Models of Assurance

Assurance—an intermediary’s guarantee of compliance with regulatory standards—is critical for legitimate governance within the sustainability field. This legitimacy classically depends on the degrees of separation that are needed between the RIT roles to create trust in regulators and enforce the compliance of targets. Following the emergence of the ISEAL Alliance—an apex organization of sustainability standards-setters—there has been a general shift in the sustainability field whereby standard-setters have delegated some of their authority to certifiers and accreditors. This article examines this movement, through the analysis of four different models of assurance, and reveals increasing complexity being built into private systems of regulation in the sustainability field. There is an increasing incidence of multiple actors who engage in processes of intermediation and accreditation, which is rising in importance. The result is empirical and conceptual confusion around previously sacred notions such as independence and conflict of interest as measures of regulatory effectiveness.

r esearch on regulation standards explores emerging governance configurations among public, private, and civic actors who increasingly take responsibility for regulating transnational activities, such as sustainability and labor rights (Abbott and Snidal 2009;Levi-Faur and Starobin 2014). This article focuses on changes in the regulatory structures applicable to sustainable agriculture, particularly the emergence of multiple layers of governance by Allison Marie Loconto is a sociologist, a researcher at the French Institute for Agricultural Research, and a visiting scientist at the Food and Agriculture Organization of the United Nations. Her recent book is Innovative Markets for Sustainable Agriculture: How Innovations in Market Institutions encourage Sustainable Agriculture in Developing countries (FAO 2016). correspondence: amloconto@versailles.inra.fr standards. Based on empirical work on standards within the global agri-food system, scholars have noted the emergence of a network of actors-standards development organizations, certification bodies, and accreditation bodies 1 -that exert governing power through multiple layers of oversight (Loconto and Busch 2010;Hatanaka, Konefal, and constance 2012).
Oversight and assurance of compliance with sustainability criteria became necessary because of increasing fraud and food safety concerns in long supply chains. As defined by the ISeAL Alliance (an apex organization of standardssetters in the sustainability field), assurance refers to a process of rule enforcement that guarantees compliance with standards, focusing on "demonstrable evidence that specified requirements relating to a product, process, system, person or body are fulfilled " (ISeAL 2012, 5). Types of assurance differ based on who is declaring conformance-first-party is a self-declaration, second-party is by an actor involved in the commercial transaction or by the standard-setter, and third-party is by an independent actor.
First-and second-party assurance, such as a first-party product warranty or a second-party guarantee through a franchise model of brand ownership, have long histories of use, particularly when the first and second parties have direct relationships with the buyer in a commercial transaction or when they have built trustworthy reputations. With the additional requirement of accreditation of third-party certifiers by generic, national accreditation bodies, third-party certification has become dominant in the current global food system (Hatanaka, Bain, and Busch 2005;Loconto and Busch 2010). This dominant assurance model is based on the concepts of organizational independence, scientific objectivity, and competitive pricing, all of which allow regulatory intermediaries to avoid conflicts of interest and manage risk (cf. Lytton 2014). These "relational" attributes of assurance are standardized within the International Organization for Standardization (ISO) 17065 standard for conformity assessment; 2 with the european New Approach to product safety, they have also become enshrined in the european market governance of sustainable agriculture (Fouilleux and Loconto 2016;Galland 2017).
Despite its dominance, the accredited third-party model of assurance is contested, particularly within the field of sustainability governance (Dingwerth and Pattberg 2009;Loconto and Fouilleux 2014). Debates over the legitimacy of NOTe: The author thanks Kenneth Abbott, David Levi-Faur, and Duncan Snidal for the extensive and constructive comments on this article, as well as the participants of the "Politics of regulatory Intermediaries" workshop at the Leonard Davis Institute, The Hebrew University of Jerusalem, May 2014, for adding their constructive feedback. The author acknowledges funding received from the French Institute for research Innovation and Society (IFrIS) for the research conducted with the ISeAL Alliance. Data collection on participatory guarantee systems received funding from the Food and Agriculture Organization of the United Nations and from the european Union (eU) through the ec/FAO Global Programme "Improved Global Governance for Hunger reduction" (GcP/INT/130/ec); and the res-AGorA project (responsible research and Innovation in a Distributed Anticipatory Governance Frame: a constructive Socio-normative Approach) under the eU's Seventh Framework Programme for research, technological development and demonstration (grant agreement no. 321427). different models of assurance (including hybrid first-, second-, or third-party forms) focus on the ability of private standards systems to enforce the compliance of producers (mostly agricultural) with the values and practices encoded in the standards and, therefore, to be trustworthy authorities on sustainability (McDermott 2012;Hatanaka 2010).
These debates began in the 1990s, when a number of social and environmental standards became formalized with certification systems (e.g., Fairtrade International, Marine Stewardship council, Forest Stewardship council, and national organic regulations). During this period, each scheme developed its own standards and conformity assessment systems and trained its own auditors (cf. Auld and renckens, this volume). This first period was marked by severe competition between schemes and accusations of "greenwashing," often based on the methods standard-setters used to verify compliance with their standards (i.e., self-reporting) (Bartley 2003).
Beginning in 2000, a group of standard-setters and accreditors 3 created an apex organization-the ISeAL Alliance-and began to collaborate to resolve common problems in their assurance systems. In 2010, multiple stakeholders (standard-setters, certifiers, accreditors, consultants, researchers, retailers, and NGOs) came together to create a standard for conformity assessment systems: the ISeAL Assurance code (IAc). Through this process, ISeAL members harmonized their assurance requirements.
Following the emergence of the ISeAL Alliance, there has been a general shift in the field. When discussing environmental and social governance of global value chains, some scholars claimed that the state delegated its regulatory function to intermediaries (standard-setters) who intermediate between the state and targets (usually producers) (e.g., Islam 2008;cashore 2002). Over the past 10 years, we see a shift where these standard-setters have delegated some of their authority to certifiers and accreditors. There is a double movement within this shift. First, delegating intermediary roles away from the standard-setter signifies the domination of the ISO model of accredited third-party assurance within the sustainability field; second, we see a move away from the ISO model toward new modes of regulating, which have begun to be standardized by the ISeAL Alliance into four models of assurance. In both movements, we see secondary intermediaries (Abbott, Levi-Faur, and Snidal, this volume)-namely, certifiers (first, second, and third party) and accreditors (public and private)-gaining importance in the relationships between regulators and targets.
This article examines these movements to better understand the dynamics of regulatory intermediation. Specifically, I show that there are two layers of regulatory intermediation within sustainability governance. First, and in line with Havinga and Verbruggen (this volume), there is a subset of rIT (regulator, intermediary, and target) actors who are positioned in the intermediary space between state regulators and producers, who are the targets of rules and the main beneficiaries (B) of sustainable agriculture. 4 Second, within this space, the different intermediary roles identified in the rIT framework are not carried out by a single intermediary at a single point but by many intermediaries. The identification of intermediation processes within this group of secondary intermediaries and their analysis using the rIT framework provides insights into the ambiguous roles and the shifting norms for assurance within the sustainability field.

Method and Analytical Framework
According to Abbott, Levi-Faur, and Snidal (this volume), regulatory intermediation is concerned primarily with the "relational work" that occurs between regulators (r), targets (T), and those additional actors who facilitate these relationships (I). The rIT framework envisions the regulatory process as a three-party game, whereby intermediaries (I) facilitate regulatory action between r and T. This article mobilizes the rIT framework to explore the implications of different combinations of actors in the provision of assurance for sustainability standards.
I explore four models of assurance that are currently used in sustainability governance: • Model A: Accredited third-party assurance, exemplified by Fairtrade International (FLO); • Model B: Second-party attestation and third-party determination, demonstrated by the Sustainable Agriculture Network/rainforest Alliance (SAN/ rA); • Model c: First-party attestation and third-party determination, as used in the Global coffee Platform (GcP); 5 and • Model D: Second-party attestation and determination, as found in the International Federation of Organic Agriculture Movement's (IFOAM) Participatory Guarantee System (PGS).
I analyze the actors participating in these schemes, their capacities and goals, and how each system allocates intermediary roles to different actors. I then examine the intermediation processes in each model, focusing on role performance. Steyaert et al. (2015) describe intermediation as "a process of situational articulation" (callon, Millo, and Muniesa 2007), a communication process between actors that works in two ways. First, intermediaries clarify regulatory goals and coordinate actions. These two activities together are meant to objectify the issues at stake so as to create evidence of regulatory performance. This is the classic function of audits. Second, intermediaries enable actors to recognize their own roles (their own subjectivity) in the regulatory process. In this second process, rIT actors engage each other orally (or in writing) about their roles in the process and confront various points of view, values, knowledge, and perceptions about the effectiveness of regulatory action. Practically, this is the process of negotiating and allocating responsibilities of standard-setting, accreditation, certification, and verification.
By examining how actors relate to one another in the intermediation process, I explore how each model focuses on or enhances different combinations of the intermediary properties identified by the rIT framework, specifically related to legitimacy and independence. In so doing, I provide a careful contextual analysis of the "quality of regulation." I ask: How do intermediaries differentially perform their roles in each of the four models of assurance, and what does understanding these differences contribute to debates over the quality of regulation?
Data were collected between 2010 and 2015. Between 2010 and 2012, I sat on the steering and technical committees, participated in meetings, conducted background research and followed field tests of the IAc with the GcP, FLO, Accreditation Services International (ASI) and IFOAM. 6 Between 2013 and 2015, I conducted a multisite ethnography of PGS in five countries and of the SAN/rA model in Tanzania.
Opening Up "Models of Assurance" Models of assurance consist of the (re)definition of regulatory roles between standard-setters (r), accreditors (I 1 ), certifiers (I 2 ), targets (T), and beneficiaries (B). Before the 2014 revision of ISeAL's Standards code 7 and the creation of the IAc in 2012, 8 ISeAL membership rules required scheme owners (standard-setters) to use the dominant model of assurance. From 2005, then, member schemes needed to separate their standard-setters and certifiers into independent legal entities; standard-setters were required to comply with ISeAL's code for setting standards, and certifiers had to be accredited according to the ISO guide 17065 by a national accreditation body.
This accreditation is an audit of the certifier's operations that assesses the certifier's competence to conduct audits. According to ISO requirements, the accreditor, in order to conduct accreditation audits, had to comply with the ISO/ Iec 17011:2004 guide for accreditation bodies, be a member of the International Accreditation Forum (IAF), and be authorized by the state in which it was incorporated. In this scenario (Figure 1), ISO and ISeAL act as meta-standard-setters (r 1 ) for ISeAL members (T 1 ), which include both standard-setters (r 2 ) and accreditors (I 1 ). Theses meta-standard-setters use the accreditors to audit certifiers' (T 1 ) competence (following the dotted arrows). ISeAL standard-setters then create the rules that producers (T 2 and B) must comply with and use certifiers (I 2 ) to ensure their compliance (following the solid arrows).
Over the years, ISeAL identified challenges that its members faced with the ISO approach, including a lack of sector-based auditor skills (e.g., interviewing techniques), unique practices (e.g., group certification), and the insufficiency of ISO definitions of effectiveness in terms of impartiality and replicability (ISeAL 2009). In addition, ISeAL members had long used private accreditors to ensure certifier competence, which made the accreditation element of the ISO approach cost-prohibitive and inappropriate, according to ISeAL's accreditor and standard-setter members. As early as 2009, ISeAL members sought additional guidance on these and other problems with ISO conformity assessment standards. 9 Put differently, ISeAL relied heavily on ISO standards for governing the conformity assessment practices of its members, but member standard-setters sought exemptions from some ISO requirements to adapt their systems to the local circumstances of their "beneficiaries" (i.e., producers) (cf. Koenig-Archibugi and Macdonald, this volume). Moreover, Derkx (2011) noted that ISeAL's accreditation body members performed poorly on ISO-based accreditation audits, which suggests that some of the desire to move away from the ISO model came from their own inability to comply with its rules.
The project of "redefining a credible guarantee" (ISeAL 2009, 7) was taken up within the IAc. ISeAL sought to provide the level of guarantee that is required for assurance to be credible to each audience for social and environmental standards, whether provided through self-assessment, peer evaluation, supply chain audits, NGO/trade union audits, or state authorized certification (ISeAL 2009). This was a deliberate attempt to move away from the ISO model, which defines credibility based on the organizational independence of the actors that attest and determine conformity with a standard. In the sustainability field, practitioners generally differentiate between self-reporting (attestation without external determination), verification (attestation with determination by a second or third party), and certification (attestation by another party and determination by a third party).
The first public draft of the IAc proposed four models of assurance, based on combinations of first-, second-, and third-party assurance, as explained in Table  1. I differentiate the models according to the types of attestation and determination that make up the intermediation processes, the roles of targets, the organizational independence between the rIT actors, and the level of guarantee according to the practitioner differentiation noted above. These elements all influence the legitimacy of each model. Model A, with accredited third-party assurance, is an ISO-compliant model; Model B combines second-and thirdparty assurance, with the standard-setter as the second party; Model c combines first-and third-party assurance; and Model D combines two types of secondparty assurance, again with the standard-setter as second party. A model based solely on first-party assurance (self-reporting by targets, Model 0) was excluded because it was "not deemed to offer sufficient credibility" (ISeAL 2011, 29).

Analysis of the Four Models
The identification of these four nonhierarchical, generic models allowed ISeAL to accommodate the existing models of assurance used by its standard-setter members. Through field testing, ISeAL members were directed to evaluate how their current models compared to the IAc four, thus testing the "quality" of their regulatory forms. In fact, no member's system fit the models perfectly. This testing is important because the models were not published in the final version of the IAc. Instead, they served as a reflexive device that helped to clarify the relationships between different actors in each member's system. In this way, they solidified the definition of a credible (and thus effective) assurance system based on a set of principles, rather than a fixed "model." I illustrate these intermediation processes as they play out in four cases.

Accredited third-party assurance (model A): FLO
Model A represents the accredited third-party assurance model: the use of third-party certifiers to attest to the target's compliance, and third-party determination by the standard-setter, which has no direct relationship with the targets. This model has an additional layer of determination, as national accreditors audit certifiers for compliance with ISO 17065. The difference between the ISO model and ISeAL's Model A is that T → I and T → R feedback loops are encouraged. The relationships between the actors illustrated in Figure 2 are explained below, using the example of FLO. R: FLO is one of the best-known sustainability standards. emerging from a charity shop movement, the fair trade concept was first established in 1988 under the label Max Havelaar in the Netherlands; it quickly spread through national labelling initiatives across europe and North America (raynolds, Murray, and Wilkinson 2007). In 1997, FLO was established as the Fairtrade Labelling Organizations International (a nongovernmental organization) and developed the first international standards for Fairtrade, which included a label and a certification scheme. FLO now operates a suite of standards that differ by type of producer (e.g., smallholder organizations, hired workers) and also apply to traders. Its standards cover production practices, treatment of workers, and terms of trade. FLO also has product-specific standards that define minimum prices for producers and a "social premium" that must be paid to producers and/or farm workers (T+B).
FLO retains control over the implementation, interpretation, and monitoring of its standard. FLO provides direct support to producer organizations to strengthen their operational capacity. FLO trainers work directly with producers (T+B) to interpret the standards and develop implementation strategies. FLO also monitors and evaluates its standards. Through its audit and producer support processes, FLO collects monitoring data on twelve key indicators; it also commissions impact and evaluation reports by external experts. 10 Following initial ISeAL rules, 11 FLO separated its standards-setting and enforcement activities, putting it in compliance with the ISO model.

FLO-CERT and DAkkS.
In 2003, FLO-cerT was created by transitioning FLO's internal certification team into a separate nonprofit certification body. FLO-cerT is contracted by FLO to conduct all certification audits on targets who pay for FLO certification. Since its creation, FLO-cerT has expanded its expertise and organizational capacity, taking on verification audits for other ISeAL members (e.g., GcP) and private codes of conduct. FLO-cerT also provides consulting services, including guidance on developing codes of conduct. The intermediary roles that FLO-cerT play within the FLO model include standards enforcement through tri-annual certification audits (site visits) and annual verification audits (document audits). FLO-cerT interprets the standards as part of its audit process, using certification checklists that facilitate its work; auditors also attend "interpretation" trainings. 12 FLO-cerT is also responsible for dispute resolution, related to the release of certificates and market surveillance of FLO-labelled products. FLO-cerT provides feedback to targets and the regulator through formal channels. There is a highly responsive relationship between FLO-cerT and targets, according to their complaints management statistics: "Two out of these [11] complaints triggered a change in our management system, thus changing the way we work henceforward." 13 However, according to one interviewee, the frequency and quality of the feedback between the regulator and intermediary has been reduced since the organizational separation of the two bodies. 14 Since 2007, FLO-cerT has been accredited to the ISO 17065 standard by DAkkS (Deutsche Akkreditierungsstelle). 15 DAkkS is the national accreditation body in Germany; under regulation (ec) No. 765/2008 and the German Accreditation Body Act (AkkStelleG), it is the sole provider of accreditation in Germany (about 4,300 accreditations per year) 16 and conducts accreditation audits in at least five other countries (Fouilleux and Loconto 2016). As a nonprofit organization delegated by the state to act in the public interest, DAkkS implements, monitors, and enforces the ISO 17065 standard, ensuring that FLO-cerT is a competent certifier. DAkkS follows the ISO/Iec 17011 standard, as well as the regulations of the meta-standard-setters in the field of accreditation. 17 According to FLO, FLO-cerT decided independently to pursue ISO accreditation: "As accreditation against ISO 17065 is not a Fairtrade requirement, any relationship with an accreditation body is the responsibility of assurance providers." 18 One manager explained the decision as a necessary "license to operate" in Germany (where FLO-cerT is registered), which also enabled FLO-cerT to expand its audit portfolio beyond FLO certification. 19 This model is considered credible because of the complete separation of attestation from the standard-setter. Furthermore, the certifier's capacities, interests, and goals (related to being a nonprofit, specialized certification body) have been checked by an independent, expert, and state-sanctioned accreditor. This is the only model where the use of the term "certification" is accurate. Nonetheless, a potential conflict of interest arises because of the direct payment to certifiers by targets and the lack of competition for attestation services.

Second-party attestation, third-party determination (model B): SAN/RA
Model B represents a deviation from the ISO model with a move toward a complex mix of second-party attestation (by the standard-setter or group administrator) and third-party determination (through the standard-setter, which does not have a direct relationship with the target). When certifiers are delegated attestation authority, they are not required to be accredited to ISO 17065, but they must be accredited by a private accreditor. Thus, beyond the use of a second party for attestation, what differentiates this model from Model A is that a private accreditor, which follows the rules created by the standard-setter, determines the competency of the auditor, as shown in Figure 3.
R: SAN and RA jointly regulate in this scheme. rainforest Alliance, Inc. is an international nonprofit organization, founded in 1986, dedicated to the conservation of tropical forests. It owns the rainforest Alliance certified TM seal, which is awarded to farms that meet the environmental, social, and economic standards of SAN, a coalition of conservation organizations (including rA) that had set the first standards for sustainable farming in rainforest areas in 1992. 20 Over the years, SAN has consolidated numerous crop standards into one whole-farm standard for sustainable agriculture and one standard for sustainable livestock production. It also maintains a standard for group certification, a chain of custody standard that ensures traceability along the supply chain, and an optional module on climate change. SAN standards cover ecosystem conservation, worker rights and safety, wildlife protection, water and soil conservation, agrochemical reduction, and education for farm children. 21 In addition, rA manages other standards systems, which can carry the rA Verified mark, related to forestry, carbon, and tourism, but unlike the rA certified seal, this mark cannot be used on product packages.
While SAN is clearly the regulator, rA plays a major role as a secondary regulator in implementing and monitoring its standards. Implementation is done through the creation and enforcement of rules regarding the use of rA labels, collaborations with the private sector to train producers (T+B) to meet the standard, and work on ecosystem-focused community projects (Loconto 2015). Since the creation of ISeAL's impacts code in 2010, all ISeAL members have begun to collect monitoring and evaluation data. rA is at the forefront of these efforts, with a research and evaluation program that includes three levels of monitoring. 22 Through local interpretation guidelines that are country-, product-and standardspecific, rA guides the interpretation of its standards instead of delegating this task to certifiers (SAN 2015). Feedback in this model occurs through research, interpretation guidelines, and conformity assessments, where producers (T) communicate their concerns to certifiers (I).
I: SAN/RA, IOAS and certifiers share the intermediation tasks of monitoring and enforcement within this model. There are two paths for producers (who pay for audits) to receive certification every three years (supplemented by yearly verification audits). First, both SAN members and the rA cert division of rA conduct audits. This work feeds directly back into the development of SAN standards, as the responsibility for revising the standards and the assurance system lie with rA cert and SAN.
The second path to certification (obligatory as of October 1, 2015) is based on the ISO 17065 standard, with additional SAN requirements developed by technical experts. ISO 17065 compliance is monitored by the independent SAN Accreditation Body, responsible for evaluating conformity of certification bodies (SAN 2015). Here two forms of accreditation are used: first, SAN/rA has designated IOAS (the official accreditation body for IFOAM's organic standards) as the accreditation body to assess and monitor third-party certification bodies that wish to provide audit services for SAN. This accreditation is specific to the ability of the certifier to audit to SAN standards. Second, if certifiers wish to determine certification, they must also be accredited against ISO 17065. Although SAN/rA delegated this responsibility to IOAS, some certifiers have opted to use DAkkS accreditation and obtain rA-specific accreditation separately (e.g., Afrocert and cereS). 23 In short, following the introduction of the IAc, the rA/SAN network is shifting its system toward model A: they see value in the role of accreditation and the use of third-party certifiers but prefer to work with private accreditors.
This model is considered credible because of the separation of attestation and determination activities between the two regulators. The internal attestation of the standard-setter can be believed because its verification capacity (or that of a delegated certifier) has been checked by an independent, nonprofit accreditor. Also, investment in research, monitoring, and evaluation has produced quantitative evidence of the positive impact of the standard on sustainability, thus building a good market reputation.

First-party attestation, third-party determination (model C): GCP
Model c relies on first-party attestation by targets (4c units comprising farmers' organizations) plus an audit by a third-party to determine the veracity of their attestations. There are similarities between models c and A, as reliance on accredited third-party certifiers to conduct audits follows the same R → I → T process at the accreditation level. There are also similarities between models c and B, where the regulators have separated the attestation and determination roles between two different organizations. The main differences between this model and the previous two are the separation of targets and beneficiaries and the creation of a new intermediary between the standard-setter and the certifier, as shown in Figure 4. R: The GCP is the new name for the common code for the coffee community (4c), which began in 2002 within the International coffee Organization (IcO), at the insistence of the German State and with the support of the German coffee Association and German Development corporation (Auld 2010). 4c became a membership association in 2006, with producers, roasters, and civil society groups making up the members. Industry was particularly strong within the process (Levy, reinecke, and Manning 2015, 19). The 4c process produced a standard that aimed to eliminate the worst practices of coffee growers. With its 2016 rebranding, GcP functions more as a platform for advocacy and dialogue. Nonetheless, as the standard-setter in this model, GcP maintains the 4c Baseline common code, a business-to-business standard without a consumer-facing label.
GcP is responsible for monitoring the assurance system that maintains ISeAL's assurance and impacts codes. GcP receives aggregated data through all accredited operators, enabling it to evaluate the impact of its standard. 24 GcP also requires its trader, exporter, roaster, and retailer members to train and provide advice to producers on implementation through field and farm education and training programs. 25 Finally, GcP seeks to create synergies-and benchmarks-with other ISeAL member standards in the coffee sector (e.g., FLO and SAN/rA). The GcP system emphasizes transparency and feedback, enabling GcP to lead efforts toward a common agenda for the industry. 26 I: CAS, 4C verifiers, and accreditors. The GcP model works as follows. Producers (B) are organized into 4c units (T), which can be cooperatives, plantations, or contract farming units. 4c units declare their compliance with the standard to the newly created (April 2016) intermediary, coffee Assurance Service GmbH & co. KG (cAS). cAS delivers all determination services-such as issuing licenses, verification visits, trainings, and commercial reporting-to 4c units, roasters, and other customers. The frequency of verification visits is determined by random sampling, a risk assessment calculation, or a combination of the two. cAS's main intermediary function is monitoring and enforcing the standard. This includes (1) approval, management, and training of 4c verifiers, certifiers that operate in coffee-producing countries and are accredited under ISO 17065 (with the optional agricultural scope) by a national accreditation body; (2) monitoring compliance of 4c units with the 4c code of conduct, reviewing their self-assessment reports, and taking license decisions; and (3) commercial reporting by final buyers. These activities provide the data needed for cAS to provide information about commercial activities and lessons learned to the standard-setter. 27 In this model, attestation by the targets is credible because they are not paying for certification (thus removing a concern about conflicts of interest), and the standard-setter has created a separate legal entity to control certification decisions based on the self-reporting of targets. These reports are verified through an audit by approved certifiers. certifiers' reports can be believed because an independent, state-sanctioned accreditor has checked each certifier's capacities and goals. Finally, as there is no on-the-pack label, only actors in the industry know about the standard, reducing the risk of fraud for financial gain.

First-party attestation, second-party determination (model D): IFOAM PGS
Model D follows model c in using first-party attestation by targets and beneficiaries, but second-party determination in this model is conducted by a PGS committee, which includes standard-setters, intermediaries, targets, and beneficiaries. Participatory Guarantee Systems (PGS) "are locally focused quality assurance systems. They certify producers based on active participation of stakeholders and are built on a foundation of trust, social networks, and knowledge exchange." 28 Key elements of the model include multiple actors with diverse responsibilities in the PGS committee; peer review rather than audits; and little or no payment for release of the certificate, as shown in Figure 5.
(r) of its standard. While not official accreditation, IFOAM operates a PGS recognition program, in which the IFOAM secretariat evaluates whether a PGS operates in accordance with the PGS principles, and verifies the integrity of the PGS vis-à-vis the principles of organic agriculture. evaluation is free, and PGS schemes that pass may use the IFOAM PGS logo.
Since 2009, IFOAM's main role in this model is to set a global agenda for acceptance of PGS as a viable assurance alternative (IFOAM 2014). As of 2015, there were 123 functioning PGS initiatives, with another 110 under development in more than seventy-two countries. IFOAM's influence is apparent in its role in the IAc development process. By the time ISeAL began developing the IAc, IFOAM had already left the ISeAL Alliance and was represented officially via IOAS (Loconto and Fouilleux 2014). IFOAM created IOAS in 1997 to separate the standard-setting and accreditation roles and to standardize organic accreditation around the world (Katto-Andrighetto 2012). Nonetheless, IFOAM was represented on the IAc committee and field tested the code. Its influence is apparent in the inclusion of model D, which most participants did not consider a valid option because, within a PGS, a single group carries out all three rIT roles. has adopted three standards with corresponding labels (organic, in conversion, and organic ingredients).
The PGS is made up of a mix of targets (T+B), who take on the r-I roles in this model. The NOA board carries final responsibility through ratification of decisions of the assessment team and authorizes use of the registered NOA trademark. The NOA administrative team organizes the logistics of documentation, preassessments, and assessment visits (every six months). The assessment team is made up of members who have received IFOAM/IOAS training. It is responsible for conducting preassessments and on-farm assessments, compiling assessment documentation, and making recommendations to the board. All NOA members (consumers, traders, or other parties) may act as observers to ensure transparency. These may include nonorganic farmers, market representatives, or consumers (additional beneficiaries in this model). NOA PGS relies on farmers to participate in the peer review system.
A number of elements make this model credible. First, audits occur more frequently (every six months on average). Second, the localized nature of the social control provides significant peer pressure to comply with commitments. Third, the body that determines compliance is not remunerated for this service, removing a significant conflict of interest. Fourth, the involvement of multiple actors (consumers, researchers, and public officials) in local systems and audits provides direct incentives for compliance, because these actors are the buyers of the certified products and communicate any dissatisfaction with production practices. effects on the Quality of regulation As a way to objectify the issues at stake, members of the ISeAL technical and steering committees created a matrix to evaluate the capacities and goals of the different types of intermediaries used in each model. As Table 2 illustrates, the independence of the actors who attest to and determine conformity is the most important intermediary capacity in sustainability standards. However, if we examine ISeAL's justifications for accepting different degrees of intermediary independence, it is clear that questions of legitimacy (i.e., "scalable," "widely accepted," and "accessible"), cost (i.e., "expensive" and "minimal cost"), and interests (i.e., "risk" and "fraud") were also key concerns for the quality of regulation. This reflects a shift in thinking about regulatory quality from focusing only on independence, to focusing on independence in relation to impartiality and transparency.
With the IAc, an intermediary between the regulator and the targets became mandatory for ISeAL members. However, the obligation of organizational independence, central to the ISO standards, was not replicated. Instead, ISeAL introduced a measure of flexibility to balance the capacity of intermediaries to be independent and legitimate against the costs of compensation and interest, while requiring transparency. Thus, in terms of the rIT framework, we see trade-offs between the capacities and goals of intermediaries. ISeAL also opened up intermediation beyond certifiers alone: even in its ISO-inspired model A, room was made for accreditors, standard-setters, targets, and beneficiaries to play intermediary roles. For example, in FLO, intermediary roles are split between the standard-setter and certifier. Because the accreditor is a national, state-sanctioned body, there is no direct relationship between the standard-setter and the accreditor. This has resulted in the standard-setter retaining some intermediary roles that might otherwise have been delegated to the accreditor.
Through the reflexive processes of revising the rules for assurance systems and testing individual systems against the new rules, standard-setters, certifiers, and accreditors could recognize their own roles in the intermediation process. For example, GcP claimed: Although none of the models described in Section four matches fully the 4c setup, we found that Model B is closest to how the organization is made up. The administrative entity of the so called "4c Unit" can be considered the "second party," which conducts a self-assessment for all the "Business Partners" belonging to the unit, while "third party" verifiers are contracted for verifying correctness of the self-assessment. The final decision is then actually made by the verification department of the 4c Association, which if looked at the Association as a whole is the "first party." Others, however, pointed out that the use of self-assessment by 4c units, and their accepting certification decision by 4c rather than by verifiers, pushes the GcP system into the realm of model c. recent changes in the system following its rebranding bring GcP closer to model A.
The above discussion illustrates how intermediary roles are not always limited to core intermediary actors. Moreover, as additional intermediaries are added to the network, their relationships become fluid. The blurring of these boundaries raises the question: Which models are most resistant to capture or takeover and, therefore, provide a higher quality of regulation?
In the cases described above, we see four different approaches. FLO relies on the institutional infrastructure of the national accreditation body system to ensure that FLOcerT remains only a certifier. But this model empowers accreditors to intervene in areas of regulation not anticipated by FLO. The SAN/rA model focuses on choosing certifiers that are mission-driven: "In the SAN/rA accreditation system there is a pre-evaluation process with requirements for certifiers around mission fit, organizational and financial stability, as well as our goal to reach broader geographic coverage." 30 In this way, SAN/rA ensures that its intermediaries are committed to following its rules. The external control on certifiers, by accreditors that audit to a scope defined by the standard-setter, reduces regulatory capture by creating competition between public and private accreditors.
Models c and D are more vulnerable to regulatory capture, as they are designed to provide intermediaries and targets with a much stronger role in regulation. In both models, intermediaries are responsible for the rules that govern the conformity assessment practices. During a technical committee debate, an accreditor reflected on the differences between models A and D: "PGS has no separation and SA8000 was forced by ISeAL to create a separation between their standard-setter and their accreditor. For PGS, they value participation, while the SA8000 model values independence." After discussion about the values of participation and independence, the committee concluded that the prioritization of these values, and thus the role of intermediaries, depends on the other actors in the process. For example, "if the user of the standard is carrefour, then they need that level of assurance based on independence because of national legislation, but the famers and consumers in France who are trading directly don't need it … but there is a need for transparency in both instances because you cannot claim independence if you don't have it." 31 This recognition of different values (participation or independence and transparency) within an assurance system brings to light a shift in the regulatory goal of ISeAL members: from compliance to learning. The current agenda of ISeAL and its standard-setter members is to facilitate learning by all actors, rather than enforcing rule compliance. This is seen in the examples of FLO and SAN/rA, which retain responsibility for feedback and learning despite significant outsourcing of other intermediary roles. In the GcP and PGS systems, intermediaries are less independent, and their aim is not a "robust market" declaration; learning is thus considered to be how "compliance" is best assured. Indeed, by recognizing these different values, ISeAL encoded the concept of learning into the IAc and provided guidance on knowledge sharing as part of the assessment process. This is considered to be "a form of risk mitigation, because informed clients are more likely to follow the standard if they understand it. rather than prohibit this activity, which can be beneficial for all parties, scheme owners need to ensure that the advice provided to clients is accurate and is available to all clients in a consistent fashion. This way, there is less opportunity for one client to be favored over another." Thus, impartiality of intermediar-ies and transparent information can be achieved within all models, as long as attestation is separated from determination. This can be done by including different actors in the intermediation process.
conclusion The models of assurance presented in this article demonstrate the increasing complexity being built into private systems of regulation. Not only do I demonstrate that there is a strong role for regulatory intermediaries within these systems (as the rIT framework suggests), I also show that multiple actors engage in processes of intermediation. Thus, while reconceptualizing private regulatory arrangements as a threeparty game is a clear theoretical advance, this characterization does not go far enough to capture the dynamic processes at play in the provision of assurance (cf. Havinga and Verbruggen, this volume). In the case of sustainability standards, we see an increasingly important role for accreditation, which adds a second level of intermediation to models of assurance. The result is empirical and conceptual confusion around previously sacred notions such as independence and conflict of interest as measures of regulatory effectiveness (Lytton, this volume).
envisioning the organization of their schemes as "models of assurance" enabled ISeAL members to create evidence of regulatory performance and clarify their roles in the regulatory process. While the models represented a collective effort to escape from ISO path dependency (cf. Kruck, this volume), 32 the IAc fell short. The 2014 state of sustainability initiatives (SSI) report found that "75 per cent of the initiatives reviewed were either ISO 17065 compliant or apply an accreditation process, emphasizing credibility as a primary driver in the voluntary sustainability standard sector" (Potts et al. 2014, 51). While the SSI data were collected before the IAc became fully operational, they point to a growing trend in the sustainability field toward regulatory innovations and the need for ISeAL members to continue to push the boundaries of the ISO mold to "improve [their standards'] effectiveness and achieve greater scale." 33 Notes 1. Standards development is the elaboration of technical norms (standards) for products, services, or systems; certification is "the provision by an independent body of written assurance (a certificate) that the references