Normative, rational, and relational motives in crosscoalition coordination for a VET refugee programme

Abstract Shared beliefs are seen as a basis for policy coordination in the literature. Actors sharing beliefs coordinate their activities in order to translate their beliefs into policies. However, the literature shows that actors also coordinate for policy change across such belief coalitions for diverse reasons. Drawing on the literature on incentives in collective action organisations, we systematise these motives. We argue that rational motivations, such as access to material resources, as well as relational motivations, including power and reputation gains, may convince actors to coordinate. Based on 25 semi-structured expert interviews, we illustrate our propositions with a case study on the motivations that led actors to coordinate and support a vocational education and training (VET) programme for refugees in Switzerland. Coordination between a coalition of VET actors and a coalition of migration actors succeeded despite divergent policy beliefs, mainly due to rational motivations.

Shared beliefs are seen as a basis for policy coordination in the literature. Actors sharing beliefs coordinate their activities in order to translate their beliefs into policies. Perhaps most prominently, the advocacy coalition framework (ACF) explains how actors sharing normative beliefs coordinate in order to render these beliefs into policies (Jenkins-Smith and Sabatier 1993;Sabatier 1988Sabatier , 1993Weible et al. 2009). Following the ACF, coalition members' beliefs are structured hierarchically: Fundamental ontological and normative beliefs (deep core beliefs) constrain mid-level policy beliefs and more specific or operational beliefs (secondary beliefs) (Jenkins-Smith et al. 1991, p. 852). These lowest-order secondary beliefs include rational concerns and are attributed little significance.
However, the literature analysing crosscoalition coordination argues that actors coordinate for other reasons than shared beliefs to promote a policy. Resource access, actors' expertise and trustworthiness, as well as their power in the policy process, play a role in deciding whether or not to coordinate (Calanni et al. 2015;Heinmiller and Pirak 2016;Ingold et al. 2017;Koebele 2020;Henry 2011;Henry et al. 2010;Ingold 2011;Sandström 2011, 2013;Weible 2005). Consistent with these contributions, we argue that actors engage in crosscoalition coordination for diverse reasons. Drawing on the literature on incentives in collective action organisations, we group the motivations for actors to coordinate in order to promote policy change into three categories: normative beliefs, rational considerations, and relational motives (Clark and Wilson 1961, p. 133;Knoke 1988, p. 315;Puffer and Meindl 1992, p. 425).
We find support for the relevance of these motives by analysing a case of crosscoalition coordination involving actors adhering to very different beliefs: The introduction of a training programme meant to integrate refugees into Switzerland's standard vocational education and training (VET) system. This policy change required a coalition of VET actors to coordinate its actions with a coalition of migration actors despite their divergent beliefs. The 1-year preparatory VET programme for refugees (Integrationsvorlehre, INVOL) was initiated by the federal migration authority (State Secretariat for Migration, SEM) in 2018. Initiated by a migration coalition, the programme was strongly opposed by VET actors. However, policy implementation required active participation from VET actors.
The federal migration authority managed to convince a number of VET actors (professional training organisations (PTOs), cantonal VET authorities) to support the programme. We trace the emerging coordination between the two coalitions, which eventually enabled refugees to enter Switzerland's standard vocational training system and labour market. We draw on 25 semi-structured expert interviews, conducted before and after programme implementation, with federal VET and migration authorities, cantonal VET authorities, and PTOs. Focussing on the dynamics of coalition formation, we found that rational incentives, such as financial resources convinced actors to coordinate. Relational motives including power, and reputational gains also played a role, albeit a minor one.
We make three contributions to the policy literature. First, we systematise the types of motives that actors might pursue when coordinating with other actors in order to attain a policy goal. We do so by connecting the ACF to the literature on incentives in collective action organisations. We provide evidence that, apart from shared beliefs, rational motives (financial resources) and relational motives (power and reputational gains) foster crosscoalition coordination.
Second, we examine which actors pursue which motives. We build on a typology developed in various contributions by Jenkins-Smith and Sabatier, who distinguish material interest groups (e.g. business associations) pursuing rational motives from purposive groups (e.g. environmental interest groups) pursuing normative beliefs and also from more mercurial government actors (Jenkins-Smith and Sabatier 1994, p. 196;Jenkins-Smith et al. 1991, p. 852;Sabatier 1998, p. 116). We suggest that formal decision-making power rather than actor type enables distinguishing actor motivations.
Third, our case study offers insight into the interaction of different types of motivation. We find that coordination among coalitions based on rational motives might be short-lived because actors ultimately compete for resources. Sustainable coordination might therefore rely on shared beliefs. Nonetheless, alternative motives, such as rational interests in material resources, or relational motives, such as improving one's reputation and maintaining one's power over others, might convince opposed actors to engage in a first instance of coordination.
The remainder of our article is structured as follows. First, we discuss the literature on crosscoalition coordination, structuring it based on the literature on incentives in collective action organisations. We deduce five propositions about actor motivation to engage in coordination. Second, we present our case and data. Finally, we discuss the implications of our findings and suggest avenues for further research.
Theory: Normative, rational, and relational coordination motives Shared beliefs are seen as a basis for policy coordination in the literature. Actors sharing beliefs coordinate their activities in order to translate their beliefs into policies. Perhaps most prominently, the ACF explains how actors sharing normative beliefs coordinate in order to render these beliefs into policies Sabatier 1988Sabatier , 1993Weible et al. 2009). As mentioned, coalition members' beliefs are assumed to be structured hierarchically in the ACF. Fundamental ontological and normative beliefs constrain more specific or operational beliefs (Jenkins-Smith et al. 1991, p. 852). First, deep core beliefs are the fundamental normative and ontological principles constituting a person's philosophy. Examples of such deep core beliefs include perceptions of human nature, various ultimate values such as freedom or security, and basic criteria of distributive justice (Sabatier 1988, p. 144). Changing deep core beliefs is similar to religious conversion and very difficult to achieve (Sabatier 1988, p. 144).
Second, policy core beliefs (i.e. intermediate beliefs) determine the strategies and positions adopted by actors to implement their deep core beliefs in related policy areas (Sabatier 1993, p. 29ff). Such policy core beliefs include the proper scope of governmental versus market activity, the proper distribution of authority among various governmental units (or levels), and the basic choices concerning policy instruments (Sabatier 1988, p. 144). Although actors do not easily change their policy core beliefs, these may be adaptedunlike deep core beliefs.
Third, secondary beliefs determine the instrumental decisions required to implement policy core beliefs in a specific area (Sabatier 1993, p. 29ff). Such secondary beliefs may, for instance, concern decisions about administrative rules, budgetary allocations, statutory revisions, as well as information about programme performance and problem seriousness (Sabatier 1988, p. 144). These secondary beliefs can be changed comparatively easily. The role of secondary beliefs in informing actors' choices in the ACF is limited, since such beliefs are constrained by policy core beliefs (hierarchical structure of a belief system). Following the ACF, actors favour their core beliefs if core beliefs and secondary beliefs conflict.
The existing literature has examined cases where actors coordinate across different belief systems. These cases suggest that alternative motives convince actors to coordinate and contribute to a common policy goal (Calanni et al. 2015;Heinmiller and Pirak 2016;Ingold et al. 2017;Koebele 2020;Henry 2011;Henry et al. 2010;Weible 2005). We group the motives found in this literature into three broad categories: normative considerations, rational motives, and relational motives. We base this categorisation on the literature on incentives in collective action organisations (Clark and Wilson 1961, p. 133;Knoke 1988, p. 315;Puffer and Meindl 1992, p. 425).
The first category is normative beliefs, which correspond closely to the mid-level policy core beliefs in the ACF. This type of motivation refers to actors' adherence to standards, their tendency to follow conduct grounded in socially instilled values about principled behaviour, or to the enactment of certain laws (Clark and Wilson 1961, p. 133;Knoke 1988, p. 315;Pecorino 2015, p. 241). Most studies analysing alternative coordination motives also confirm the relevance of shared beliefs (Fischer et al. 2010;Heinmiller and Pirak 2016, p. 180;Jasper 2004, p. 8;Kim and Roh 2008, p. 674;Koebele 2020, p. 743;Weible and Ingold 2018, p. 333;Ingold 2011;Sandström 2011, 2013). Thus, although alternative motives matter, normative beliefs remain important. We hence suggest that actors' distance from the other coalition's core beliefs influences whether actors engage in crosscoalition coordination.
Proposition 1: The closer actors are to another coalition's core beliefs, the more likely they engage in cross coalition coordination.
Second, actors pursue rational motives and compare costs and benefits. We define rational motives as directed at pursuing private, exclusionary goods. Rational motives may for instance include access to material benefits (Knoke 1988). The ACF introduces certain rational motives under "secondary beliefs" (e.g. "budgetary allocations" correspond to financial resources). However, following the hierarchical nature of belief systems actors are unlikely to compromise their higher-order beliefs in order to pursue rational motives. This results in an uneasy relation between rational motives and normative beliefs, which has been discussed in the ACF literature (Luxon 2019;Schlager 1995). On the one hand, values and beliefs take centre stage in the ACF. Individuals and coalitions are assumed to orient their actions toward ideals and beliefs (Ameringer 2002, p. 547). Rational motives correspond to secondary beliefs and thus are merely a subcategory in the belief system so that actors are not driven primarily by simple economic or political incentives (Nohrstedt 2010, p. 317). On the other hand, however, the same actors are also assumed to act rationally (Ingold and Varone 2012;Nohrstedt 2010, p. 312;Sotirov and Winkel 2016;Winkel and Sotirov 2011, p. 145). If an external event increasing resources for one coalition leads to policy change, these resources are evidently very important. Actors may also be assumed to care about these resources and not merely about their beliefs (Nohrstedt 2010, p. 312). This tension between normative beliefs and rational motives has been observed in a number of ACF applications. It is unclear when individuals pursue ideals and when they act based on more rational motives (Haelg et al. 2019;Luxon 2019, p. 108).
In contrast, the literature on crosscoalition coordination has found that diverse rational motives are relevant, including tangible ones such as material resources and intangible ones such as professional competence and expertise (Calanni et al. 2015;Koebele 2020;Weible et al. 2018, p. 6). In the case of crosscoalition policy coordination, various rational motivations have been found to play a role. Expertise has been found to be particularly important when deciding to coordinate (Calanni et al. 2015;Heinmiller and Pirak 2016;Koebele 2020). The role of financial resources in crosscoalition coordination is less clear-cut (Calanni et al. 2015;Cappelletti et al. 2014;Kim and Roh 2008, p. 680;Weible et al. 2018, p. 14). Hence, we are particularly interested in the role of financial resources, which leads to proposition 2: Proposition 2: Actors coordinate across coalitions because of rational motives, such as financial resources.
Third, actors follow relational motives. These motives differ from rational ones, as they are not directed at pursuing private, excludable goods but rather at the company or the approval of others, or, in the case of power, at their subordination. Relational motives include the need to form emotional attachments with others (Clark and Wilson 1961, p. 133;Knoke 1988, p. 315;Puffer and Meindl 1992, p. 428). Emotional attachment may be less important for organisational actors. Another example of a relational motive may be gained in power. As a rule, actor hierarchy solves a collective action problem as a powerful actor can simply impose their preference on others (Elster 1989, p. 40). Hence, joining a powerful actor increases one's chances of implementing one's policy ideas. According to the literature on crosscoalition coordination, actors prefer coordinating with powerful actors (Calanni et al. 2015;Heinmiller and Pirak 2016). Complementing the literature on crosscoalition coordination, we are interested in the preferences of powerful actors. It may be difficult to convince such actors to coordinate because they may not need additional actors to carry through their beliefs. Power has been conceptualised in different ways, corresponding to various aspects important in policy making: access to legal authority, public support, the ability to mobilise supporters, and leadership (Weible and Ingold 2018, p. 330). In our study, we define power as decision-making authority over a certain domain.
Proposition 3: Powerful actors, i.e. actors with decision-making authority are less likely to engage in crosscoalition coordination.
Relational motives may also include reputation conferred by others, convincing actors to coordinate and contribute to a collective (Clark and Wilson 1961, p. 133;Knoke 1988, p. 315;Puffer and Meindl 1992, p. 428). Gaining a positive reputation greatly depends on others, since their affirmation of positive qualities is much more credible. Building up a reputation with other, external partners is therefore a prominent motive for relating to and coordinating with others (Aerne 2020; Lazega and Pattison 1999;Raub and Weesie 1990). In addition, coordination is facilitated by reputation: knowing that others are reliable partners is important when selecting whom to coordinate with (Dijkstra and Assen 2013;Granovetter 1985;Macy and Skvoretz 1998;Raub and Weesie 1990;Tomochi 2004).
To our best knowledge, reputation has not been considered in previous studies on crosscoalition coordination as a coordination motive. It might, however, be relevant, as a positive reputation legitimises an organisation not only through internal participants but also through stockholders, the public, and the state (Gordon and Babchuk 1959;Meyer and Rowan 1977). In particular, gaining reputation has been identified as a relational motive for public organisations, to which policy implementation is delegated (Bertelli and Busuioc 2021, p. 44), and also for economic interest groups, such as employer associations (Schmitter and Streeck 1999, p. 86). This leads to proposition 4: Proposition 4: Actors coordinate with another coalition in order to gain reputation.
The literature has furthermore differentiated coalitional actors according to how consistently they pursue their beliefs. Material interest groups (e.g. business federations) differ from purposive groups (e.g. environmental groups) and also from government actors (Jenkins-Smith and Sabatier 1994, p. 196;Jenkins-Smith et al. 1991, p. 852;Sabatier 1998, p. 116). In theory, material interest groups promote policy goals less consistently. They struggle to convince coalition members to contribute to a common goal and collective welfare because coalition members are profitoriented and care little about shared beliefs (Sabatier 1998, p. 116). Accordingly, material interest groups are more flexible in their policy positions: As long as there is a financial gain, coalition members, and hence the coalition itself, are open to changing their policy position. The second type of actor (purposive groups) attracts members based on an idea or shared belief, and pursues goals benefiting some wider public or even society at large (Jenkins-Smith et al. 1991, p. 852). Accordingly, the members of these purposive groups are motivated by beliefs, which makes it difficult for the coalition to deviate from the pursued policy positions (Jenkins-Smith et al. 1991, p. 852). Government actors are described as "mercurial" in their support for certain policy positions. Since administrative bodies respond to changes in government, their position changes with electoral outcomes as well as with nonelectoral, exogenous events, which increase problem pressure (Jenkins-Smith et al. 1991, p. 873). Evidence of this typology seems mixed with respect to crosscoalition coordination. Financial resources are more important to government actors and purposive groups (Cappelletti et al. 2014;Weible et al. 2018, p. 14). This leads to proposition 5, which examines how the influence of rational incentives differs across actor types.
Proposition 5: Material interest groups are more likely to be induced to coordinate across coalitions by rational motives, such as financial resources, than purposive groups and governmental actors.
In summary, while shared beliefs play a role in crosscoalition coordination, they are not the sole reason for actors to engage in coordination. Other rational motives include access to resources. The importance of these rational motives likely varies depending on the functional role of policy actors. In addition, relational motives (e.g. power and reputational gain) may also motivate actors to coordinate across coalitions.

Context, case, and data
Context VET is the backbone of the Swiss education system. Sixty percent of young people attend a dual training programme at upper secondary level. Such VET programmes combine school-and workplace-based learning. Typically, Swiss youths in VET spend 3 days a week at a workplace and 2 days in school. Accordingly, VET systems involve employers and their associations on one side and the state on the other (and partly also labour unions) in the provision, financing, administration, standardisation, and reform of vocational education (Busemeyer and Trampusch 2012, p. 351).
In Switzerland, three VET actors are of central importance. First, formal responsibility for VET programmes lies with the State Secretariat for Education, Research and Innovation (SERI). Second, the prevalence of dual training in Switzerland means that professional training organizations (PTOs) play a central role in VET. These are often industry associations, which design curricula and final examinations (subsequently authorised by the federal authority for education). Third, the cantonal VET authorities oversee individual training contracts and operate VET colleges, whose courses complement practical training.
In 2015, Switzerland faced a massive influx of refugees. In response, additional integration measures were adopted. The most significant and also the most controversial policy change was the so-called integration preapprenticeship, a 1-year preparatory VET programme combining workplace training with classroom instruction.
Despite being a dual training programme with workplace-and school-based learning, the integration preapprenticeship presents a major policy change in the Swiss education system. Untypically, the programme was aimed exclusively at a particular group of trainees (refugees). Instituting a preparatory training track lowers the entry barriers for certain groups and potentially diminishes the selectivity and prestige of the VET system. Constituting a major policy change, the integration preapprenticeship was opposed by all VET actors, whose support was needed, however (see our case study in Section 4).

Case description
We analysed how coordination emerged between two coalitions jointly advocating the new VET refugee programme. Eventually, the programme was supported by two coalitions: VET actors on the one hand and migration actors on the other. These two coalitions held quite different policy core beliefs and initially opposed each other.
Although initiated by the State Secretariat for Migration (SEM), the policy hinged on the contribution of VET actors. Both, cantonal VET offices and professional training organizations were free to participate in the programme or not. Surprisingly, a majority of cantonal VET offices (18 out of 26) and a sufficient number of PTOs (12 out of 154) supported the programmedespite initial opposition. This allowed implementing the programme and granted around 600 refugees a training place in 2018. The professional training organizations (PTOs), together with the cantonal VET offices, designed the curricula and also motivated individual employers to accept refugees.
Our case exhibits various ACF characteristics. First, a major policy change occurred in response to an external event (a refugee crisis). Second, two opposing policy beliefs clashed. VET actors wanted to keep the VET system selective in order to protect its prestige, whereas migration actors sought to award refugees the necessary credentials to be successful on the Swiss labour market. Third, crosscoalition coordination between migration and VET actors emerged and eventually led to policy change. Fourth, a policy broker mediated between the responsible federal migration authority (SEM) and the VET actors: the Swiss Federal Institute for Vocational Education and Training (SFIVET), a federal institute conducting research on VET, advised the federal migration authority (SEM) on programme design and implementation.
The federal VET research institute (SFIVET) played a crucial role in facilitating coordination between the two coalitions. It contributed to the successful coordination between the two coalitions mainly in three ways. First, it made sure that the federal migration office used the right wording when talking to the VET coalition (SVIFET representative, Zollikofen, 1 April 2019). For instance, the name of the programme had led to serious irritations on the side of the VET coalition. Naming it preapprenticeship generated fears that regular apprenticeship tracks might come to be associated with this additional, less demanding programme (Basel City representative, Basel, 20 February 2019, Thurgau representative, Frauenfeld, 13 November 2018, Swissmem representative, Winterthur, 29 January 2019). Second, the federal VET research institute pushed the VET coalition to provide enough occupation-specific content so as to facilitate refugees' labour market integration (SVIFET representative, Zollikofen, 1 April 2019). There was a tendency to organise generic programmes with some work-related component that would however not prepare candidates for specific occupations so as to keep the access to occupations selective. Third, the federal VET research institute made sure that the selection of candidates into the programme was not overambitious, as the VET coalition preferred a very strict admission into the programme (SVIFET representative, Zollikofen, 1 April 2019).
Some characteristics, however, are less typical of ACF application. First, as discussed in detail in the next section, actors were far from sharing a common belief. Second, the programme is concerned with integration policy. Benefiting a very specific group, integration policy might resemble redistributive policies, which are particular with respect to coalitional work (Leifeld 2013). Third, the programme involved policy change in Switzerland, a consociational and corporatist democracy, whereas the ACF was originally developed for a pluralist setting (Ingold 2011). However, a number of ACF applications have demonstrated its value also in the area of redistributive policies and in consociational settings (Sabatier 1998).
Fourth, the time period under study was shorter than in most ACF applications. Typically, ACF scholars hold that analysing policy change requires observing a policy subsystem over a decade (Sabatier 1993, p. 16). Observing coalitional learning requires decades rather than years. We challenge the idea that coordination occurs merely as a consequence of shared beliefs. Observing longer time intervals likely means overlooking rational and relational motives, which may change more quickly and are important for stimulating coordination. We therefore zoom in on the relatively short period (2015-2019) in which policy change happened. We analyse how much coordination in our case can be attributed to shared beliefs and to what extent rational and relational motives were relevant.
Fifth, we analysed implementation in addition to policy design. In the VET domain, implementation is perhaps the most difficult and controversial phase. VET programmes depend on voluntary and active contributions from employers and their associations. Consensus during policy design is not necessarily indicative of the degree of agreement or the coordination needed for the VET programme to succeed beyond the paperwork. The capacity of state authorities to force employers to offer vocational training is limited. Under such circumstances, training might lead to trainee exploitation. Instituting collective skills training is challenging because employers fear that the returns on their investment will be reaped by other companies when their apprentices leave their training firm (Culpepper 2000, p. 224). Consequently, employers invest in training only reluctantly. Our case is further complicated by training being extended to refugees. Presumably, training refugees is more costly than training local youths and potentially increases the losses incurred as a result of poaching.
Moreover, the federal migration authority (SEM) decided to move to policy implementation in order to gain support for its programme. Alternatively, migration actors could have launched a parliamentary motion demanding such a programme or a popular initiative, which would have increased public pressure. However, the federal migration authority (SEM) chose to move directly to the local level in order to implement the programme. This explains our decision to trace policy design and implementation, as well as the discourse surrounding these stages.

Sampling strategy and data
Identifying relevant actors, their beliefs, and corresponding coalitions in a policy subsystem is far from straightforward. We faced the additional difficulty of separating policy beliefs from actors' alternative motivations. We addressed this difficulty in three steps. We first identified the coordinating actors and second their coordination motives. Consequently, we triangulated our interview data from coordinating actors with interviews with actors not coordinating with the migration coalition.
As mentioned, we first identified the coordinating actors. Our initial sample of interviewees thus consisted of key actors (i.e. those engaging in coordination in order to support the programme): the federal migration authority (SEM), cantonal migration offices, the cantonal VET offices participating in the programme, and professional training organizations (PTOs). Interviews lasted between 45 minutes and 2 hours. All interviews were transcribed and coded by hand by the same person, with the help of atlas.ti.
In a second step, we examined actors' beliefs and motivations and found that they were split along two core beliefs. Migration actors favoured refugee integration regardless of the potential effect on the training system whereas VET actors favoured maintaining the selective nature of the VET system over refugee integration. We thus established that this was a case of crosscoalition coordination where coordinating actors were far from sharing beliefs. Rather, additional motives were important in convincing actors to engage in crosscoalition coordination. To better understand actors' underlying motivationstheir beliefs as well as their concerns about resources, power, or reputationwe decided to code the interviews a second time. In looking for patterns that explained actors' motives for coordinating with the migration coalition despite opposing core beliefs, we focussed on the theorised alternative motives: rational motives (material resources) and relational motives (power and reputational gains) (for a code book, see Supporting Information).
Having found diverging beliefs in the policy coalition, we expanded our sample. We included VET actors who could have implemented the programme but did not. Including VET actors who did not promote the programme ensured that the divergence of the coordinating actors' beliefs was not coincidental. Rather, actors who did not support the programme shared the same concerns as VET actors supporting the programme but were more adamant about their beliefs. Actors who did not support the programme also explained their policy preferences more openly.
In total, we conducted 25 interviews. On the federal level, we interviewed representatives of the State Secretariat for Education, Research and Innovation (SERI) and of the State Secretariat for Migration (SEM). On the cantonal level, we interviewed the VET offices of nonimplementing (Grisons, Thurgau) and implementing cantons (Basle-City, Berne, Geneva, Jura, Lucerne, Neuchâtel, St.Gallen, Vaud, Valais, and Zurich). We selected the cantons to maximise variation on several dimensions known to play an important role in VET. The first dimension is language, as vocational training is more prevalent in German-speaking Switzerland. We selected German-speaking cantons (Basle-City, Berne, Grisons, Lucerne, St.Gallen, Thurgau, and Zurich) and French-speaking ones (Geneva, Jura, Neuchâtel, Vaud and Valais). Three cantons (St.Gallen, Zurich and Jura) were among those in which vocational training is most common. In contrast, Vaud and Geneva tended to rely more on academic education. We also varied canton size, from the largest (Zurich) to one of the smallest (Jura).
We also conducted interviews with 8 out of 12 professional training organizations (PTOs) supporting the programme. Four professional training organizations (PTOs) preferred not to be interviewed. At the federal offices (SEM, SERI), and at most cantonal VET offices and professional training organizations (PTOs), only one person was responsible for the programme. We interviewed these officials (for a detailed list of interviewees, see Supporting Information).

Case study
This section shows that the actors coordinating with one another did not share policy beliefs. Next, it traces the reasons why VET actors supported the policy nonetheless and considers our propositions (Section 2): actor distance from core beliefs, rational motives (financial resources) incentivizing collective action, and relational motives (power and reputation).

Diverging beliefs of the two coordinating coalitions
Our interviews revealed that coordinating actors had very different ideas about the labour market integration of refugees. In essence, two core beliefs about programme implementation clashed: providing refugees with the credentials needed to succeed on the Swiss labour market and maintaining the Swiss VET system selective and prestigious.
Despite offering dual training, the proposed programme was strongly opposed by all VET actors: the federal VET office, cantonal VET offices, and professional training organizations (PTOs). Their opposition was based on two concerns. Originally, the federal migration office (SEM) had proposed a programme that would lead to a nationally recognised diploma. The federal VET office considered this as a risk. If a considerably shorter programme, so its reasoning, led to a nationally recognised diploma, this would threaten the value of existing VET diplomas: A parallel program presented a risk for the regular vocational education tracks, because it was aimed at a specific group yet was ultimately supposed to be equivalent to the regular tracks. (SERI representative, Berne, February 6, 2019) Offering a shorter, parallel training programme might result in disparate diplomaawarding standards, and thereby potentially reduce the uniformity of the certified skills. In the long run, such a scheme might lead employers to depend less on a VET diploma as proof of certain skill levels. It might thus also reduce the ability of VET graduates to find employment. As the federal migration office (SEM) could not introduce the programme against the opposition of the federal VET office, it agreed to a compromise: An additional 1-year preapprenticeship programme would be introduced, yet without a formally recognised diploma. While the federal migration office (SEM) managed to launch the programme, the federal VET office prevented access to a formal VET diploma by introducing a shorter programme.
Not leading to a federal diploma, the alternative programme eased refugee access to the regular VET track rather than to the labour market. Facilitating VET access automatically reduced its selectivity. This in turn sparked fears that VET might lose some of its prestige. The loss of status might be exacerbated by granting access to minority groups at the bottom of societal hierarchies (Hagendoorn 1995). In general, migrants are expected to take up jobs at the bottom of the occupational hierarchy regardless of their previous position in their home societies (Auer et al. 2019). Accordingly, they usually experience considerable downward mobility and accept jobs in their receiving countries that are neither well-paid nor sought-after (Chiswick 2005;Moreh 2014;Piore 1979). Switzerland takes great pride in its occupational skills development system. Facilitating refugee access to particular occupations and to the Swiss vocational education system in general was therefore seen to undermine the prestige of the VET system. Accordingly, VET actors opposed programme introduction, thus fuelling controversial discussions. VET actors were keen to protect their occupations, among other measures, by admitting only the most competitive candidates to their tracks.
The concerns of VET actors about how the programme would affect VET prestige are also clearly evident in the statements of a cantonal VET office that decided against implementation. The Canton of Thurgau preferred to integrate refugees in alternative preparatory classes for 3 years before allowing them to attend regular VET tracks. While these preparatory classes appear to benefit refugees, they prevent them from pursuing an occupation or a VET track. The preparatory classes are structured as a three-stage programme: Refugees take a test at the end of every year and, provided they pass, can advance to the next level (representative of VET Office Thurgau, November 13, 2018). The share of workplace training increases annually (representative of VET Office Thurgau, November 13, 2018). The Thurgau VET Office also advises field instructors and training enterprises to keep the bar high for candidates aiming to enter a regular VET track: [They should] not accept candidates who are not sufficiently strong in either German or Maths. Such candidates should first be sent to integration classes one through three. (Representative of VET Office Thurgau, November 13, 2018) The example of the Canton of Thurgau shows that VET actors sought to maintain admission standards to protect the prestige of occupations. An additional programme, designed to facilitate access, was seen as a threat to the standing of the Swiss VET system.
In contrast, migration actors promoted the programme. They wanted refugees to earn the credentials needed to succeed on the Swiss labour market, where job seekers lacking formal qualifications have very limited long-term perspectives. Candidates accepted into a VET track establish contacts within the local community, are able to earn a living, and do not depend on social welfare (representative of Service des formations postobligatoires et de l'orientation, Neuchâtel, August 20, 2018). Unlike the VET actors, the migration actors were less concerned about the prestige and functioning of the VET system (representative of the federal office of VET, Berne, February 6, 2019). As a cantonal programme implementer in Neuchâtel explained: Those responsible for asylum seekers in social work would of course prefer for us to absorb more candidates, because every time they can place someone in an apprenticeship, then for them it's a win in inverted commas. The difference in orientation of cantonal migration and VET offices becomes perhaps most clear in the case of Basle-City. In Basle-City, the VET office refused to implement the programme due to concerns about the prestige of the VET system. In contrast, the office of social welfare responsible for refugees in Basle-City decided to implement the programme (Representative of the office of social assistance Canton of Basle-City, Basle, 12 February 2019). From the perspective of the social assistance office, holding a VET diploma is primarily a key to long-term integration for refugees. However, for refugees, the way to such a diploma is often difficult. Therefore, it is sometimes preferred to directly integrate refugees into the labour market. The integration preapprenticeship was perceived as a chance to increase refugees' awareness of the use and importance of a VET diploma by exposing them to a workplace while also guiding them to such a VET training. The representative of the office of social assistance in Basle-city described the value of the programme combining work and training as follows: The goal is long-term integration, not fast labor market placement. But the readiness to invest two years to learn the language, to do an internship, to attend two to three years of training is sometimes lacking. ( : : : ) To young adults, the way to a VET diploma seems so long that they hardly ever start it. That's why we say: Education AND work. [rather than education before work, comment by the authors] ( : : : ) (Representative of the office of social assistance Canton of Basle-City, Basle, 12 February 2019) Thus, two policy beliefs clashed. Whereas the VET actors prioritised the quality and prestige of the VET system, the migration actors favoured refugee integration and inclusiveness. These clashing beliefs might also be seen as the same policy confronted from opposite sides. It is difficult to establish how far the two coalitions' beliefs overlapped. However, total oppositionno overlapping policy core beliefs at allmakes coordination even less likely. Below, we trace why VET actors coordinated with the migration coalition.
Coalition heterogeneity: VET actors' distance from opposing core beliefs We found that not all VET actors opposed the programme to the same extent. Convincing VET actors to support the programme was facilitated by the heterogeneity of their policy core beliefs. In line with Proposition 1, VET actors closer to the migration coalition's beliefs were more willing to coordinate. We observed systematic variation between the concerns of the cantonal VET offices and professional training organizations. French-speaking cantons were not as adamant about VET selectivity, nor were occupations that traditionally recruit less academically inclined candidates. Accordingly, the programme was more readily implemented in Frenchspeaking cantons and in occupations recruiting academically weaker school leavers (logistics, construction, hospitality industry).
In French-speaking cantons, vocational education is less prevalent. Compared to German-speaking cantons, a larger share of students earns one of several types of upper-secondary baccalaureates, which enables them to easily continue their education. Thus, the best student admitted to a VET track in a French-speaking canton is less academically inclined than the best student admitted to a VET track in German-speaking Switzerland. Accordingly, the VET tracks in French-speaking Switzerland are slightly less selective than those in German-speaking cantons. VET diplomas are also less esteemed compared to German-speaking Switzerland (Bolli et al. 2018).
Consequently, instituting a preparatory track (potentially undermining VET selectivity) was less opposed by French-speaking cantonal VET offices. This tendency was reflected by programme support. All French-speaking cantons implemented the programme whereas the eight nonimplementing cantons were all located in German-speaking Switzerland. This pattern is even more remarkable given that implementation was probably easier to achieve in German-speaking cantons. In general, these cantons have higher training rates and more dual VETs. Cantons with a strong dual VET tradition are better equipped to implement a collective skills development programme. This finding is confirmed by national PTOs being reluctant to implement the programme in French-speaking Switzerland because they expected an insufficient number of their members to provide training places (VISCOM representative, Berne, August 10, 2018).
In addition, maintaining the selectivity of VET tracks was not equally important in all occupational fields. Prestige hierarchies among occupations arise from a division of labour, which entails unequal access to resources and power (Hodge et al. 1964;Inkeles and Rossi 1956;Nakao and Treas 1994;Treiman 1977;Kraus et al. 1978;Zhou 2005). Some recent studies have established that occupations also enjoy different levels of prestige in Switzerland (Abrassart and Wolter 2020;Bolli et al. 2018).
Tracing programme implementation reveals a pattern that corresponds to the idea that professional training organizations were concerned about the occupational prestige of their tracks to different degrees. Less selective and hence less prestigious occupations were more likely to participate in the programme and invest in curriculum development because they did not fear lowering their occupational standing by facilitating access (Aerne 2021).
In the construction sector, catering and hotel services, cleaning and agriculture, the professional training organizations were very actively designing curricula and recruiting training firms. In these occupations, curricula were rich in occupational, trade-specific skills. Teaching refugees these trade-specific skills facilitates access to a regular training VET track upon completing this specific preparatory programme. Polybau, the professional training organization for construction, developed a detailed curriculum, including a specially designed course at their training centre. It also provided refugees with appropriate shoes and work clothes so they could be employed from day 1 without causing friction in companies (Polybau representative, Uzwil, March 13, 2018). The professional training organization was aware of the varying selectivity and prestige of different training tracks and that it was recruiting from a pool of less academically inclined school-leavers. As its representative explained: Who integrates difficult young people into the workplace? Where do the difficult adolescents go? They go into construction. Those who are not academically talented do not go to commercial school (KV) or become a laboratory chemist. ( : : : ) The academically weaker students go into construction. And academically weaker students are those who are a bit difficult. We have a large share of young people who are hard to educate, and we socialize them. All of these young people who complete their apprenticeship have been socialized by the construction sector. ( : : : ) The construction sector is largely responsible for the successful socialization and integration of difficult youngsters and migrants. This is neither recognized nor supported. ( : : : ) Does the commercial sector participate in the integration pre-apprenticeship? -Certainly not! (Polybau representative, Uzwil, March 13, 2018) The quote shows that professional training organizations are aware of the varying selectivity and prestige of different training tracks. Knowing that it would be welcoming these refugees in its trade sooner or later, this professional training organization in construction industry was prepared to impart essential occupational skills.
By contrast, more selective occupations were more hesitant about providing support and eliminating barriers for refugees to enter their occupation. For instance, the competency profile of mechanics and automation specialists envisages that refugees learn to put a workpiece into a machine, start the machine, remove the workpiece, clean it, measure whether it meets the specifications, and return it to storage (representative of Swissmem, Winterthur, January 29, 2019). This simple cycle does not reflect the complexity of the profession and seemingly amounts to little content for 1 year of training even if refugees are not fully proficient in the local language. Or, as the representative of Swissmem explained: The purpose of these integration pre-apprenticeships is not to teach any professional, occupational content. ( : : : ) Content-wise they do not have to learn anything, but they have to work, so that they have all these side-effects, such as language, culture, punctuality and so on. (Representative of Swissmem, Winterthur, January 29, 2019) The quote suggests that professional training organizations of more prestigious and selective occupations ensured that refugees could not access their regular VET track too easily by limiting the occupational content of their preparatory programmes. They effectively enforced their role as gatekeepers.
Summarizing the insights on shared beliefs, we find that VET actors less concerned about the prestige of the training system coordinated more willingly with the migration coalition. The programme was implemented more readily in French-speaking cantons and in academically less selective training tracks.

Rational motives: Material resources
Financial resources played an important role in convincing VET actors to coordinate with the migration coalition. Surprisingly, resources were considered less important by professional training organizations, which can be described as material interest groups, compared to cantonal VET authorities.
The programme received rather generous funding. The federal migration office (SEM) motivated VET actors to join its coalition by granting CHF 13,000 (EUR 13,000) per candidate a yeara barely negligible financial incentive. Funds were disbursed to the canton providing the training. Cantons could also remunerate professional training organizations . The CHF 13,000 likely exceeded the costs of training provided in the programme. Similar programmes cost around CHF 10,000 (EUR 10,000) for 10 refugees per semester (representative of cantonal VET office St.Gallen, April 9, 2019). In addition, the requirement that every canton had to contribute funds matching the sum of CHF 13,000 paid by the federal migration office (SEM) for every candidate per year was not rigorously enforced.
The funding structure meant that the programme was somewhat overfunded, leading cantons to participate for financial reasons. This was confirmed by an interview with the SFIVET, which had advised the federal migration office (SEM) on programme implementation (SERI representative, Berne, February 6, 2019; SFIVET representative, Zollikofen, April 4, 2019): It is a strong incentive if you can collect CHF 13.000 per candidate and year for four years for training purposes, given that these candidates needed to be schooled anyway. The costs play a role for the cantons. They were happy to participate. (SFIVET representative, Zollikofen, April 4, 2019) The canton of St.Gallen confirmed that the SEM resources incentivised participation (representative of cantonal VET office St.Gallen, St.Gallen, April 9, 2019).
Cantonal offices not implementing the programme faced public criticism for missing this opportunity and for conceding federal resources to other cantons (representative of Basel-City Welfare Office, Basle-City, February 20, 2019). This was particularly true for Basle-City, an economically thriving canton that had generated a surplus in previous years and did not necessarily depend on federal funding to implement an additional programme. However, cantonal authorities felt that it would be politically unwise not to benefit from federal resources and therefore decided to implement the programme.
Our case also shows the limitations of a coalition based on rational motives. Material resources being excludable goods, coalition actors will compete for the same resources rather than coordinate to generate them, thus complicating longterm coordination. In the view of some professional training organizations, the resource orientation of some cantons limited the degree of coordination (representative of Swiss Association of Meat Specialists, SFF, Zurich, August 2, 2018). Given that programme implementation fell under cantonal responsibility, the professional training organization of meat specialists felt that it had to shoulder too much work: I sometimes had the impression that they [the cantons] just wanted to collect the money from the federal office of migration and were not interested in the rest. Our assessment shows that not all cantons are engaged to the same extent. Some seem to be involved on a superficial level, but when it gets to implementation, they are very reluctant. (Representative of Swiss Association of Meat Specialists, SFF, Zurich, August 2, 2018) Cantons also forwent coordination opportunities because they were keen to organise training themselves in order to be eligible for federal resources. Hence, they were unwilling to allow "their" refugees to pursue training in another cantoneven if allocating candidates to extracantonal classes was desirable, in order to fill classes and minimise overall costs. This proved to be particularly difficult for login, the professional training organization for rail construction. This organisation partnered the Canton of Zurich in delivering the refugee training programme (including classroom instruction). While the CHF 13,000 per training place were awarded to the Canton of Zurich, training places were spread across the entire railway network. Sending refugees to work in other cantons was not feasible, but other cantons could have sent "their" refugees to attend classes in Zurich. However, Zurich charged CHF 20,000 per trainee, which disincentivised most cantons, except Basel-City, from sending refugees there.
Two conclusions can be drawn about the role of rational incentives in coalition formation. First, resources helped convince VET actors to coordinate. Second, and contrary to expectation, resources were less important for material interest groups (i.e. professional training organizations) than for government actors (cantonal VET authorities). Further, crosscoalitional coordination based on rational motives rather than on beliefs may be limited. Rational incentives referring to excludable goods, coalition actors compete for the same resources rather than coordinate to generate them, thus complicating long-term coordination.

Relational motives: Power and reputation
A first type of relational motivesmaintaining formal decision-making power in the VET domaininfluenced if VET actors coordinated with the migration coalition despite opposing beliefs. Actors with formal decision-making power were unlikely to benefit from coordination and hence proved hard to motivate. Other VET actors lacking formal decision-making power, yet who could expand their authority by joining the migration coalition, were more likely to do so. Thus, in line with Proposition 4, powerful actors proved difficult to convince to coordinate with the migration coalition because they were unlikely to increase their power.
The most powerful VET actor is the federal VET office (Secretariat of Education, Research and Innovation, SERI). As VET falls under a federal jurisdiction in Switzerland, the federal VET office has formal decision-making power over the VET system. Unsurprisingly, it was very reluctant to coordinate with the migration coalition. Besides actively opposing a nationally valid certificate, the federal VET authority also refrained from mediating between the cantonal VET offices and the federal migration office (SEM) (SERI representative, Berne, February 6, 2019). The federal migration office (SEM) had not been in direct contact with the cantonal VET offices. Despite the federal VET office having regular contact with the cantonal VET offices and the professional training organizations, it consciously refrained from mediating between the federal migration authority (SEM) and the other VET actors. Interaction with VET actors was seen to provide the federal migration authority (SEM) with a "valuable learning experience" (SERI representative, Berne, February 6, 2019). In the context of programme implementation, which went ahead on a tight schedule, such reluctance on the part of the federal VET office might have easily jeopardised programme success. In the end, the federal migration authority (SEM) sought advice from the SFIVET (SFIVET representative, Zollikofen, April 4, 2019).
Professional training organizations also have formal decision-making powernot over the whole system but over single tracks. Due to the federal VET office's intervention, the programme was not a formal VET programme. Hence, other actors than the professional training orgnaizations could implement it in their professional fields. Other actors encroaching on their field motivated two professional training organizations to implement rather than oppose the programme. Initially, construction technology (Suissetec) and mechanics and automation (Swissmem) did not plan to implement the programme but decided to do so when another actor suggested developing a curriculum in their occupational field (Swissmem representative, Winterthur, January 29, 2019). Having another actor designing preparatory curricula would have meant losing the capacity to select VET candidates, and hence control over diploma standards and prestige. These professional training organizations therefore implemented the programme chiefly because they feared losing their gatekeeping function (i.e. access control).
Our case thus shows that actors with formal decision-making power are more difficult to convince to coordinate across coalitions. The federal VET office did not coordinate with the migration coalition. In contrast, other VET actors (professional training organizations) lacking absolute authority coordinated with the migration coalition because failing to implement the programme would have meant losing control over their VET domain.
A second type of relational motiveleading VET actors to coordinate with the migration coalitioninvolves reputational gains. While such gains played a role in convincing VET actors, this motive was less important compared to the other two motives. Certain professional training organizations were motivated to implement the programme in order to enhance their reputation: retail trade, printing technology, and rail construction. The interviews with their representatives suggest that either the association or the member companies were concerned about their image in regard to social standards. The participation of the retail trade is particularly insightful. Switzerland's two largest retailers, Migros and Coop, implemented the programme themselves, independently of their professional training organization (SEM representative, Berne, December 11, 2018). Both retailers announced their involvement in their weekly newsletters, thus demonstrating their social commitment to a wider public and presenting themselves as responsible employers.
VISCOM, the professional training organization for printing, is also actively involved in political lobbying as a trade union (VISCOM representative, Berne, August 10, 2018). At the height of the refugee crisis, VISCOM opposed new public procurement rules that would have enabled federal authorities to print documents and reports abroad. It is not unlikely that VISCOM was hoping to create political goodwill also for this issue by implementing the programme (SERI representative, Berne, February 6, 2019).
Login, the professional training organization for rail construction, implemented the programme primarily because the state-owned federal railways (Schweizerische Bundesbahnen, SBB) were instructed to do so directly by the Federal Council (login representative, Olten, July 3, 2018). Accordingly, the professional training organization responsible for occupational training at SBB had direct contact with the federal migration authority (SEM) about programme implementation (SEM representative, Berne, December 11, 2018).

Discussion
Shared beliefs are seen as an important precedent for policy coordination. The policy literature, and in particular, the ACF literature, argues that actors sharing beliefs coordinate in order to translate these beliefs into policies (Jenkins-Smith and Sabatier 1993;Sabatier 1988Sabatier , 1993Weible et al. 2009). Following the ACF, coalition members' beliefs are structured hierarchically in that most fundamental ontological and normative beliefs constrain more specific or operational beliefs (Jenkins-Smith et al. 1991, p. 852). This leaves little room for the lowest-order beliefs, which include alternative motives, such as rational concerns.
A first contribution of our paper to the literature consists in a categorisation of the motivations leading to such crosscoalition coordination. Our categorisation includes rational, relational, and normative motives, based on the literature on incentives in collective action organisations (Clark and Wilson 1961, p. 133;Knoke 1988, p. 315;Puffer and Meindl 1992, p. 425). We then analysed to what extent normative, rational, and relational motivations led VET actors to coordinate with a migration coalition in order to support a VET programme for refugees in Switzerland. We found support for each type of motivation, but rational motivations (access to financial resources) were more important than relational ones (reputation and power).
Normative motivations shaped whether coalition actors coordinated. VET actors coordinated with a migration coalition despite a strong concern for the selectivity of the VET system, a normative belief that diverged from the migration coalition. VET actors less concerned about the selectivity of their occupation (occupations traditionally recruiting less academically inclined candidates) or about VET selectivity in general (French-speaking cantons) were more inclined to coordinate with the migration coalition. This finding is consistent with the literature on crosscoalition coordination, which highlights the relevance of shared beliefs (Calanni et al. 2015;Heinmiller and Pirak 2016;Koebele 2020;Weible and Ingold 2018).
The most important motivation to convince VET actors to coordinate with the migration coalition in our case study was rational motives in the form of financial resources. The programme was overfunded and VET actors joined the migration coalition for financial reasons. This finding partly aligns with the literature on crosscoalition coordination. Some studies have found support for the importance of financial resources (Kim and Roh 2008, p. 680), while others have not (Calanni et al. 2015;Weible et al. 2018, p. 14). The result corresponds to the literature on resource dependency arguing that the pursuit of resources motivates coordination (Borgatti and Foster 2003, p. 997;Gulati and Gargiulo 2000, p. 1;Mizruchi 1993, p. 47;Oliver and Ebers 1998, p. 565;Pfeffer 1987).
Relational motives also played a role in convincing VET actors to coordinate with the migration coalition, but a minor one compared to rational motivations. A first type of relational motive consisted in maintaining formal decision-making power in the VET domain. Actors with formal decision-making power unlikely to benefit from coordinating with the migration coalition were hard to motivate (federal VET office). Powerful actors coordinated when they were threatened with losing some of their formal decision-making power by refraining from coordinating (professional training organizations). This finding complements the literature on crosscoalition coordination, which suggests that powerful actors have preferred coordination partners (Calanni et al. 2015;Heinmiller and Pirak 2016;Weible and Ingold 2018).
Likewise, reputational gains played a role in motivating VET actors to support the programme. To our best knowledge, reputation has thus far not been extensively analysed in the literature on crosscoalition coordination. However, it is theorised as an important motive to relate to others, since a positive reputation is more credibly asserted by third parties (Aerne 2020;Lazega and Pattison 1999;Raub and Weesie 1990). Our interviews with representatives of professional training organizations suggest that some of these organisations or some of their members were motivated by concerns about their image in regard to social standards. However, this motive was less important than anticipated.
The role of reputation in stimulating crosscoalition coordination might depend on the salience of the policy issue. The refugee crisis in 2015 attracted great media attention so that resolving this issue became pressing. Opposing a solution, or refraining from coordinating in such an instance, may prove more difficult than in policy contexts attracting less attention. Thus, our finding that reputation plays a limited role may be even clearer in cases with lower salience.
Our second contribution to the literature is that we analysed which actor types reacts to which type of motivation. Surprisingly, we found that resources were more important to government actors (cantonal VET offices) than to material interest groups (professional training organizations). This is in line with studies on crosscoalition coordination (Cappelletti et al. 2014;Weible et al. 2018, p. 14), but contrasts with a typology developed by the ACF literature that argues that financial resources are most relevant to material interest groups (Jenkins-Smith and Sabatier 1994, p. 196;Jenkins-Smith et al. 1991, p. 852;Sabatier 1998, p. 116). We also observed that actors most likely following rational motivations were the only ones lacking formal decision-making power (cantonal VET authorities). In contrast, VET actors with formal decision-making power (federal VET authority, professional training organizations) were more likely convinced by a threat to lose power or reputational gains. Thus, formal decision-making power rather than actor type may determine which motives actors follow.
The importance of financial resources in crosscoalition coordination likely depends on the number of actors without formal authority involved in a policy domain. The number of actors without formal decision-making authority is likely to be higher in federalist polities where government responsibilities are shared across different levels. In such federalist systems, financial incentives might thus more likely bring about crosscoalition coordination (Cappelletti et al. 2014).
Third, our study has generated some unanticipated insights. We found crosscoalition coordination based on rational motives, such as material resources, to be limited. Insofar as rational motives mean pursuing private goods, actors participating in a coalition for rational reasons simultaneously deny other actors the same motive. There is competition forrather than coordination ofprivate goods such as resources. Thus, while rational motives (e.g. resource access) may promote shortterm coordination, even resulting in policy change, beliefs may need to be aligned to ensure sustained coordination. While coalitions based on rational motives may be short-lived, they may bring policy actors into contact with new ideas and policy solutions. Subsequently, actors coordinating for other reasons than shared beliefs may engage in crosscoalition learning and adapt their beliefs. Further research would be needed to better understand the role and sequence of various motives in promoting coordination.

Conclusion
Shared normative beliefs are seen as an important precedent for policy coordination in the literature. Actors sharing beliefs coordinate their activities in order to translate their beliefs into policies. Perhaps most prominently, the ACF explains how actors sharing normative beliefs coordinate in order to render these beliefs into policies (Jenkins-Smith and Sabatier 1993;Sabatier 1988Sabatier , 1993Weible et al. 2009). However, actors may coordinate for other reasons than shared beliefs Calanni et al. 2015;Heinmiller and Pirak 2016;Ingold et al. 2017;Koebele 2020;Henry 2011;Henry et al. 2010;Ingold 2011;Sandström 2011, 2013;Weible 2005).
We contribute to the literature by categorising the various motives that facilitate coordination across coalitions into normative, rational, and relational motives. We have illustrated the relevance of these alternative motives with a case of crosscoalition coordination, where VET actors coordinated with a migration coalition despite opposing beliefs. In addition to shared beliefs, our case shows the importance of rational motives (e.g. financial resources) and relational motives (power and reputation) in convincing actors to coordinate. In our analysis, rational motives were most important to convince actors to engage in crosscoalition coordination. Relational motives were also found to play a role, but a minor one compared to rational motives.
We also analysed who reacted to what kind of incentives. We found that formal decision-making power of actors, rather than actor type to be decisive regarding the motivation actors follow. Subfederal government actors without decision-making power reacted to financial resources much more strongly than material interest groups and federal government actors, both of which have formal decision-making power. Actors with formal decision-making power were convinced by a threat of losing power or reputation gains. This contrasts with a typology developed by the ACF literature that argues that financial resources are most relevant to material interest groups (Jenkins-Smith and Sabatier 1994, p. 196;Jenkins-Smith et al. 1991, p. 852;Sabatier 1998, p. 116).
Coordination based on rational motives seems to be rather short-lived. In our case study, we found that coordination in coalitions based on rational motives is limited because actors compete for the same resources rather than coordinate. Establishing a stable, long-term coalition may thus depend on shared beliefs. However, rational motives may still play a role in catalysing coordination. Actors pursuing rational motives might encounter new policy solutions, which may be followed by policy learning and a possible change in belief in the long run. Further research would be needed to better understand how rational and relational motives, as well as normative beliefs, interact in policy coordination.
As the motives discussed are quite universal, our findings might be generalisable to other cases. Financial incentives, losing control over one's domain, and reputation might also be relevant in other policy areas. Similarly, distance from opposing core beliefs is also likely to play a role in other policy domains. However, the importance of these various motives may be context-dependent. The policy analysed here concerned a highly salient issue (refugee integration). Refraining from coordinating to resolve a pressing issue may involve considerable reputation losses. Reputation may thus be more relevant in crosscoalition coordination if the policy issue is highly salient. In addition, Switzerland is a federalist system. As discussed, financial incentives may be particularly important to convince actors lacking formal decision-making power (e.g. lowerlevel public actors) to engage in coordination. Thus, financial incentives may be more important in federalist systems.
Supplementary material. To view supplementary material for this article, please visit https://doi.org/10. 1017/S0143814X22000290 Data availability. This study does not employ statistical methods, and no replication materials are available.