Forums and coordination across sectors

This article focuses on forums, understood as a particular type of collaborative governance arrangements (Fischer and Leifeld Reference Fischer and Leifeld2015).Footnote ² Forums can facilitate exchange across sectors, that is the public administration sector (including administration and politicians), the interest group sector (including trade associations, professional associations, civil society groups, as well as individual firms and private persons), and the research sector (universities, research centres, research societies). Indeed, forums tend to be established to support cross-sectoral coordination (Bryson et al. Reference Bryson, Crosby and Middleton Stone2006; Maag and Fischer Reference Maag and Fischer2018). While the definition of forums in the literature is not always limited to those including actors from different sectors (Lubell et al. Reference Lubell, Henry and McCoy2010; Mewhirter et al. Reference Mewhirter, McLaughlin and Fischer2019b; Angst et al. Reference Angst, Mewhirter, McLaughlin and Fischer2021), cross-sectoral forums are especially important for creating interaction and exchange between public actors from public administration, interest groups, and research organisations. Forums thus potentially contribute to more holistic and integrative governance in polycentric governance systems (Lubell Reference Lubell2013; Berardo and Lubell Reference Berardo and Lubell2019).

Some forums are part of the official, state-initiated governance mechanisms. This holds for example for advisory committees (Vasseur et al. Reference Vasseur, Lafrance, Ansseau, Renaud, Morin and Audet1997, Krick Reference Krick2015). However, many forums are better understood as self-organising communities of interested actors from different sectors, focusing on knowledge exchange, debate of values and interest, development of common ground, capacity building, or raising public awareness for the issue at stake. Forums are more than informal networks in the sense that they have some sort of formal organisational structures (membership rules, statutes, often an administrative office, etc.) (Fischer and Leifeld Reference Fischer and Leifeld2015; Angst et al. Reference Angst, Mewhirter, McLaughlin and Fischer2021). Forums can produce different types of outputs (Emerson and Nabatchi Reference Emerson and Nabatchi2015b; Fischer and Leifeld Reference Fischer and Leifeld2015; Fischer and Schläpfer Reference Fischer and Schläpfer2017), providing different types of benefits to actors. However, although forums might have formally announced goals and outputs, individual forum participants might have different goals and value different outputs (Emerson and Nabatchi Reference Emerson and Nabatchi2015b; Fischer and Leifeld Reference Fischer and Leifeld2015). Among other factors, the contributions of forum members, or their inputs into the forums, are important for understanding the capacity of forums to produce different types of outputs.

Actors’ inputs to forums

Forum members can contribute to forums in several ways, which all have related costs (Angst et al. Reference Angst, Mewhirter, McLaughlin and Fischer2021). First, forum members provide financial support and in-kind contributions through member fees, service agreements, or through overhead and infrastructure contributions. Second, members provide working time by participating in meetings, contributing to working groups, or supporting the forum administration. Third, members contribute nonmaterial resources such as knowledge or leadership or – more generally – “engagement” (Berardo et al. Reference Berardo, Heikkila and Gerlak2014). A fourth type of cost that forum members might incur is loss of “agency autonomy” (Agranoff Reference Agranoff2006), meaning that forum participation can constrain the freedom of actors to pursue their missions. In a similar context, Warner (Reference Warner2006) speaks of “political costs,” that is, the risk of being coopted. From the perspective of forums, these inputs are factors of production that allow them to create outputs. From the perspective of the forum members that provide the inputs, they are costs: the resources spent on forums are not available for other purposes.

This article focuses on work inputs in terms of working time because this is the most important input for many forum members and one that produces the highest opportunity costs. Direct financial costs by contrast are comparatively low for most members (e.g., a relatively low annual member fee), except for a few members that act as main sponsors.Footnote ³ Nonmaterial inputs such as disclosing information or providing leadership during meetings have low opportunity costs. Finally, costs in the form of loss of agency autonomy are less significant for the forums considered in this article, as most of them do not take binding decisions.

Benefits from forum outputs

The literature on collaborative governance has thus far paid relatively little attention to the question why actors are willing to bear the costs of forum inputs, that is, what benefits motivate actors to bear such costs (Berardo et al. Reference Berardo, Heikkila and Gerlak2014; Gustafson and Hertting 2016; Blakeley and Evans Reference Blakeley and Evans2009). Instead, the literature has focused on initial conditions favouring the emergence of collaborative governance, the processes that sustain it on the level of collaborative arrangements and forums, and their outcomes and impacts (Bryson et al. Reference Bryson, Crosby and Middleton Stone2006; Ansell and Gash Reference Ansell and Gash2008; Koontz and Thomas Reference Koontz and Thomas2006; Emerson and Nabatchi Reference Emerson and Nabatchi2015a; Spekkink and Boons Reference Spekkink and Boons2016). The same tends to be true for the (earlier) literature on polycentric governance, ecology of games, and forums. Although the existence of a variety of forums was recognised as important, and related polycentric governance systems were analysed as a whole, less focus has been put on actors’ transaction costs when choosing among those forums, and their potentially individual interests – independent of mutual or collective interests – when participating in forums (Lubell Reference Lubell2013; Fischer and Leifeld Reference Fischer and Leifeld2015; Berardo and Lubell Reference Berardo and Lubell2019; Oliver and Berardo Reference Oliver and Berardo2021). Similarly, the respective literature tends not to take into account how different forum characteristics might affect actors’ decision of participation and work input (e.g. Fischer and Maag Reference Fischer and Maag2019). Recognising the importance of distinguishing between different units of analysis when evaluating the performance of collaborative governance regimes, Emerson and Nabatchi (Reference Emerson and Nabatchi2015b) emphasise the importance of evaluating different types of outputs and outcomes for participant actors, collaborative regimes and forums as a whole, and specific target goals.

Forum participation and providing work input into forums are usually voluntary contributions (Ansell and Gash Reference Ansell and Gash2018, 24). Therefore, as a basic premise, it is useful to think in terms of a cost-benefit rationale to explain actors’ work inputs into forums (Fischer and Leifeld Reference Fischer and Leifeld2015). It is plausible to assume that actors’ willingness to contribute work inputs positively relates to the benefits they gain from forum outputs and that actors’ work input is not purely altruistic (Emerson and Nabatchi Reference Emerson and Nabatchi2015b). The causal mechanism behind this positive relationship is likely to be circular: meaningful outputs attract inputs (Ansell and Gash Reference Ansell and Gash2008) which in turn allow forums to produce outputs that are even more attractive. Ansell and Gash (Reference Ansell and Gash2018) refer to this circular causality as “positive feedback.”

Two dimensions of forum benefits

It is therefore fundamental to understand the kinds of benefits that actors gain from forum outputs. Two basic distinctions repeatedly appear in the existing literature. The first one separates benefits from exchange in the form of improved knowledge and networks, and benefits in the form of actual or prospective influence on policy and practice. Benefits from exchange are mostly intangible and hardly quantifiable (Lubell et al. Reference Lubell, Henry and McCoy2010; Crona and Parker Reference Crona and Parker2012; Kowalski and Jenkins Reference Kowalski and Jenkins2015). They help actors to do their job in a more informed way and expand their networks, but they do not change the broader framework conditions under which they operate (policies, regulations, patterns of behaviour), at least not in the short and medium term. This is the major difference between the policy benefits that follow from more “advanced” forum outputs such as group decisions (compromises, agreed action plans), or joint action (sensitisation campaignsFootnote ⁴ , lobbying) (Fischer and Schläpfer Reference Fischer and Schläpfer2017). These forum outputs potentially allow members to adapt the framework conditions of governance and policymaking on the issue the forum is dealing with and reach commonly agreed goals, for example, through policy change (Emerson and Nabatchi Reference Emerson and Nabatchi2015b).

Different flavours of this distinction appear in the existing literature. Glasbergen (Reference Glasbergen2011) for example ranks forum outputs on a “ladder of partnership activity” where trust building is at the bottom and changing the political order is at the top. Ansell and Gash (Reference Ansell and Gash2008, Reference Ansell and Gash2018) argue that incentives for participating and investing time into forum work increase as members see “concrete, tangible, effectual policy outcomes” but decrease if outputs are “merely advisory or largely ceremonial.” Warner (Reference Warner2006), in his analysis of multistakeholder platforms for integrated catchment management, finds that knowledge exchange is enough for some stakeholders but that other want “food on the table.” Bauer and Steurer (Reference Bauer and Steurer2014) as well as Lubell (Reference Lubell2004) focus on the question of whether regional partnerships for climate adaptation are catalysts of change, that is do they produce tangible results, or are they “pure talking shops.” Fischer and Schläpfer (Reference Fischer and Schläpfer2017) distinguish joint position papers as more tangible forum outputs from less tangible outputs such as networking or learning in forums in Swiss environmental governance. By contrast, Emerson and Nabatchi (Reference Emerson and Nabatchi2015b), while mentioning “intermediate” outputs, focus on three different types of policy benefits that they call outputs, outcomes, and adaptation.

The second distinction is between individual and collective benefits. Individual benefits serve the self-interest of individual forum members, while collective benefits serve the common interests of most or all forum members and the groups they represent. Examples for the first are actors’ influence on forum decisions or their visibility, examples for the second are the elaboration of compromise solutions or implementation plans. The distinction between individual and collective benefits is important because it might affect the capacity of forums to produce certain kinds of outputs. If benefits from work input to forums are mostly individual, collective action – and forums being more than pure “talking shops” (Lubell Reference Lubell2004; Bauer and Steurer Reference Bauer and Steurer2014) – is only achievable if the self-interests are more or less compatible. By contrast, if collective benefits strongly drive actors’ work input, chances are better that a forum produces collective action. Findings by Blakeley and Evans (Reference Blakeley and Evans2009) suggest that actors that remained committed over a longer period are strongly motivated by collective benefits, while actors mostly attracted by individual benefits were more likely to drop out. Thus, actors’ work input motivated by producing collective benefits is important for the continuity of forums.

The distinction between individual and collective benefits is widely recognised in the literature. For example, it appears in the well-known conceptual frameworks developed by Emerson et al. (Reference Emerson, Nabatchi and Balogh2012), Emerson and Nabatchi (Reference Emerson and Nabatchi2015a; Reference Emerson and Nabatchi2015b) and Thomson and Perry (Reference Thomson and Perry2006). Other authors do not make the distinction explicitly but instead focus on only one type of benefits. For example, in their study on collaborative environmental management in the South Florida Ecosystem Restoration Task Force, Berardo et al. (Reference Berardo, Heikkila and Gerlak2014) only consider individual benefits in a strict rational-choice sense. Warner (Reference Warner2006) by contrast focuses almost exclusively on collective benefits. However, there are only a few empirical insights on the distinction. Most notably, Gustafson and Hertting (2016) find that both types of benefits have about equal weight in the kind of collaborative governance institutions they studied. Blakeley and Evans’ (Reference Blakeley and Evans2009) draw a similar conclusion in their study of citizen participation in a neighbourhood regeneration programme in east Manchester. By contrast, Simmons and Birchall (Reference Simmons and Birchall2005) find that collective benefits outweigh individual benefits. Emerson and Nabatchi (Reference Emerson and Nabatchi2015a) argue at the theoretical level that the work towards collective benefits might replace the work towards individual benefits as a collaborative governance arrangement matures.Footnote ⁵ The distinction between individual and collective benefits is most meaningful for policy benefits (i.e. individual or collective influence on policy or practice), but applies less clearly to exchange benefits. Exchange benefits, such as improved knowledge and networks, cannot be clearly attributed to the individual or collective levels.

Other factors influencing actors’ dedication and work in forums

Existing literature emphasises various other factors influencing how much work actors contribute to forums. First, on the forum level, some forums – for example expert advisory committees convened by government agencies (Krick Reference Krick2015; Vasseur and Lafrance Reference Vasseur, Lafrance, Ansseau, Renaud, Morin and Audet1997) – have an official government mandate or provide official support for a government agency (Imperial Reference Imperial2005; Warner Reference Warner2006; Emerson and Nabatchi Reference Emerson and Nabatchi2015a). Such forums provide additional benefits to members, for example reputational gains, direct influence on policy, or monetary compensation (Emerson and Nabatchi Reference Emerson and Nabatchi2015a, 167). Second, on the level of forum members, members’ contribution depends on their capacity of influencing governance through alternative channels outside the forum (Ansell and Gash Reference Ansell and Gash2008; Scott and Thomas Reference Scott and Thomas2017). Actors with few alternatives might derive particular benefits because a specific forum is their only option of influencing governance. Third, government actors might have a special interest in entertaining forums. Given that governments have a strong interest in enhancing coordination among actors, they might be ready to provide more work input than others into forums (Imperial Reference Imperial2005; Ansell and Gash Reference Ansell and Gash2008, Reference Ansell and Gash2018). Furthermore, other actors with a leadership function within the forum have additional incentives such as the search for influence and prestige. They might also simply have the job to invest more input, as part of the employment at the member organisation. Fourth, the actual reputation of forum members for being important for forum functioning and decisionmaking is crucial. Actors with more reputation might provide more input (Scott and Thomas Reference Scott and Thomas2017; Mewhirter et al., Reference Mewhirter, Coleman and Berardo2019a; Angst et al. Reference Angst, Mewhirter, McLaughlin and Fischer2021). Actors can use their reputation to shape the forum outputs according to their preferences, thus increasing the benefit of forum work. Vice versa, actors with more work input might increase their reputation. Finally, Ansell and Gash (Reference Ansell and Gash2008, Reference Ansell and Gash2018) and Warner (Reference Warner2006) underline the importance of leadership for motivating actors that might not be self-motivated enough to provide inputs. Actors that perceive a given forum to be well lead and managed are likely to have a higher work input into it.

Selection of forums and member survey

The eight forums selected for this analysis are drawn from an encompassing list of forums active in Swiss environmental governance. We apply a purposive selection of forums. Forums end up on our encompassing list if at least one representative of at least two of the following societal sectors are forum members: the public administration sector, the interest group sector (public and private interest groups as well as individual firms), and the research sector. Our encompassing list is the result of several iterations of document and website analysis and snowballing among forum managers. Forums with exclusive goals of either fostering private interests or scientific knowledge were not included. From this list, to have a comparable set of forums in terms of the issues dealt with, we then selected those forums that deal with habitat and natural hazard governance. Table A1 in the supplementary material provides short descriptions of the eight forums as well as the membership composition of the forums.

The population of this analysis consists of all official members of these eight forums. The official forum members were identified based on lists accessible on the web pages of the forums. Thus, our units of analysis are individual persons participating in the forums, but we also know which collective actors they represent (except for a minority of individuals that do not represent any collective actor when participating in the forums). The number of forum members ranges from 7 to 30, with 159 actors in total in the eight forums. A total of 132 actors responded to our online survey during the winter of 2016/17, resulting in a response rate of about 83 per cent (see Table A1 in the supplementary material for response rates per forum). Six of these 132 actors were excluded from the analysis because they are salaried employees of the forums. The analysis thus relies on 126 cases.

Dependent variable

The dependent variable – actors’ work input into forums – is operationalised as the number of days per year that forum members spend on working for the forum and forum-related activities. This includes days of actual physical presence in forum meetings, but also time for preparation work, etc. The exact survey question used to assess the work input of actors into forums was “How many working days per year (full time) do you (and collaborators from your organisation) invest for forum XY (preparation, participation, etc.)?” Answer categories were a) one working day or less, b) two or three working days, c) about one week, d) about two weeks, e) about one month, f) more than one month. For the regression models below, we took the upper limits of these categories (1, 3, 5, 10, 20, 30 days) to create a quasi-continuous variable. For the regression analyses, we log transform the dependent variable in order to avoid a disproportional impact of the highest category (30 days).

Independent variables

Based on the literature relevant to the two dimensions of forum outputs, we operationalise the four categories through several output items. We presented the respondents with a list of thirteen possible forum output items representing the four theoretical categories. We then asked the respondents to indicate whether each output was important (variable value 1), whether the output was unimportant (variable value 0), or whether the actor considered this output not to be an output of the forum (variable value NA). Figure 1 describes the 13 outputs and shows how they operationalise either exchange or policy benefits, and individual or collective benefits. We only classify variables belonging to policy benefits as either individual or collective benefits, because the distinction between individual and collective benefits is not clearly applicable to exchange outputs.

Figure 1. Thirteen outputs and their classification.

Empirically, the 13 output items do not exactly cluster into our theoretical categories, as appears from the results of a multiple correspondence analysis (MCA, see supplementary material, A4). Indeed, we would not expect the items in one category to correlate, but rather to be alternative realisations of the broader theoretical category. Still, we perform robustness tests that consider the most important insights from the MCA and modify categories accordingly.

We aggregate the 13 forum benefits by simply adding the individual variables (1 for “important” output, 0 for “unimportant output”) composing the four different outputs (exchange, policy, individual, collective). As the four types of outputs are composed of different numbers of variables (e.g. individual benefits are represented by only two variables, whereas exchange benefits are represented by 7 variables), we standardise the output variables to correct for different numbers of composing variables.

Control variables

Eight control variables capture the five types of potential alternative explanations described above. First, whether forums have an official mandate by a government agency (government mandate) is assessed by studying the forum’s self-description in documents and on their websites. For example, the Platform Natural Hazards is an extra-parliamentary committee founded by the Swiss government. This variable is a dummy variable with the value 1 (government mandate) and 0 (no government mandate). Second, the likelihood of actors having alternative channels outside of the forum for influencing governance in the respective issue sector is assessed by three variables. A first dummy variable (organisational actors) separates individual forum members without a specific organisational affiliation and representatives of individual firms (0) from forum members that represent an organisation (1). This information stems from the official lists of forum members, as well as a survey question asking respondents to indicate whether they participate in the forum representing an organisation or not. Organisations, as compared to individuals, have more resources, both in terms of time and finances that allow them being active on more alternative channels. A second indicator variable for alternative channels is given by the number of employees that an organisational actor has, as assessed through a survey question on how many employees the organisation has.Footnote ⁶ This information allows for taking into account differences between organisations in terms of resources and general capacity. More (human) resources allow an organisation to be potentially present in various forums (and beyond). A third indicator variable assessing the likelihood of actors having alternative channels outside of the forum is based on the survey question “Besides [forum], there exist many other channels through which your organisation can get involved in discussions on the [forum issue] (e.g. direct contact to public administration actors or international organisations, public campaigns, or participation in other forums). As compared to these alternative channels, how important is [forum] for your organisation to present its positions?” with answer categories “(rather) unimportant,” and “(rather) important.” The variable is thus a dummy variable (other channels), assessing whether the forum is (rather) unimportant for the actor, and it thus has other channels (value 1), or if the forum is rather important in comparison to other channels, indicating that the actor has few other channels (value 0). Third, government actors are distinguished from other actors by a simple dummy variable, indicating whether a forum member is a government agency (1) or not (0). With another dummy variable, we assessed whether actors are in a leadership position based on their formal role as forum managers or board members, or similar (1), as opposed to normal membership (0). Fourth, the reputation of actors is assessed based on reputational power (Fischer and Sciarini Reference Fischer and Sciarini2015). Survey respondents were asked to indicate whether the other forum members were influential or not.Footnote ⁷ For each forum member, reputational power is then calculated as the percentage of other forum members considering the given forum member as influential, expressed as a continuous variable between 0 and 1. Finally, the perception of leadership by forum members relies on a survey question asking forum members whether forum leadership was rather too strict (dummy variable value of 0), rather too little strict (also 0), or just right (1) (satisfaction leadership).

Inputs to forums

On average, the 126 members work about five days per year for their respective forums.Footnote ⁸ However, there is high variation between forums and within the individual forums. Figure 2 plots this variation. It shows for each forum separately the number of working days that its members spend on forum-related activities. The areas of the circles correspond to the number of respondents in a given category (the numbers next to the circles show this number). Figure 2 shows that there is a “lower bound” (minimum days of work) for each forum that varies between one and five days. Interestingly, the forums with an official government mandate (EFBS and PLANAT) or that are government-initiated (Forum Early Detection) have the greatest lower bounds. Only one member works less than five days for these three forums. By contrast, in three other forums, a significant share of members works no more than one day.

Figure 2. Number of working days that members spend on forum-related activities.

Beyond these cross-forum differences, there is considerable variance within most of the forums. A small number of forum members work significantly more than their colleagues do. This finding holds in particular for the three members that work 20 days, but also for the three members of the AG W&W and the AG Renat that work 10 days. Set aside these six actors, the within-forum range in the number of working days committed to forums is about 5 to 8 days. Within the Forum Early Detection, there is no variance at all.Footnote ⁹

Benefits from work in forums

Our first, descriptive question is which kinds of benefits do forum members derive from the forum they participate in. Figure 3 shows for each of the thirteen output categories the share of forum members that consider the respective output important (darkest part of the bars).

Figure 3. Importance of outputs.

Figure 3 shows that exchange outputs are considered important by almost all forum members (>90%), the only exception being studies which are considered important by only about half of the forum members. Apparently, formalised knowledge (studies) is valued less than informal knowledge exchange. The appreciation for all outputs referring to policy benefits is markedly lower than for exchange outputs. This means that informal exchange is enough for a significant part of forum members. Actors work in forums to gain new insights and connect to other actors in the field, but their interest in influencing the broader framework conditions is limited.

Among the six outputs referring to policy benefits (bottom half of Figure 3), the patterns are less clear. Within this category, we distinguish individual and collective benefits. While the individual benefits and two of the collective benefits (position papers and sensitisation) are considered important by roughly two-thirds of the forum members, compromises and implementation plans are appreciated by less than half of the members. Figure 3 also provides evidence that some actors consider the more tangible outputs unimportant. This is possible since the costs of forum participation are relatively low for most actors; in many cases, it amounts to just one or two days per year and no financial contributions.

The patterns in Figure 3 are almost identical if disaggregated for the three actor types (government, interest groups, science). Only a few differences are worth noting (see A2 in the supplementary material): First, government actors seem to consider the more tangible outputs (compromises, implementation plans, consulting) slightly more important than the other two actor types. This is consistent with their role in the political system where they are responsible for developing pragmatic, broadly acceptable solutions (Crona and Parker Reference Crona and Parker2012; Ingold and Varone Reference Ingold and Varone2012). Second, actors from the science sector consider consulting and influencing markedly less important than the other actor types.

Contrary to the stability of patterns across types of actors, the patterns of which output is considered important vary substantially between forums, reflecting the different activities of the forums (see Figure A3 in the supplementary material). For example, the variation across forums is larger for policy outputs than for exchange outputs. Furthermore, one forum (fifth bar) consistently scores lower on any of the outputs.

Benefits of work input into forums

We now address the second questions posed at the beginning of the article: Which kinds of benefits motivate certain actors to dedicate working time to forums? When specifying the regression models, we need to take into account two methodological aspects: First, given the scale on which working days are measured (1, 3, 5, 10, 20 days), we cannot assume the proportional odds assumption to hold for our dependent variable. Therefore, we log transform the dependent variable. This also corresponds to the argument that the perceived cost of an additional working day is smaller for someone that is already working several days compared to someone that is only working one day. The second crucial aspect when specifying the regression models is the fact that the average level of work input differs widely between the eight forums. Therefore, we apply a random intercept model, being aware that standard deviations might not be reliable given the low number of forums. The results appear in Table 1. Two alternative models (basic model without random intercepts (Table A5.1), ordinal logit regression without transforming the dependent variable (Table A5.2)) appear in the supplementary material.

Table 1. Results of regression models (linear mixed effects, logged dep var)

Values in brackets represent standard deviations, coefficient estimates with p-values lower than .05 appear in bold. Given the log-transformation of the dependent variable, the size of effects of these coefficients should be interpreted based on their exponentiation. The first three models (1a–1c) examine the effects of exchange and policy benefits, both separately and combined. Models 2a–2c do the same with individual and collective benefits. Notice that exchange benefits are excluded entirely in models 2a–2c, given that they could be individual or collective. The final model (3) is a refinement of Model 1c, with the policy benefits split into individual and collective benefits.Footnote ¹⁰

Results reported in Table 1 show quite clearly that outputs that correspond individual benefits rather than collective benefits are correlated with higher work input. Whenever this distinction is made (models 2a, 2b, 2c, 3), individual benefits appear as significant, whereas collective benefits do not. By contrast, there is no clear difference between exchange and policy benefits. Both are related to work input when entered into the model as individual categories (models 1a, 1b), but the combined model 1c does not pick up any significant difference between both categories. Still, when the category of individual benefits is separated into individual and collective benefits – that is if this category is separated into meaningful subcategories of individual and collective benefits – exchange benefits seem to be related to work input. Overall, results clearly show that individual benefits are related to work input, that is actors that value individual benefits provide more work input into forums. Results with respect to the distinction between exchange and policy benefits are less straightforward. Although the alternative model specification in Table A5.1 (models without random intercepts) confirms these results, models in Table A5.2 tend to suggest that rather policy benefits than exchange benefits are related to work input into forums. However, given the inconsistent findings across models, we refrain from giving too much weight to this result.

As further robustness tests, we use different aggregations (Table A5.3 in the supplementary material). In the exchange category, we remove “mutual understanding” and “networks” because they are social capital outputs rather than pure knowledge (“networks” also lies less close to the other items in the empirical clustering analysis, as can be seen in Figure A4) (Model R1 in Table A5.3). In another robustness test, we take out “studies” from the exchange category and classify it as “policy” and “collective” output, also based on the clustering, see Figure A4) (Model R2 in Table A5.3). Generally, the results confirm the results from the main model, but R2 provides another suggestion that exchange benefits might be less related to work input than policy benefits.

The effects of control variables are consistent across the different models (see models in the supplementary material for some exceptions). First, there is more work input in forums with a formal government mandate. Besides the fact that a formal mandate by the government might spur more work input because of the more direct channel of forum work into government decisions, it is likely that these forums provide additional benefits to their members, such as reputation, or financial compensation. This finding somehow contrasts the argument in the literature suggesting that grassroots/ voluntary collaborations are more effective overall at attracting and retaining members (e.g. Ulibarri et al. Reference Ulibarri, Emerson, Imperial, Jager, Newig and Weber2020). Second, the forum members’ capacity of influencing governance through alternative channels outside the forum is assessed with three variables. The number of employees has a positive effect on work input, meaning that actors with more overall resources are also able to provide more work input into any forum. Actors with other channels tend to provide less input into forum work, suggesting that the existence of alternatives decreases actors’ work input in a given forum. Third, actors in a leadership position within the forum provide more work input, but government actors do not. Finally, forum members’ reputation within the forum clearly relates to work input, but the degree of satisfaction with forum leadership is not.