Navigating Input and Output Legitimacy in Multi-Stakeholder Initiatives: Institutional Stewards at Work

: Multi-stakeholder initiatives (MSIs) are a form of private governance sometimes used to manage the social and environmental impacts of supply chains. We argue that there is a potential tension between input and output legitimacy in MSIs. Input legitimacy requires facilitating representation from a wide range of organizations with heterogeneous interests. This work, however, faces collective action problems that could lead to limited ambitions, lowering output legitimacy. We ﬁnd that, under the right conditions a relatively small group of motivated actors, who we call institutional stewards, may be willing to undertake the cost and labor of building and maintaining the MSI. This can help reconcile the tension between input and output legitimacy in a formal sense, though it also results in inequalities in power. We test this claim using a case study of organizations’ activities in the Roundtable on Sustainable Palm Oil (RSPO). We ﬁnd that a small group of founding members—and other members of long tenure—account for a disproportionate level of activity in the organization.

Often, studies of MSIs revolve around questions of legitimacy [7] (p. 5), focusing on the formal structure of the MSI as a whole, rather than individual members [15]. Though their membership diversity makes it difficult for single actors to control MSIs, [4], these organizations still are not particularly democratic in practice [11,13,16]. International nongovernmental organizations (NGOs) based in the global North, for example, frequently are overrepresented, while smaller organizations from the global South are marginalized [13,17,18].
Other work on MSIs focuses on the organizational strategies and interests driving standards' development [19,20]. Several of these document tensions between MSIs' legitimacy as a whole and their characteristics and the conditions that produce them. Lawrence and Suddaby's [20] institutional work approach argues that the creation of new institutions requires the support of a wide range of actors [62]. A practice-oriented account [63], institutional work focuses on how actors engage with each other by examining the "purposive action of individuals and organizations aimed at creating, maintaining, and disrupting institutions" [20] (p. 215). Through numerous case studies, the institutional work literature documents how individuals' actions affect institutions on a daily basis, thus bridging the literatures on legitimacy and strategic action [26,27]. For our purposes, institutional work is particularly helpful for identifying practices aimed at affecting institutions; the distribution of such activities, in turn, identifies who does what within an MSI.
Whereas institutional work turns attention to mundane actions, it still considers power and the capacity to undertake institutional work [64]. Echoing discussions on MSIs and institutional entrepreneurship, recent studies often find that institutional work is undertaken by elites seeking a central institutional role [65][66][67]. Often, these actors have in-depth knowledge of the broader institutional field and use their social capital and skills to support their work [61,68].
Though the concepts of institutional work and institutional entrepreneurship have clear similarities, the former is more inclusive because it considers not only institutional creation, but also maintenance and disruption. Battilana et al.'s [69] (p. 68) influential definition of institutional entrepreneurship holds that entrepreneurs initiate divergent changes that "break the institutional status quo in a field of activity" (p. 68). This theoretical construct clearly precludes those who participate in institutional maintenance, despite that this, as well, is strategic work. In response, we coin the term institutional stewards as institutional maintenance's analog to the entrepreneur, defining stewards as actors who initiate non-divergent change, extending fundamental institutional logics into novel circumstances. Similar to institutional entrepreneurs [69] (p. 68), institutional stewards actively participate in implementing change (Table 1), but those changes are reforms intended to preserve the institution as a whole. Table 1. Institutional entrepreneurs versus institutional stewards [21,60,69].

Concept Definition
Institutional Entrepreneurs Actors who leverage resources to initiate and implement divergent change Institutional Stewards Actors who leverage resources to initiate and implement non-divergent change While conceptually distinct, there are good reasons to expect a strong empirical overlap between institutional entrepreneurs and institutional stewards [20]. Considering the resources institutional maintenance requires, we would expect only actors with a considerable stake in the game to engage in stewardship [56]. Accordingly, the willingness to participate and serve as stewards may depend on institutional resources, although it is difficult to determine a common metric because MSIs span across sectors. For firms, Dang et al. [70] suggest total assets, total sales, and market value of equity as different measurement proxies for size, but these measures do not apply readily to civil society [48,49]. In any case, while organizations' size may contribute to institutional work, MSIs can also represent an opportunity to change the status quo for smaller actors [13,42].
As noted above, the institutional entrepreneurship literature presumes actors create institutions because they see this to be in their interest. We also observed that the consolidation of like-minded actors in the initial phases of institutional creation might be a means of making the process more manageable. Taken together, these points suggest that the organizational coalition initiating an MSI is likely to be critical to its maintenance. That is, the constraints of institutional entrepreneurship should select a set of actors or organizations that will later be predisposed to act as stewards. In the case of the RSPO, for example, we would then expect stewards to frequently be organizations that have participated since the MSIs' origins. These initial actors, presumably, determined the initial rules of the game, giving them an early adopter advantage which should increase their stake in the MSI [7]. Considering the effort required to build institutions [26,27], we would expect these actors to preserve their influence on the future development of the RSPO through subsequent maintaining work. In this way, the actors who initially served as institutional entrepreneurs are more likely to go on to become institutional stewards.

Operationalizing Institutional Work
As there has been relatively limited attention to the internal dynamics of MSIs, we turn to the literature on institutional work, operationalizing organizational activities using the categories identified by Lawrence and Suddaby [20]. We focus on two broad categories of work discussed above: creating new institutions, and maintaining extant ones. More specifically, we focus on the categories of advocacy, defining, and vesting within creation work, and enabling, policing, and deterring within maintenance work.
Advocacy, defining, and vesting together compose the "overtly political work in which actors reconstruct rules, property rights and boundaries that define access to material resources" [20] (p. 221). Specifically, advocacy serves as a means for initially acquiring the legitimacy needed to create institutions through direct representation of specific actors and their interests [20]. According to Suchman [37], advocacy is of particular importance for marginalized actors, since it provides an opportunity to shape the institutional environment through lobbying for additional resources, advertising, and litigation [71]. Next, defining work mainly centers on the creation of constitutive rules, such as membership requirements or standards, which establish the parameters for developing institutional practices [20,72]. Together, advocacy and defining provide the foundation for vesting, mainly pertaining to the allocation of property rights [20,73].
The maintenance work of enabling, policing, and deterring is primarily aimed at "ensuring adherence to rule systems" [20] (p. 230). Because MSIs primarily operate via explicit principles or rules, we focus on maintenance work undertaken to ensure rules remain efficacious. Enabling fosters certainty and agreement in institutional arrangements by introducing rules that facilitate, supplement, and support institutions [20]. Policing is meant to ensure compliance with the rules of an institution through means of enforcement, auditing, and monitoring that may take a "carrot or stick approach" [74] with either sanctions or inducements [20,73]. Finally, deterring seeks to guarantee compliance with the rules by establishing coercive barriers that limit or prohibit institutional change [20].

Data Sources
Our analysis takes a mixed-methods approach, beginning with qualitative coding and content analysis of RSPO documents to outline governance patterns, and to generate datasets that allow us to quantitatively analyze organizational activity within the MSI. This corpus is composed of material available from current and past versions of the RSPO website stored by the Internet Archive. These include key documents about the Membership Rule, RSPO Code of Conduct, Principles & Criteria, and Statutes [75]. Much of our quantitative analysis relies on data available on the RSPO's membership as of February 2017 [76]. To collect data to analyze membership composition, we utilized the RSPO website's membership search engine to extract the full list of its members. As new organizations frequently join the RSPO, we included the ordinary members from the period 2004 until February 2017 (the cut-off point of data extraction). Due to the absence of archival versions of the RSPO membership list, these data only contain organizations that were members at the time of data collection. As membership departure rates appear to be relatively low, however, we do not believe this significantly affects our conclusions.
We utilized minutes of RSPO General Assembly (GA) [77] and Board of Governors (BoG) [78] meetings from 2004 to 2015 to compile lists of BoG members during that time period, as well as current lists of governing body members to identify membership in those groups. These minutes were also used to identify members that propose resolutions in the GA and to provide context on internal RSPO debates. In addition, we combined attendance summaries posted on several recent RSPO GA websites to construct the history of attendance at a sample of recent GAs. Finally, we drew on membership lists presented on the RSPO website [79] to identify members of standing committees, working groups, and task forces as of 2017.
As a means of comparing the RSPO with peer MSIs, we reconstructed the number of prior service years for each organization on the board of the Forest Stewardship Council (FSC), Better Cotton Initiative (BCI), and Roundtable on Responsible Soy (RTRS) since the year 2000. To collect these data, we used a combination of general body meeting minutes and Internet Archive website caches from the respective organizations, all of which are freely available on their websites.

Statistical Modelling and Data Visualization
To test our hypotheses, we estimated regression models predicting the propensity to propose resolutions at the general assembly and serving on an RSPO governing body. Given that resolution proposal is a rare event, we account for zero inflation by estimating zero-inflated negative binomial (ZINB) regression models. An extension of Poisson regression, ZINB models allow for overdispersion in the response variable [80], while the zero-inflated component in the model addresses distributional problems that can arise from observations of zero proposals. [81,82]. The former part, in other words, models the probability of having strictly zero counts, whereas the latter uses a negative binomial regression to model non-zeroes [82] (p. 720). Both excess zeroes and overdispersion are frequent issues in real-life data [83], and accounting for these reduces estimation biases [82]. Since rare events also raise risks of biased coefficient estimates in standard logistic regressions, we used Firth's [84] penalized likelihood estimation approach, as implemented in logistf [85] in R 3.4.1 [86], to estimate logistic regression models to check the robustness of our negative binomial and ZINB estimations.

Origins of the Roundtable on Sustainable Palm Oil
Global production of palm oil is expected to reach double its 2010 volume and area by 2025 [87]. Palm oil production has historically been geographically concentrated in Indonesia and Malaysia, while consumption is distributed globally (Figures 1 and 2). With growing demand, suitable land in the leading producer countries is becoming scarce [88], and plantings have expanded in Africa and Latin America [89][90][91][92], often promoted by national governments [93]. While evidence on direct livelihoods impacts are limited, some recent studies find that smallholder oil palm production improves both village-level assets [94] and household-level livelihoods and nutrition, though the effects are stronger for more affluent households and are driven in large part by farm expansion [95].  [96]. Visualized using ggplot2 [97], in R 3.4.1 [86]. Growth of global palm oil and palm fruit oil production since 1980 [96]. Visualized using ggplot2 [97], in R 3.4.1 [86].
Although the RSPO's founding members include organizations from both the global North and South, it was a group of Northern organizations that started preparations prior to the first GA in 2004 [108]. In 2001, the World Wildlife Foundation (WWF) began engaging stakeholders in the palm oil supply chain, starting with informal collaboration with Aarhus United UK Ltd., Migros, the Malaysian Palm Oil Association, and Unilever [23]. In the same year, responding to increased attention to palm oil's externalities, WWF-Switzerland hired an external consultant to consider a partnership model with businesses to regulate the industry. Having decided on a roundtable format, the next step was to look for potential partners [109]. The process culminated in a meeting in London in September 2002 attended by 16 different European stakeholder organizations, intended to gather support for the proposed initiative and discuss preliminary goals, structure, and working plans [108]. During the meeting, participants agreed that a program driven purely by  [96]. Visualized using ggplot2 [97], in R 3.4.1 [86].
Although the RSPO's founding members include organizations from both the global North and South, it was a group of Northern organizations that started preparations prior to the first GA in 2004 [108]. In 2001, the World Wildlife Foundation (WWF) began engaging stakeholders in the palm oil supply chain, starting with informal collaboration with Aarhus United UK Ltd., Migros, the Malaysian Palm Oil Association, and Unilever [23]. In the same year, responding to increased attention to palm oil's externalities, WWF-Switzerland hired an external consultant to consider a partnership model with businesses to regulate the industry. Having decided on a roundtable format, the next step was to look for potential partners [109]. The process culminated in a meeting in London in September 2002 attended by 16 different European stakeholder organizations, intended to gather support for the proposed initiative and discuss preliminary goals, structure, and working plans [108]. During the meeting, participants agreed that a program driven purely by European interests would potentially be illegitimate [108] (p. 10) and participation of stakeholders from Asia, in particular, would be necessary to secure critical mass and subsequent acceptance by growers [16,108]. These acknowledgements aside, the absence of organizations from the global South at the preparatory meeting in London meant Northern stakeholders drove preliminary defining work for the RSPO: the meeting established the framework for sustainability assessment [108] (p. 9) and the roundtable format [109] intended to "develop and to agree on criteria and standards for sustainable palm oil" [108] (p. 13). WWF's meeting agenda effectively laid the RSPO's foundation, establishing the organization as the RSPO's original institutional steward.
Upon concluding the initial meeting, WWF and Unilever together created an organizing committee for the first roundtable, planned for 2003 [109]. WWF reached out to the Malaysian Palm Oil Association (MPOA) to ensure representation of producing countries [108]. Eventually, the committee included six organizations-the founding members of the RSPO (two of these organizations, Aarhus United UK and Karlshamns AB, later merged to form AAK, which we subsequently identify as a founding member). Each contributed €10,000 to fund the first roundtable meeting in Kuala Lumpur [110]. Participation in the organizing committee allowed the founding members to have additional influence on defining work [20], considering their responsibility for developing the founding rule system [109]. As noted above, these foundational choices can leave important legacies. The RSPO's organizational structure of seven different sectors, for example, was defined before the first GA and has not changed [111]. Similarly, the organizing committee established the BoG's responsibility for resources and programs [109].

Development of the RSPO
As of February 2017, the RSPO boasted over 3400 members divided into four categories: Ordinary, Affiliate, Supply Chain Associate, and Honorary [112]. Members directly participate in RSPO governance through two primary mechanisms: the annual GA meetings, where members propose and vote for resolutions, which direct the RSPO's development; and the Governing Bodies (GBs), including the BoG who oversees implementation and special projects [113]. Only Ordinary Members may propose and vote for resolutions at the GA, and if a resolution is approved, it is further processed by the BoG, a group of sectoral representatives elected by the Ordinary Members and responsible for day-to-day administration [114]. Given their central role in RSPO governance, our analysis focuses on the Ordinary Members.
The RSPO's Ordinary Membership has increased markedly over time, while the overall composition has changed ( Figure 3). The RSPO began with 36 Ordinary Members in 2004, a number that has climbed to over 1400. Following a slow initial uptake of new members, growth has increased rapidly since 2010. While some sectors, particularly Banks and Investors and NGOs, have had relatively stable numbers, membership ranks have grown especially for firms operating in the middle of the supply chain, particularly consumer goods manufacturers and processors and traders. The sizeable growth in consumer goods manufacturers, in particular, has shifted the composition of Ordinary Members toward organizations registering their membership in Organisation for Economic Co-operation and Development (OECD) countries. These firms now account for about 70% of total member numbers ( Figure 4). It is possible that the requirements of Ordinary Membership drive some of the differential growth observed in Figures 3 and 4. Oil Palm Growers, for example, are required to submit information on land clearing and other plantation activities upon application [75].

Institutional Stewards in the Roundtable on Sustainable Palm Oil
While the RSPO's membership structure allows heterogeneous groups of members to "express their positions and interests in specific and specialized domains of expertise" [7] (p. 11), past studies have questioned the degree to which power is distributed between and within these sectors [16], with particular attention to the dominance of stakeholders from the private sector and the global North [115]. Our concerns here are slightly different. Given our argument that founding coalitions of institutional entrepreneurs are likely to be willing and able to put in the labor required for institutional maintenance, and, hence, become institutional stewards, we assess the degree to which

Institutional Stewards in the Roundtable on Sustainable Palm Oil
While the RSPO's membership structure allows heterogeneous groups of members to "express their positions and interests in specific and specialized domains of expertise" [7] (p. 11), past studies have questioned the degree to which power is distributed between and within these sectors [16], with particular attention to the dominance of stakeholders from the private sector and the global North [115]. Our concerns here are slightly different. Given our argument that founding coalitions of institutional entrepreneurs are likely to be willing and able to put in the labor required for institutional maintenance, and, hence, become institutional stewards, we assess the degree to which founding and longstanding members continue to perform the lion's share of key institutional work in the RSPO. As noted above, there are two primary avenues by which Ordinary Members can participate directly in RSPO governance: participation in the GA and service on the BoG and other GBs. In this section, we analyze participation in institutional work in both these arenas.

Participation in the General Assembly
In a formal sense, the GA, which oversees the secretariat and passes resolutions that affect the future direction of the institution, holds fundamental power in the RSPO. Many MSIs, such as the Forest Stewardship Council (FSC), weight voting power to balance different sectors [34]. The RSPO, by contrast, has an unweighted voting system, granting numerically dominant groups, such as consumer goods manufacturers and palm oil processors and traders, significant voting power. Nevertheless, as these sectors have grown, most of the newer members are not bothering to attend GAs ( Figure 5). Indeed, the proportion of members attending GAs fell for all sectors between 2012 and 2015.
This general trend, however, hides an important geographic feature of GA attendance. GAs are held in Southeast Asia, so growers, processors, and traders tend to have lower transportation costs, a fact reflected in these sectors' share of recent GA attendees ( Figure 6). Specifically, despite that consumer goods manufacturing sector membership numbers have increased significantly over the past several years, their share of total attendees in GAs has declined. Retailers, similarly, have declined as a proportion of actual attendees. While this is offset somewhat by the option of electronic voting, generally only around 25% of Ordinary Members registered at GAs have exercised this option in recent years. held in Southeast Asia, so growers, processors, and traders tend to have lower transportation costs, a fact reflected in these sectors' share of recent GA attendees ( Figure 6). Specifically, despite that consumer goods manufacturing sector membership numbers have increased significantly over the past several years, their share of total attendees in GAs has declined. Retailers, similarly, have declined as a proportion of actual attendees. While this is offset somewhat by the option of electronic voting, generally only around 25% of Ordinary Members registered at GAs have exercised this option in recent years.    . RSPO sectoral attendance at recent GAs, as a percentage of total attendees. Visualized using ggplot2 [97], in R 3.4.1 [86].
Of course, members can influence GAs not only by voting but, also, by proposing resolutions to be voted upon. As resolutions set the RSPO's overall direction, they are an important part of institutional work. As a matter of organizational culture, resolutions brought to the GA floor have generally been heavily vetted and, as a result, tend to be likely to pass, making resolution proposal itself an indicator of influence. Figure 6. RSPO sectoral attendance at recent GAs, as a percentage of total attendees. Visualized using ggplot2 [97], in R 3.4.1 [86].
Of course, members can influence GAs not only by voting but, also, by proposing resolutions to be voted upon. As resolutions set the RSPO's overall direction, they are an important part of institutional work. As a matter of organizational culture, resolutions brought to the GA floor have generally been heavily vetted and, as a result, tend to be likely to pass, making resolution proposal itself an indicator of influence.
If, as we argue, founding members are more likely to act as institutional stewards for the RSPO, we would expect to find that founding members, as well as members with longer tenures, would be more likely to propose resolutions in the GA. To test these claims, Table 2 presents the results of a series of one zero-inflated negative binomial and three logistic regression models predicting the number of resolutions proposed in a given year and the probability that an organization proposes a resolution in a given year, respectively, spanning from the first GA in 2004 until GA 12 in 2015. During this period, only 63 non-procedural resolutions were voted on and, of these, only 55 (87%) had sponsors identified in the minutes. The final three models are aggregated across the observation period and predict whether or not an organization has ever proposed a resolution. * = sig. at 0.05, ** = sig. at 0.01, *** = sig. at 0.001. Several significant patterns can be discerned from these models. First, it is clear that founding and BoG members frequently propose resolutions. Based on the zero-inflated negative binomial model, founding members are more likely to propose resolutions, while BoG members are more likely to propose multiple resolutions in a single GA, a finding consistent with the role of BoG in redrafting resolutions. With the exception of NGOs, which in several models appear to be more likely to propose resolutions than other groups, there are no other systematic organizational predictors of proposing resolutions. Importantly, when previous resolution proposal is included in the logistic regression models, none of the other variables are statistically significant, pointing to the concentration of proposal activity in a very small share of Ordinary Members across the period studied.

Participation in Governing Bodies
Given the high degree of overlap between founding members, long-term members, and BoG members, it is possible the models in Table 3, while largely as we would expect, partially mask the role of long-term and founding members, attributing some of their activity to service on the BoG. In this section, we examine participation in the BoG and other governing bodies (GBs). Based on our understanding of institutional stewards, we would expect to find that founding and long-term members are also more likely to serve on GBs. Unlike the GA, where all members can participate and vote, the BoG has a predetermined number of seats that must contain between three and sixteen Ordinary Members [114]. BoG members are elected by the GA for a period of two years, and Ordinary Members may only vote for a representative within their own sector and sub-sector [108]. Generally speaking, the predetermined seats on the BoG help to ensure a more balanced representation of the different stakeholder groups [13], in comparison to the GA. Even though some sectors have more representatives than others [114], the BoG-as well as the other GBs-aims for "decision[s] taken in absence of any sustained objection" [113] (p. 27).
Participation in the BoG is particularly important in driving the RSPO's development, and BoG members are inherently able to become institutional stewards. For example, while other Ordinary Members may forward resolutions at the GA, any changes to the standard must be endorsed by the BoG in accordance with the Standard Operating Procedure for Standards Setting and Review [113].
Turnover in the BoG has been quite slow, with elections staggered following a decision taken at GA3, such that there was never more than 25% turnover in any of the years analyzed. Many member organizations stay on the BoG for multiple terms; around 25% of GA members have been founding members or members of the original BoG since 2008 (Figure 7). Indeed, all but one of the RSPO's founding members have remained on the board consistently. In the remaining seats, organizations frequently serve consecutive terms as BoG members. GA minutes from 2004-2015 [77] reveal the vast majority of elections were uncontested, and the seat reserved for smallholders remained vacant for the RSPO's first two years before being allocated. This pattern can be seen in Figure 8, which presents the distribution of years of service on the BoG for organizations with at least one year of service. The distribution has a long tail, with peaks at around two and ten years of service, indicative of a small group of very long-serving members. In contrast to patterns of GA attendance, the distribution of sectoral representation in the BoG tilts toward OECD countries, whose share of BoG members has been slowly climbing since the RSPO was founded (Figure 7). Throughout the study period, Northern organizations have occupied approximately two thirds (62.3%) of BoG seats, in comparison to a little over one third (37.7%) for the South. NGOs initially followed a trend of equal distribution between North and South, but since 2009 for Environmental or Nature Conservation Organizations, and 2013 for Social or Development Organizations, Northern members have occupied both seats. The composition of BoG seats held by the remaining sectors has remained fairly consistent with their sectoral composition: the two seats for Palm Oil Processor and/or Traders have been evenly divided between North and South members, whereas consumer goods manufacturers, retailers, and banks and investors have all been from the North. In contrast to patterns of GA attendance, the distribution of sectoral representation in the BoG tilts toward OECD countries, whose share of BoG members has been slowly climbing since the RSPO was founded (Figure 7). Throughout the study period, Northern organizations have occupied approximately two thirds (62.3%) of BoG seats, in comparison to a little over one third (37.7%) for the South. NGOs initially followed a trend of equal distribution between North and South, but since 2009 for Environmental or Nature Conservation Organizations, and 2013 for Social or Development Organizations, Northern members have occupied both seats. The composition of BoG seats held by the remaining sectors has remained fairly consistent with their sectoral composition: the two seats for Palm Oil Processor and/or Traders have been evenly divided between North and South members, whereas consumer goods manufacturers, retailers, and banks and investors have all been from the North.

Other governing bodies
Whereas BoG members are elected by Ordinary Members at the GA, the remaining GBs are appointed [15,40]. Officially, the Standing Committees (SCs) are composed of "members from the BoG, or their respective alternates, and RSPO members and strives for balanced representation of all membership categories" [113] (p. 6). Nevertheless, it is at the sole discretion of the BoG to determine what counts as balanced representation [113]. Subsequently, the SCs are responsible for deciding whether it is necessary to establish a Working Group (WG) or Task Force (TF) to undertake more detailed work [113]. Whereas the annual GA has been hosted in Southeast Asia since the founding of the RSPO, GB meetings take place throughout the world. From the available minutes, the majority of WG and TF meetings appear to take place in an Asia, while the locations of BoG and SC meetings are more evenly divided between Europe and Asia [75]. Figure 9 depicts the membership overlap among the GBs (including the BoG) for which membership rosters were available as of February 2017. Founding members are labelled, while BoG members are displayed in black. The visualization utilizes Fruchterman and Reingold's [117] algorithm to place GBs and organizations, which locates organizations serving on GBs with overlapping membership more closely together. Both GBs and organizations are sized by their network degree, which for organizations is the number of GBs on which they serve and, for the GBs, is the number of organizations with a seat. The visualization suggests a relatively small group of organizations serve on several GBs, forming a core with the potential to connect discussions taking place on different issues. Two of these organizations are founding members, while a third founding member, AAK, serves on two GBs in addition to the BoG (We have excluded the Finance Committee from the visualization, as the membership is not listed on the RSPO site. However, AAK has an ex officio seat on the Committee by virtue of its role as Treasurer on the BoG).

Other Governing Bodies
Whereas BoG members are elected by Ordinary Members at the GA, the remaining GBs are appointed [15,40]. Officially, the Standing Committees (SCs) are composed of "members from the BoG, or their respective alternates, and RSPO members and strives for balanced representation of all membership categories" [113] (p. 6). Nevertheless, it is at the sole discretion of the BoG to determine what counts as balanced representation [113]. Subsequently, the SCs are responsible for deciding whether it is necessary to establish a Working Group (WG) or Task Force (TF) to undertake more detailed work [113]. Whereas the annual GA has been hosted in Southeast Asia since the founding of the RSPO, GB meetings take place throughout the world. From the available minutes, the majority of WG and TF meetings appear to take place in an Asia, while the locations of BoG and SC meetings are more evenly divided between Europe and Asia [75]. Figure 9 depicts the membership overlap among the GBs (including the BoG) for which membership rosters were available as of February 2017. Founding members are labelled, while BoG members are displayed in black. The visualization utilizes Fruchterman and Reingold's [117] algorithm to place GBs and organizations, which locates organizations serving on GBs with overlapping membership more closely together. Both GBs and organizations are sized by their network degree, which for organizations is the number of GBs on which they serve and, for the GBs, is the number of organizations with a seat. The visualization suggests a relatively small group of organizations serve on several GBs, forming a core with the potential to connect discussions taking place on different issues. Two of these organizations are founding members, while a third founding member, AAK, serves on two GBs in addition to the BoG (We have excluded the Finance Committee from the visualization, as the membership is not listed on the RSPO site. However, AAK has an ex officio seat on the Committee by virtue of its role as Treasurer on the BoG). . Visualized using ggplot2 [97], in R 3.4.1 [86]. GBs displayed as circles; GB members displayed as triangles. BoG members and BoG displayed in black.
Given the relatively small number of GBs, organizations' centrality in Figure 9 is driven in large part simply by the number of GBs upon which an organization serves. Again, in keeping with our assumptions about institutional stewards, we would expect founding members and members of long tenure to be more likely to serve on GBs than other organizations. Table 4 tests this claim, using models that predict either the number of GBs (other than the BoG) on which an Ordinary Member is currently serving, or the probability that an Ordinary Member serves on at least one GB. We estimate zero-inflated Poisson and negative binomial models, as well as Firth's [84] penalized likelihood logistic regression estimation [85] for the same reasons as described above in the case of resolution proposal. Summary statistics for the variables used are presented in Table 5. Given the relatively small number of GBs, organizations' centrality in Figure 9 is driven in large part simply by the number of GBs upon which an organization serves. Again, in keeping with our assumptions about institutional stewards, we would expect founding members and members of long tenure to be more likely to serve on GBs than other organizations. Table 4 tests this claim, using models that predict either the number of GBs (other than the BoG) on which an Ordinary Member is currently serving, or the probability that an Ordinary Member serves on at least one GB. We estimate zero-inflated Poisson and negative binomial models, as well as Firth's [84] penalized likelihood logistic regression estimation [85] for the same reasons as described above in the case of resolution proposal. Summary statistics for the variables used are presented in Table 5. Table 4 demonstrates that while, as expected, members of longer tenure are more likely to be members of GBs other than the BoG, founding members are not more likely than others to serve on a GB, controlling for BoG membership, which, as expected, has a statistically significant, positive relationship with further GB membership. Interestingly, we do not find an improvement in model fit when moving from a zero-inflated Poisson to a zero-inflated negative binomial regression model, suggesting that, controlling for the variables in the model, there is no additional overdispersion. In other words, the propensity of organizations that serve on multiple GBs to serve on quite large numbers of GBs is effectively accounted for by the variables in the model. Table 4. Key variables of interest from estimated models predicting current GB membership. Based on a likelihood ratio test (χ 2 = 1.39, p = 0.239), we find no statistically significant improvement in model fit from the Poisson to the negative binomial models. * = sig. at 0.05, ** = sig. at 0.01, *** = sig. at 0.001. Table 5. Summary statistics for variables used in Table 4.

Variable Summary
Total Seats on Governing Bodies

Discussion
Across a range of indicators, our findings support the claim that the RSPO is largely administered by a relatively small group of members, generally of long standing. Beginning with the observation that input legitimacy is insufficient to understand the role of private authority in economic governance [7], it is important to consider a range of daily institutional work in order to better understand power and participation within MSIs [115] (p. 419). While institutional work is often understood as a typology of activities based on empirical cases [118], here we have used the literature as a theoretical framework, directing our attention to the distribution of everyday activities to understand which organizations are running the show. As our analysis indicates, the distribution of everyday institutional work may create power relations quite different from those that might be expected from formal institutional arrangements.
Early RSPO members remain critical for the institution's continued operations. Six organizations, all of whom were either founding members or served on the first official BoG, for example, sponsored or co-sponsored 60 percent of the resolutions approved during the GAs covered by our dataset. Resolution proposal, however, is not the only way in which entrepreneurs continue to influence the RSPO. They act as institutional stewards in making outsize contributions to workaday concerns through service on the BoG and other GBs. Involved both in creating and maintaining institutions, these organizations have remained central to the RSPO throughout its history.
We believe this path-dependent source of cleavage may be as important as the more commonly remarked North/South divide in MSI politics [11,17,45]. In terms of formal membership composition, to be sure, the RSPO appears consistent with literature that suggests MSIs are driven by actors originating from the global North [13,15,19,30]. Nevertheless, some of the most central actors are from the global South, and in practice participation is more complex. In specific, the GBs generally seem to be comprised of larger organizations, regardless of their country of origin. It is noteworthy that, while we find Ordinary Members headquartered in the OECD are more likely to serve on GBs, this relationship is even stronger for members headquartered in Singapore, suggesting a combination of affluence and geographic location may be at play. In addition, we find that, controlling for BoG membership and membership tenure, organizations from the OECD are no more likely to propose resolutions in the GA than other members. Further, the profile of members' attendance at recent GAs suggests direct participation in governance may be more equitably distributed across sectors in practice than formal membership rolls might suggest.
These observations return us to the question of legitimacy. While the MSI literature often points to inequitable North/South representation as a challenge to input legitimacy [7,15,17,19], our analysis indicates challenges to legitimacy may be more complex. On the one hand, limited participation in the GAs, as well as the questions of sectoral and geographic representation raised above, could be risks to input legitimacy, but, on the other, institutional stewards' investments in the institution may be necessary for performance and, as a result, output legitimacy.
On this interpretation, the presence of a strong core of institutional stewards might shield the RSPO-and similar MSIs-from the collective action problems [55,56] that might otherwise lead to the institutional collapse. One suggestive point that speaks to this claim is the fact that so many BoG elections are uncontested (indeed, the smallholder seat remained open for the first two years of the RSPO's existence). While it is possible this is because BoG members' positions are so secure as to render them unassailable, the lackluster attendance and low level of activity at the GAs (there are only 61 distinct resolution sponsors in our dataset), tends to suggest foot-dragging, rather than fear, better explains the lack of competition.
A final consideration here is whether or not this pattern is characteristic of most MSIs. Some support for this proposition is found in Figure 10, which presents the distribution of years of tenure for a selection of top MSI governing boards, based on publicly available records. While this is only a tentative indication, it seems clear that the centrality of institutional stewards noted in the RSPO is found in some, but not all, of the top governance bodies in comparable MSIs. The well-established (and term-limited) Forest Stewardship Council (FSC), for example, exhibits relatively stable median membership tenure in their top governance bodies. Longer, non-consecutive, terms of service are possible but very rarely achieved. The more recently established Better Cotton Initiative (BCI), Roundtable on Responsible Soy (RTRS) and the RSPO itself exhibit a pattern of lengthening median tenure, each exhibiting a small number of outlier board members that are reelected continuously. While these patterns are of course affected by the diverse electoral and term rules of the MSIs represented, this straw-in-the-wind test suggests comparative analysis of the distribution of institutional work within a range of MSIs could be a promising avenue for further research on the conditions favorable to the emergence of institutional stewards. The above analysis offers the following key contributions to the literature. First, through a longitudinal study of the RSPO, we shed light on the different phases of institutional development. Whereas the work aimed at creating institutions is well understood, institutional maintenance has gained relatively little attention despite its importance [20]. Scholars tend to focus on successful implementation of institutional change [119], often in a single type of organization [120]. Second, we introduce the concept of institutional stewards to characterize the different roles that actors assume throughout the lifecycle of an institution. We find that the RSPO has largely been governed by a few organizations that started out as the initial entrepreneurs who went on to become its stewards.
Although limited participation of organizations in the development process may raise concerns over legitimacy, our findings suggest we must consider input and output legitimacy in tandem to gain a comprehensive understanding of the role of private authority in economic governance [7]. While there is a strong overlap between entrepreneurs and stewards, BoG elections are usually uncontested. The lack of competition is thus unlikely to be due to inherently low input legitimacy. Moreover, the predetermined seats for each sector help promoting a balanced stakeholder representation, although it should be noted that these are often occupied by large organizations within the sectors, further highlighting necessity of access to resources. Still, the ability of every Ordinary Member to propose and vote on resolutions enables smaller organization to assert a degree of influence over institutional development, as well. Finally, whereas organizations may gain greater influence by expending resources, their willingness to do so is central in maintaining institutions in the face of collective action problems.
Naturally, our study entails certain limitations. The reliance on public records means restricted access to data may have implications for our validity. In terms of BoG participation, for example, it was possible to establish an accurate overview of how the composition has developed over time, and as such we are fairly confident in the results. Conversely, we could only obtain complete The above analysis offers the following key contributions to the literature. First, through a longitudinal study of the RSPO, we shed light on the different phases of institutional development. Whereas the work aimed at creating institutions is well understood, institutional maintenance has gained relatively little attention despite its importance [20]. Scholars tend to focus on successful implementation of institutional change [119], often in a single type of organization [120]. Second, we introduce the concept of institutional stewards to characterize the different roles that actors assume throughout the lifecycle of an institution. We find that the RSPO has largely been governed by a few organizations that started out as the initial entrepreneurs who went on to become its stewards.
Although limited participation of organizations in the development process may raise concerns over legitimacy, our findings suggest we must consider input and output legitimacy in tandem to gain a comprehensive understanding of the role of private authority in economic governance [7]. While there is a strong overlap between entrepreneurs and stewards, BoG elections are usually uncontested. The lack of competition is thus unlikely to be due to inherently low input legitimacy. Moreover, the predetermined seats for each sector help promoting a balanced stakeholder representation, although it should be noted that these are often occupied by large organizations within the sectors, further highlighting necessity of access to resources. Still, the ability of every Ordinary Member to propose and vote on resolutions enables smaller organization to assert a degree of influence over institutional development, as well. Finally, whereas organizations may gain greater influence by expending resources, their willingness to do so is central in maintaining institutions in the face of collective action problems. Naturally, our study entails certain limitations. The reliance on public records means restricted access to data may have implications for our validity. In terms of BoG participation, for example, it was possible to establish an accurate overview of how the composition has developed over time, and as such we are fairly confident in the results. Conversely, we could only obtain complete information on a limited number of GBs. We initially considered including previous GBs in our sample, but decided to focus on those currently active, as those were the ones for which the most complete data were available [121]. Additionally, 13 resolutions outlined in the GA minutes did not indicate a clear sponsor.
Overall, it is important to note that our results pertain to a single case study. Future studies should undertake comparative analysis of other MSIs or similar arrangements to test the robustness of our conclusions here. The discussion section provides some tentative evidence for a few select initiatives, but further research clearly is required to further test these first impressions. A final consideration relates to organizations' willingness to participate in institutional development. Whereas we identify a range of defining characteristics for the actors that are more likely to become institutional stewards, our study cannot explain in detail why they are willing to do so. Throughout this study, we assumed new institutions are created when organizations with sufficient resources find an opportunity to further their interests [60] (p. 14). Maintenance work is thus simply perceived as a means to secure the persistence of organizational interests. More nuanced exploration of the motivations for participating in institutional work, however, remains a key question for future research.

Conclusions
Our case study of the distribution of institutional work in the RSPO suggests the organization's development has been driven in large part by relatively few organizations that undertake the lion's share of institutional work intended to create new and maintain extant institutions. While this dominance raises questions about input legitimacy, critical to institutional survival [15,20], it is unclear the degree to which this dominance is driven by exclusion or by apathy. To the extent that institutional stewards run the show in the RSPO, they may help balance the tension between input and output legitimacy in the face of collective action problems.
Whereas our analysis allows us to identify institutional stewards, it does not explain why these organizations take up these roles. While there is certainly an incentive for first movers to put in the labor to build institutions that will subsequently govern them as a means of expanding influence [13], future research would do well to identify the conditions under which institutional entrepreneurs have the motivation, and the endurance, to stay on as institutional stewards. In addition, it would be helpful to undertake comparative analysis of multiple MSIs to better understand the situations in which institutional stewards can contribute positively to organizations' functioning.   The final three models are aggregated across the observation period and predict whether or not an organization has ever proposed a resolution. * = sig. at 0.05, ** = sig. at 0.01, *** = sig. at 0.001. Based on a likelihood ratio test (χ 2 = 1.39, p = 0.239), we find no statistically significant improvement in model fit from the Poisson to the negative binomial models. * = sig. at 0.05, ** = sig. at 0.01, *** = sig. At 0.001.