how predispositions influence activism has received attention (Gundelach, 1995; Gundelach & Toubøl, 2019; Klandermans, 2014; McAdam, 1986; van Deth & Scarbrough, 1995b), there is a void in the literature when it comes to how predispositions relate to outcomes (Converse, 1964; Schwartz, 2007). This void relates to the problematic marginalisation of values and other attitudinal, moral and ideological factors from the field of social movement studies (Walder, 2009). Furthermore, studying the complete process of how predispositions directly and indirectly through the mediating context of participation in activism influence attitudinal outcomes and how this creates new predispositions needs to be theorised and explored empirically. In this chapter, we set out to investigate this process, focusing on how value predispositions’ relationship through mediating contexts of activism relates to changes of attitudinal outcomes in the process as depicted in Fig. 5.1.

Fig. 5.1
figure 1

The relationship between value predispositions and attitudinal outcomes mediated by contexts of activism

Full size image

This paper advances our knowledge of the complex relationship between value predispositions, participation and attitudinal outcomes by (1) developing theoretical hypotheses specifying the process of how value predisposition relates to different kinds of activism and attitudinal outcomes and (2) by empirically testing the hypotheses of value predispositions’ relationships with attitudinal outcomes as mediated by social movement contexts of participation. Limiting our focus to the question of mediated relationships between values and attitudinal outcomes implies that we do not concern ourselves with the potential direct effects of values on attitudinal outcomes. We analyse two sets of value predispositions: (a) conformity measured by the religious affiliation of (i) non-believers, (ii) self-identified passive Christians and (iii) active Christians and (b) values of self-transcendence (strongly related to altruism) and self-enhancement (strongly related to egoism) measured on the basic human value scale. Outcomes of attitudinal change are (c) loss of trust in the political institutions representative of the partisan political system in the form of (i) Parliament and the repressive state apparatus represented by (ii) the judiciary system and (iii) the police. Also, we analyse the relationship to changes in the central issue of concern to the movement, namely, the (d) political view of immigration policies. Analysing these relationships, we focus on how they are mediated by the movement contexts of (e) participation in activism of three different kinds, namely, (i) humanitarian activity, (ii) political protest and (iii) civil disobedience. This is achieved by defining a statistical model that can handle several dependent variables and capture the layered process of participation and its subsequent outcomes and how predispositions influence both participation and outcomes. Our case is the Danish refugee solidarity movement, and we analyse Danish residents’ activism to help refugees and how their predispositions influence participation and outcomes hereof. In short, the result suggests that predispositions, in general, do not directly influence changes in attitudes but instead influence attitudes as mediated through different contexts of activism in the sense that specific contexts of activism connect certain values to certain attitudinal outcomes.

These findings highlight the importance of predispositions in the process of recruitment and participation and individual-level outcomes. The complex findings of interactions and indirect effects mediated by specific and distinguishable contexts call for more practice-oriented theorising of the process of recruitment, participation and outcomes. It is also important to stress that the exploratory pioneering nature of the study implies that the findings are, first and foremost, hypothesis generating. This is the case because the design is based on a cross-sectional survey that, in general, does not allow for making causal claims concerning the processes underpinning the correlations. Therefore, more studies are needed to test and develop hypotheses.

Our case is the Danish refugee solidarity movement, which is constituted of people who organise to support refugees and their rights in Denmark. The movement is deeply embedded in humanitarian ideology and is concerned with a political topic that concerns a fundamental Western set of interlinked moral values of human rights, human dignity, and the sacredness of human life (Joas, 2013; Toubøl, 2017). Given the topics’ strong relations to conflict over basic morality central to democratic institutions and liberal societies, value predispositions are likely to be central to the movement participants. This makes the movement a strategic case for exploring the role of value predispositions, and we will focus on two value dimensions salient to the movement. Self-transcendence is a value of particular relevance because it is intimately related to the movement’s core ideals and activities of solidarity and political altruism. Second, a distinguishing feature of the movement is its heterogeneous activist base consisting of both activists from the political left, local groups with no particular political affiliation as well as people from conservative religious groups who all share a common concern and compassion for refugees (Toubøl, 2015). The unusual fact that the movement mobilises both traditional and conformist individuals from religious networks and progressive non-conformist left-wing activists motivates our choice of also focusing on the value of conformity, which we will measure in terms of affiliation with the dominant religion of Danish society, Christianity. Also, the movement’s broad and varied collective action repertoire makes it strategic for studying how value predispositions’ influence on attitudinal outcomes is mediated by different kinds of activism. In turn, the broad repertoire creates interaction with a wide array of political institutions, making it relevant to consider how different kinds of activism relate to different attitudinal outcomes of institutional trust. In sum, being the most likely case for observing variation in values, movement repertoire and involvement with political institutions, the movement is strategic and suitable for our purpose of exploring how values, activism and attitudinal outcomes are interrelated.

In the following section, we discuss theories about the value-attitude-action triad and develop hypotheses regarding the relationship between values, contexts of activism and attitudinal outcomes. “Data and Methods” section details data and methods, including operationalising the theoretical model into a recursive block structure, which allows analysing the complex set of hypotheses. “Results” section presents and elaborates the empirical results, and finally, in “Conclusion and Discussion” section, we conclude and discuss the implications of the findings for the literature on social movement outcomes for individuals.

Value Predispositions, Activism and Attitudinal Outcomes

In general, questions concerning values and their role in the mobilisation process and for the strategies of movements, as well as the question of which values movement ideologies are concerned with, have remained understudied (McAdam, 1986; Walder, 2009). This theory section is guided by the very general theoretical model in Fig. 5.1 and explains how we theorise the relationships based on the existing literature. In doing this, we pay specific attention to the indirect effects that are mediated by contexts of participation in activism and develop a set of hypotheses. Finally, the resulting five hypotheses are summarised in Fig. 5.2.

Fig. 5.2
figure 2

Overview of theoretical hypothesis

Full size image

Values, Context and Attitudes

Before considering theories of the relationship between predispositions, activism and attitudes, we briefly deal with defining, on the one hand, the predisposition of values and, on the other hand, the outcome of attitudes. There is no clear consensus on the concepts in the social sciences (van Deth & Scarbrough, 1995a), and to complicate matters, values and attitudes are often confounded (Schwartz, 2007). However, they can be defined as quite distinct. Following Schwartz’s definition, values refer to desirable goals that motivate action and transcend specific contexts (Schwartz, 2007). In contrast, attitudes are specific to issues, objects, actions and situations; they are thereby non-transcendent but tied to specific contexts. Thus, across a variety of contexts, the same values may result in different attitudes, or as van Deth and Scarbrough put it, values are the ‘underlying orientations, which are relevant for or inform the process of, arriving at attitudes’ (van Deth & Scarbrough, 1995a, p. 32).

However, not all values are equally important for attitude formation in all contexts, and which values are activated depends on the context. The competition is determined by the understanding of the given situation in which the actor arrives. This implies that values are ordered by importance relative to each other in any given context, which, according to Schwartz, is another characteristic that distinguishes them from attitudes. An attitude toward a given object, situation or event is not in competition with other attitudes. However, an attitude may be ambiguous when it combines more than one value activated in the given context.

This understanding of the relationship between values, contexts and attitudes suggests a process where values are transformed into attitudes in a specific context. However, in empirical studies of values’ relation to attitudes or action, the contextual factor is rarely, if ever, considered beyond very general macro-context variation at the level of countries. In the following, we shall discuss how meso-level contexts of different forms of activism may mediate the relationship between values and attitudes.

The Mediating Role of Contexts

Introducing the intermediate context of participation in movement activities adds a layer of complexity to the theory. On the one hand, we must consider how value predispositions relate to participation in activism and, on the other hand, how participation may influence outcomes of changes in attitude. We deal with each step in turn.

We have already touched upon how values inform motivations for action. This has been confirmed in the literature on social movements. For instance, Dauphinais et al.’s (1992) study of ‘Predictors of Rank-and-File Feminist Activism’ concludes that predispositions are vital predictors of active versus non-active feminist activists. These findings are supported by studies of other movements such as Barkan et al.’s (1995) study of the antihunger movement and Stern et al.’s (1995, 1999) work on environmental movements, and more recently Lahusen and Grasso (2018) published the edited volume Solidarity in Europe , which provides ample evidence of the close relationship between values, attitudes and activism (Fernández, 2018; see also Fernández G. G.’s contribution in chap. 4).

As argued above, values are mediated through different contexts, which are of consequence to participation (see also Passy and Monsch’s contribution in chap. 6). Processes related to group culture and collective identity formation have been shown to have a substantial impact (Carlsen et al., 2021b; De Weerd & Klandermans, 1999; Klandermans, 2015; Klandermans et al., 2002; Passy & Giugni, 2000, 2001; Passy & Monsch, 2020), but emotional reactions have also been found to be very influential (Gundelach & Toubøl, 2019; Jasper & Poulsen, 1995). Moreover, and of particular importance to the solidarity movement under study, encounters between activists and deprived others may forge solidary relationships that influence future participation (Carlsen et al., 2020b; Maggini & Fernández, 2019).

Studies also suggest that values are not equally important to all kinds of movement participation. As suggested by McAdam (1986), we can think of recruitment and movement participation as a process beginning with movement activities that entail low levels of risks and costs before the activists gradually, often facilitated by a process of socialisation with movement goals and culture, move on to activism that entails higher risks and costs. When considering this distinction between low- and high-risk/cost activism, values appear to be particularly important to the initial stages of low-risk/cost activism, but when moving on to high-risk/cost activism, values tend to lose importance relative to processes of network embeddedness, socialising and learning processes (Dauphinais et al., 1992; Gundelach & Toubøl, 2019; McAdam, 1986). Such observations warrant closer scrutiny of the relationship between values and different contexts of activism.

In this study, we distinguish between three qualitatively different contexts of participation in the refugee solidarity movement that vary along the dimensions of (1) contentiousness concerning the degree of confrontation with other political actors, (2) risk concerning the individual risks of participation in the activities (Toubøl, 2017, 2019) and (3) civil disobedience. The first kind of participation is the most common in the movement, namely, humanitarian activities that aim to alleviate the suffering of the refugees. Humanitarian activities stand apart by being both non-contentious (it is not per se related to a political conflict, even though it might be) and low-risk. The second form of participation is political protest, a classical contentious form of activism that implies a low to medium level of risk. Finally, civil disobedience, such as helping refugees go underground or obstructing deportations, is contentious and high-risk. Elsewhere, we have argued for the substantial and theoretical relevance and meaningfulness of focusing on exactly these three kinds of activism in relation to the particular case of the refugee solidarity movement (Toubøl, 2017, 2019).

In relation to the values and types of activism under study, we first hypothesise that (1) strong values of self-transcendence relate to strong engagement with the low-risk activism like humanitarian activities and political protest but do not influence participation in high-risk civil disobedience because prior research has pointed to this kind of activism being the result of network embeddedness and related processes of socialisation (della Porta, 2018; Wiltfang & McAdam, 1991). Second, we expect that (2) values of conformity are negatively associated with participation in civil disobedience due to its inherent rejection of existing institutional orders and also political protest because that activity also constitutes a challenge to society’s hierarchical orders and an unconventional extra-institutional approach to democratic institutions. This implies that we expect the non-conformists to participate in political protest and civil disobedience because they do not take society’s traditional order for granted.

Turning to how activism produces outcomes of attitudinal change, the literature is scarcer. However, studies of biographical consequences of movement participation often have a change of attitudes and values at the centre (McAdam, 1988, 1989, 1999b; McAdam & Kloos, 2014). While there is a consensus that activism profoundly influences the activists, the social movement literature is limited when it comes to formalised theories and models. However, other literature on the specific attitudinal outcomes of this study, institutional trust and political views, have valuable insights to offer.

Following Max Kaase (1999), we view trust as relational, and therefore loss of institutional trust involves interaction between individuals and institutions. The interactions that constitute trust differ between different kinds of institutions. Bo Rothstein and Dietlind Stolle distinguish between partisan institutions, pertaining in this case to Parliament representing the political system, and order institutions like the legal system and the police (Rothstein & Stolle, 2008), which represent the oppressive state apparatus (Althusser, 1971) with which movements and activists often find themselves in conflict. A core distinction in relation to institutional trust is that citizens expect political bias from partisan institutions but impartially and neutrality from the order institutions. For partisan institutions like the Parliament, the interactions that constitute trust can be understood in terms of political efficacy (e.g., Craig et al., 1990; Pollock, 1983), in particular external efficacy, which is the political institutions’ responsiveness to the activists’ demands (not to be confused with giving in to the activists’ demands). Thus, if experiencing non-response or lack of willingness to engage in debate and dialogue from politicians, even though the politicians and activist might disagree, it will likely lower trust due to the institution’s lack of responsiveness. For order institutions like the police and legal system, trust stems mainly from the procedural aspects of justice rather than their perceived performance (Tyler & Huo, 2002). Therefore, what matters for trust in legal institutions are personal experiences of high levels of fairness in the exercise of legal authority even though the final verdict may go against oneself (Jackson et al., 2012; Nix et al., 2015).

For both partisan and order institutions, it all boils down to the fact that trust is constituted through interactions between individuals and institutions. Therefore, forms of activism that imply interaction with the particular institutions are more likely to result in a loss or gain of trust (gain is rarely observed in data and, therefore, not considered here). Political protest is a form of interaction with political institutions, and therefore we hypothesise that (3) participation in political protest activities will lower institutional trust, particularly in political institutions. It is likewise for civil disobedience, which, however, entails interaction with the order institutions of police and judiciary to a higher degree. Therefore, (4) we expect participation in civil disobedience to cause a loss of trust in political institutions, both partisan and order institutions. In contrast, humanitarian activities rarely involve direct interaction with political institutions. Therefore, we do not expect humanitarian activities to create changes in institutional trust.

Finally, we consider the outcome of a change of political view. This outcome is not tied to interaction with a particular institution but rather with engaging with a political topic. Social movement studies contain multitudes of observations of how participation in movements, including refugee solidarity movements (e.g., Cunningham, 1995), develops political consciousness and awareness through socialisation processes (e.g., McAdam, 1988). What seems to be the common denominator of activism involving altering political consciousness and views is the presence of a contentious dimension; that is, the meaning of the activities relates to a contested political topic (Carlsen et al., 2021b; Eliasoph & Lichterman, 2003). Thus, we hypothesise that (5) the two contentious contexts of activism, political protest and civil disobedience, are positively related to a change of the political view of immigration.

Figure 5.2 sums up the complex set of hypotheses in a theoretical model that further unpacks the model’s operationalisation presented in Fig. 5.1. It also hints at a need for a sophisticated statistical model, which we will explain in detail below.

Data and Methods

We use a non-representative convenience sample collected in activist groups on the social media Facebook May–July 2016. As described elsewhere (Carlsen, 2019; Toubøl, 2017), the Danish refugee solidarity movement experienced a massive revitalisation during 2014 and, in particular, in relation to a wave of refugee migration through Europe in the summer and fall of 2015. Facebook groups were a central component in the movement’s organisation, which allowed us to sample the movement activists from these groups, constituting our sample frame. We invited members to participate by posting invitations, including a link to the online questionnaire, in the groups. Also, from data accessed through the Facebook API, we were able to produce measures of the sample frame, which allows us to assess the sample’s representativity.

We identified 165 relevant groups, and in 150 (91%), we obtained permission to field the survey. In addition, the survey was posted on 137 Facebook pages related to the movement. This resulted in a total of 2289 complete responses to the online questionnaire (51 question pages, median completion time of 22 min). To assess representativity, we compared the response sample to the sample frame of 28,304 Facebook users who were active in the groups during the period of the survey. From the social media data, we were able to produce two measures allowing us to compare representativity, namely, gender from a name classifier and a mean position on a political left-right scale from ‘like’ behaviour.Footnote 1 The sample reflects the sample frame, having a majority of women but also over-representing women with 84% in the sample versus 76% in the sample frame. The sample is also significantly more politically left-leaning with a mean of 3899 compared to the sample frame’s mean of 4159 on a 1–10 points scale, even though the difference of 0.26 point is small. None of the biases is alarming, but the underrepresentation of men should be kept in mind when interpreting results where gender might be an important factor.

The survey is cross-sectional and does not allow for causal inference. However, we take advantage of the fact that a dramatic event took place on 5 September 2015 when a large number of refugees in an unregulated manner started crossing the border to Denmark, primarily from Germany. This became a dramatic national moment and was immediately followed by intense mobilisation and activity in the movement. This dramatic event allows for more reliable retrospective inquiry, and several items are constructed in ways that separate activities and level of movement participation and timing of involvement between, before and after 5 September 2015. This allows for including a time dimension in the models, although we still refrain from drawing causal conclusions but see the design as strengthening the reliability of the explorative findings and consider our findings as suitable for qualifying existing and suggesting new theoretically plausible hypotheses.

Variables

The dependent variables are created from survey questions asking if the respondent, as a result of his or her engagement, changed their level of trust in the institutions of Parliament, the judiciary and the police or changed his or her view of immigration in a positive or negative way.

Table 5.1 summarises the variables regarding institutional trust. In the analysis that follows, we consider only the difference between no change and a decline in trust and leave out an increase in trust because this event is so rare (with the exception of the police). The wording of the questions suggests that the responses reflect a causal relationship between movement activity and loss of trust. Hence, the following analyses concern what variables relate to such a change in trust.

Table 5.1 Distribution of answers to the question ‘Have what you learnt and your experiences with the refugee cause changed your trust in the following institutions?’
Full size table

For the analysis, we combined the three variables into a scale measuring the number of institutions in which the respondent lost trust (summarised in Table 5.2). The scale conforms almost perfectly to the assumptions of a Mokken scale. Out of 2289 responses, only 127 (6%) do not conform to the Mokken scale hypothesis regarding data structure, and we drop these respondents from the analysis. Without these respondents, the items form a scale with Loevinger’s H = 0.74. This indicates very high scalability, and consequently we adopt the scale of loss of institutional trust as our measure of attitudinal change toward political institutions.

Table 5.2 Construction of Mokken scale of loss of institutional trust
Full size table

The measure of the other dependent variable that measures attitudinal outcomes concerning changes of political view is binary. Those reporting no change of opinion (0) comprised 87%, and 13% reported having adopted a more refugee-friendly political view (1) as a consequence of their involvement with the movement.Footnote 2

The focal independent variables of the value predispositions of self-transcendence and conformity are operationalised following different principles. Following Schwartz (Davidov et al., 2008), self-transcendence is measured by four items from the basic value orientations of universalism and benevolence. The items are adopted from the European Social Survey (ESS) and form an additive index with the principal range of 4–20. However, since the distribution is highly skewed, categories 4–13 have been collapsed, which results in a scale from 1 to 8. We also include a variable that measures self-enhancement values which may be equated with an egoistic personality. We have not formulated specific hypotheses regarding self-enhancement but include it nonetheless as an important control variable because, in Schwartz’ theory, it represents a value opposite that of self-transcendence (Schwartz, 1992; Schwartz & Bilsky, 1987). Similar to the self-transcendence scale, the variable is measured by four items from ESS. The principal scale ranging from 4 to 20 has been recoded into a 1–7 scale to maximise variation and avoid too few observations in categories at the tails of the distribution.

Conformity is measured by a proxy, religious affiliation with three categories: (1) active member of the Evangelical Lutheran Church in Denmark or another Christian religious association (11%), (2) passive member of the church (34%) and (3) not being religiously affiliated (55%).Footnote 3 As Schwartz remarks, religious behaviour may be associated with values of conformity and tradition (Schwartz, 2007), but how to interpret this matter, we may add, depends on context. For instance, studies from the US sectarian religious tradition show that religious activity drives activism and, thus, rather than conformity, inspires deviance (Cunningham, 1995; e.g., Delehanty, 2020). However, the Danish religious landscape is far from sectarian.

The Danes have been characterised as people who are ‘belonging without believing’ (Storm 2009). On the one hand, the Danes have low religiosity (Evans & Baronavski, 2018), and Denmark has been characterised as the least religious country in the world (Zuckerman, 2008, 2009). On the other hand, 74% of the population are members of the Evangelical Lutheran Church in Denmark.Footnote 4 However, churchgoing is low, and only one in six believes in a personal God (Andersen et al., 2019). The strong institutional position of the Evangelical Lutheran Church in Denmark is due to its century-long state-sanctioned monopoly on religious enterprise. Furthermore, the church is a national symbol, and church membership has a central role as a sign of integration and belonging to the Danish national identity (Iversen, 2018; Iversen et al., 2008; Sundback, 2008; Warburg, 2008). This is, for instance, reflected in the Danish name of the church, Folkekirken, which translates into ‘The People’s Church’.

Studies show the active membership in a dominant church is associated with conformity; that is, active members are more likely to trust key political institutions (Brañas-Garza et al., 2009; Kasselstrand et al., 2017; Kasselstrand & Eltanani, 2013), have more conservative political leanings (Esmer & Pettersson, 2007; Wolf, 2008) and respect authority (Proctor, 2006). Therefore, we assume that active membership in the church implies a relatively high level of conformity, while being a non-believer indicates a very low level of conformity because it is a breach of the membership norm, which, as argued above, not only relates to religious matters but perhaps even more to matters of national identity. Passive membership in the church indicates a level of conformity in between. The variables are treated as binaries. However, in the case of a minority religion, the relationship between religious affiliation and conformity might be different. Due to this uncertainty, we restrict the measure to concern members of Christian churches in Denmark and add a control for association with the only religious minority of any substantial size in the population and sample, namely, Muslims. This has implications for generalisability and transferability of findings related to conformity, which are only valid in relation to the dominant religion of the majority culture in contexts similar to that of Danish society.

The context in which values are activated and motivate activism that may lead to change in institutional trust and political views is measured by three variables assessing the level of participation in three types of activism: (1) humanitarian activity, (2) political protest and (3) civil disobedience (Toubøl, 2019). They are created from an item inquiring about the respondents’ participation in 16 activities (summarised in Toubøl (2019), Table III) during and after the mobilisation that began in September 2015. The classification of the 16 activities into three categories is motivated by theoretical and substantive considerations, including detailed knowledge of the contents of the movement’s activities from extensive fieldwork (Toubøl, 2017). As explained in the theory section, we expect the three categories of activities to imply different patterns of interactions with the political institutions. Therefore, both constitute a direct relation to the loss of institutional trust and mediators of value predispositions’ relationship with loss of institutional trust.

Finally, we add several control variables. Gender, age and level of educational attainment control for sociobiographic and sociodemographic factors. More specific to social movement activity, we control for history of activism (Wiltfang & McAdam, 1991) before the mobilisation in September 2015, which is inferred from an item asking the respondents to report different forms of political participation before and after September 2015. Retrospective inquiry is difficult in a survey, but given the iconic and dramatic status of the events in September 2015, the measure’s reliability might be acceptable (Belli, 2014). Similarly, we ask the respondents if they were active in the movement before September 2015 or became active only during the mobilisation that followed. This provides us with a measure of embeddedness in the movement networks. The two measures serve as indicators of the level of socialisation of movement identity, values, views and beliefs, as well as learning of movement practices, including activism (McAdam, 1986; van Stekelenburg, 2017). Both ideational and practical socialisations may influence risk and cost perceptions and, therefore, heighten the chance of participating in the more risky types of activism like civil disobedience (Ayanian & Tausch, 2016; Carlsen et al., 2021b; Gundelach & Toubøl, 2019). Finally, we include an index of the respondents’ emotional reaction to the events in September 2015 summarising their emotional reactions in terms of (1) compassion with the refugees, (2) feeling responsible for the refugees, (3) anger toward the authorities lack of care for the refugees and (4) ashamed by the lack of a welcoming attitude in Danish society.

Statistical Method: Discrete Graphical Models

For the statistical analysis, we use discrete graphical modelling. The model is based on a recursive block structural model designed to analyse the complex hypotheses regarding a mix of direct, indirect and mediated effects (see Fig. 5.2). The recursive block structure is depicted in Fig. 5.3, and the position of the variables in the structure is based on the theoretical considerations presented above. The logic of a recursive block model consists of the variables in block 1, the numerically lowest block, which are considered dependent on all other variables and the variables in the numerically highest block, in this case, block 4, that are independent of all other variables. The variables in the intermediary blocks are simultaneously independent of the variables of the numerically lower blocks and dependent on the variables in the numerically higher blocks. Thus, recursive block models allow for modelling complex variable relationships and estimating models, including multiple dependent variables.

Fig. 5.3
figure 3

Recursive block structure

Full size image

The block model is analysed as a chain graphical model (Lauritzen, 1996) that combines graphical theoretic analyses with Markov graphs. The statistical analysis is a probability-based adaptation of high-dimensional contingency tables based on the principles of classical elaboration analysis (Aneshensel, 2012; Davis, 1971; Lazarsfeld & Rosenberg, 1955; Rosenberg, 1968). The strategies and techniques of discrete graphical modelling and the DIGRAM softwareFootnote 5 are described and developed by Kreiner (Kreiner, 1986, 1987, 1996, 2003).

In this paper, the order of the recursive blocks is based on the assumptions that are sketched in Fig. 5.2. The focal variables corresponding to the theoretical model of Fig. 5.2 are placed in blocks 1, 2 and 4 of Fig. 5.3. Block 1 contains the dependent variables, loss of institutional trust and political views of favouring more refugee-friendly policies. Block 2 includes the three kinds of activism. In block 4, we find the value scales of altruism and egoism and measures of religious affiliation that we operationalised as proxies of conformity. Also, sociobiographic and socioeconomic controls are in block 4. The three control variables concerning the history of activism, embeddedness in the movement and emotional response are placed in an intervening block 3 as they might constitute contexts that mediate the value predispositions’ relationship to the activity variables. The reason for this is chronology. Values are considered rather stable and, therefore, are most fundamental to the model. The movement embeddedness and history of activism variables concern the period before September 2015, and the emotional response concerns the dramatic events that started the mobilisation in September 2015. Therefore, they are chronologically before the activism variables, which measure participation during and after September 2015.

The advantage of the table elaboration techniques offered by DIGRAM is that it provides a nuanced analysis where the results may be presented visually. In contrast to the traditional regression type of analysis, the variable may be nominal, ordinal or interval scale properties, and it is possible to include several dependent variables, to combine directed and symmetrical relationships between the variables and to include all variables as well as interactions in the analysis from the beginning. The method has two limitations compared to regression analysis. First, the method does not make it possible to compute R2 or any other goodness-of-fit criteria to select among models. Instead, the analysis continues until the researcher finds that the results are empirically and theoretically satisfactory. This kind of interplay between explanatory ideas and the examination of data is at the heart of data analysis (Tufte, 1974) and takes places in all kinds of multivariate analysis. Practitioners of regression analysis also ‘play’ with the data as an integral part of their research activity but rarely explicate their procedures and primarily justify their choice of model from goodness-of-fit tests (Ron, 2002). Instead, the DIGRAM software forces the researcher to make the interplay explicit rather than relying on strict formal criteria for the model’s fit to the data. The lack of a formal measure of goodness of fit leads to another limitation. The inclusion of all direct and indirect relationships between the variables results in a need to study a high number of relationships. When the model includes a large number of variables, the analysis may seem too complex and even incomprehensible. Therefore, the DIGRAM researcher—rather than performing data dredging (Bartels & Brady, 1993)—must limit the number of variables in the model and include variables only where theoretically plausible hypotheses between the variables may be explicated.

The overall strategy of analysis consists of three steps:

  1. 1.

    Based on theoretical considerations, the researcher determines the recursive block structure of the variables as done in Fig. 5.3.

  2. 2.

    Using log-linear analysis, the relationships between all variables are tested for conditional independence. The analysis depends on the collapsibility properties of log-linear and graphical models, which means that estimations of correlations and computations of test statistics may be computed in smaller marginal tables (Agresti, 2013; Kreiner, 1998). If two variables are conditionally independent given all the other variables in the model (i.e., they are partially uncorrelated), the relationship is deleted from the analysis. This changes the characteristics of the model, and the analysis is repeated in search of new cases of uncorrelated variables until all insignificant relationships are deleted. The search for an adequate model is done stepwise in a researcher-supervised semi-automated manner by deleting and adding associations to the model, based on both empirical test results and theoretical subject matter knowledge. The level of significance is tested by using Monte Carlo estimates of exact conditional tests and is assessed by taking the multiple tests performed into account. The final model, thus, includes only highly significant and/or highly theoretically relevant associations.

  3. 3.

    To measure the strength of conditional association and as test statistics to evaluate hypotheses of conditional independence, γ coefficients are used for ordinal variables, and χ2 tests are used for nominal variables.

The final model includes direct as well as mediating relationships between the blocks. We present it graphically in Fig. 5.4 in the subsequent section, where we detail the results of the empirical analysis of the five hypotheses.

Fig. 5.4
figure 4

Final model of the variable relationships of the recursive block structure

Full size image

Results

The five hypotheses can be ordered in two sets, which will structure the presentation of the results. The first set consists of hypotheses 1 and 2 and concerns value predispositions’ indirect effects through the contexts of the three forms of activism they may be activated in. The second set comprises hypotheses 3, 4 and 5 and regards how different kinds of activism imply interaction with different political institutions, which may lead to a loss of trust in particular institutions.

The results of the analysis are depicted as a graph in Fig. 5.4. Lines with arrows indicate the direction of the asymmetric variable relationships, whereas other lines show symmetric relationships. Only statistically significant relationships are reported. Coefficients are partial γ-correlations. To be clear, the dataset does not allow us to observe causal relationships, and we do not intend to make any causal claims based on the results. When we discuss the direction of asymmetric relationships, this is based on theoretical considerations.

Value Predispositions and Participation

The first set of hypotheses concerns which value predispositions are associated with participation in certain kinds of movement activity. Hypothesis 1 stipulated that self-transcendence values would positively associate with participation in humanitarian activities and political protest. In contrast, hypothesis 2 predicts that conformity would be negatively associated with political protest and civil disobedience. The results support both hypotheses. Self-transcendence values are positively associated with participation in both humanitarian activities and political protest. The non-conform respondents without religious affiliation are more likely to participate in political protest, whereas the passive church-affiliated Christians are very unlikely to participate in civil disobedience. Being an active church member indicates a high level of conformity which increases the relative likelihood of participating in humanitarian activity compared to participating in civil disobedience and political protest.

Theoretically, the relationship between values, contexts and action can be interpreted in two ways, both of which may be correct. The first theoretical interpretation assumes that values are prior to and influence action. The subject of refugee solidarity and refugee rights in the context of the movement is evaluated differently based on the individuals’ value hierarchy, which leads to different courses of action. Here the non-conform, self-transcendent person will construct the issue as a politically contentious struggle about justice for refugees, and the natural course of action is political protest and, subsequently, in rare cases, even civil disobedience. The conform person might construe the situation as concerning the suffering of unfortunate refugees but ignore the political level out of the basic propensity to trust and respect the authority of political institutions. In that case, the natural course of actions focuses on humanitarian activities that aim to reduce the suffering of the refugees.

In the second interpretation, the movement activities are not actions per se but are seen as the contexts that activate values. Here, the results tell a story of what values are likely to be activated in what contexts of the movement activities of humanitarian activities, political protest and civil disobedience, and the relationships between the variables are symmetric rather than asymmetric. Values do not lead to certain courses of action; rather, in certain contexts, certain values are activated.

Both interpretations are consistent with the theories, and both processes may be involved in generating the observed correlations. Our data, unfortunately, do not allow for separating the processes, assessing their relative validity and drawing firm conclusions.

Participation and Attitudes

The second set of hypotheses concerns the relationship between movement participation and attitudes. Hypotheses 3 and 4 concern what kind of interaction with political institutions the different activities involve. The hypotheses stipulated that political protest would involve interaction with partisan institutions, causing a loss of trust, and civil disobedience would imply interaction with order institutions resulting in a loss of trust in these institutions. The results support both hypotheses. Political protest has a positive correlation with loss of trust at 0.15, and civil disobedience has a slightly stronger association of 0.20. Also, the strong symmetric association of a 0.48 partial correlation supports the hypotheses’ claim that the two kinds of activities are entangled. Compared to political protest, civil disobedience has a slightly stronger correlation with the loss of institutional trust scale where 1 is the partisan institution of the Parliament and 2 and 3 are the order institutions of the legal system and the police. Thus, civil disobedience’s slightly stronger correlation with loss of institutional trust supports the hypotheses that civil disobedience to a higher degree than political protest involves interaction with the order institutions of the repressive state apparatus and the associated loss of trust.

Our final hypothesis, 5, concerning a change in political opinion, is not supported. Neither participation in political protest nor civil disobedience is associated with a higher likelihood of change in political view regarding immigration.

Conclusion and Discussion

In this study, we have explored the role of value predispositions for participation in activism and individual-level outcomes of movement participation in the form of attitudinal change. In line with mainstream theory on the value-attitude-action nexus, we find that values’ possible influence on outcomes of attitudinal change is mediated by the contexts of the actors: It is through the process of making meaning of the given context that certain values are activated and related to the issues at hand through a process of attitude formation. We show that certain values relate to certain specific contexts of movement activity, either because they are activated in these contexts or because they inform attitudes that call for such action. In our case, the Danish refugee solidarity movement, we find that participation in contentious activism of political protest and civil disobedience, on the one hand, are both related to values of self-transcendence and non-conformity and, on the other hand, to outcomes of loss of trust in political institutions like Parliament, the legal system and the police. For the non-contentious context of humanitarian activities, however, we do not observe any relationships to outcomes of attitudinal change.

These findings give reason to reconsider how we think of values and moral implications of movement participation and, more generally, take the context of values more seriously, both empirically and theoretically. The study goes beyond the notion that certain movements relate to certain values (Klandermans, 2015; Snow et al., 1986) and expands our knowledge by exploring how, within the same movement, different contexts of activism relate to values in different ways and, in turn, relate to different attitudinal outcomes.

Furthermore, the findings suggest that movements may produce heterogeneity within themselves because activists participating in different parts of the repertoire constituting different contexts develop different attitudes to the issue and political actors with which and whom the movement engages. This, in turn, implies that the same movement may contain different competing values that denote different desirables that guide action. Thus, one of the implications is to take a more nuanced approach to movements and recognise that the often noted heterogeneous, unorganised and diverse composition of movements is a factor of consequence.

Because of its focus on how to bridge and align different frames understood as interpretive schemata that guide meaning-making, framing theory (Snow et al., 1986) presents itself as a relevant approach to address these issues. To strengthen a focus on rank and file members’ practices, which calls for situation centred, ethnographic approaches, the concept of scene style (Lichterman & Eliasoph, 2014) readily presents itself as an important supplement. This concept invites us to investigate and explore how different contexts of movement participation can be understood as different scenes constituted of interactional styles. The cultures of different scenes bring certain values into play and relate them to various objects and issues that produce scene-specific logics of action and habits of thought, shaping the moral selves of the involved actors (Carlsen et al., 2021b).

The observed heterogeneity of values and attitudes in different within-movement contexts also complicates the consequences of movement activity for society’s basic values and moral foundations (Alexander, 2006; McAdam, 1988). In the case of the Danish refugee solidarity movement, a central internal issue of strife was whether the movement’s goal was to change institutional politics pursued through contentious practices or whether it should focus only on humanitarian activities pursuing a prefigurative political strategy (Carlsen et al., 2020a; Vandevoordt, 2019). This debate was framed as a within-movement debate, and the participants recognised each other as belonging to the same movement (Toubøl, 2017, pp. 54–57). This within-movement division along the dimension of contentiousness (Carlsen & Toubøl, 2021) is also clearly observed in our analyses. It raises the question of whether these observed variations within the overall collective identity of the movement also result in different moral visions for society. While the humanitarian activities of the movement were not associated with any of the attitudinal outcomes included in the analysis above, it still seems likely that such activities involved collective identity formation that may serve as a moral template or vision, not just for the activists involved but also in the wider society (Melucci, 1989).

The insight that what goes on at the interactional level in different parts of a movement has implications for values and attitudinal outcomes is relevant for value studies in general. Empirical research into values is dominated by survey studies. However, the survey method suffers from its long-standing weakness with regard to measuring respondent contexts (Barton, 1968; Carlsen et al., 2021a; Cicourel, 1964; Coleman, 1958), which results in an epistemological paradox in the sense that theories of values stress the consequential role of contexts, but empirical studies rarely include reliable measures of context. While this study also is limited when it comes to measuring contexts, it does provide evidence that this omission is problematic. Hence, studying and conceptualising the shaping of the moral self in different contexts and situations are pending tasks for students of values, attitudes and activism. This task is further complicated by the fact that movement contexts—both in the sense of contexts within the movements and the political institutional and cultural contexts that the movement is embedded in—are not constant, but dynamic (e.g. Tilly, 1986). While this approach is beyond the capability of the present paper, which is based on cross-sectional data, it is a highly relevant perspective for future studies based in longitudinal data to grasp such dynamic developments of contexts and their implications for individual-level participation and shaping of the activists’ morality.