Risk and accountability: Drivers for change in network governance. The case of school restaurants governance in a Swiss city

: Growing requirements for accountability and risk management put decentralized models of public governance under pressure. This article investigates the drivers for change from a completely decentralized, network-oriented model to a more centralized, and procedural governance model of school restaurants in a Swiss city. It focuses on the pressures and challenges that this municipality faces in terms of risks and accountability in order to identify the conditions in which network governance can be successful. We applied a qualitative approach that combined conducting 25 semi-structured interviews of main stakeholders and analyzing documentation. We found that increased demand for school meals from families, the perception of increasing exposure to insufficiently managed risks associated with growing accountability requirements constitute the main drivers for change to the centralization of certain risk-sensitive, costly, and low social purpose activities, thus providing the municipality authorities with more control over the system while preserving the associative function.

ABOUT THE AUTHORS Nathalie Brender, U.S. CPA (United States Certified Public Accountant), is a professor of risk management, governance, accounting and auditing at the Haute école de gestion de Genève, University of Applied Sciences Western Switzerland. She is currently teaching and conducting research in corporate governance, risk management, and auditing. Her past professional experience includes management positions in risk consulting, auditing, as well as financial reporting in the private sector.
Bledi Yzeiraj, CIA (Certified Internal Auditor), was a research assistant at the Haute école de gestion de Genève, University of Applied Sciences Western Switzerland and is currently working in the private sector. He holds a Master in Public Management and a Bachelor in Business Management from the University of Geneva. At the Haute école de gestion de Genève, he completed research on management auditing and corporate governance.
Florian Dupuy is a research assistant at the Haute école de gestion de Genève, University of Applied Sciences Western Switzerland. He is a Swiss certified expert for accounting and controlling and holds a Bachelor of Science in Business Administration. His research interest is in corporate governance, and the evolution of auditing and control professions.

PUBLIC INTEREST STATEMENT
Public organizations worldwide are coping with new forms of decentralized and participative governance. The delivery of public services has evolved from governance arrangements characterized by centralization and hierarchy towards more interactive and participative decision-making and services co-creation. Within this context, our paper contributes to the literature of governance as it describes an evolution in the opposite direction in the case of the governance of school restaurants in a Swiss city and investigates its causes. This study provides public actors with action points to reassess the roles of the participants in the network with di erent organizations of tasks and activities to be completed. The network governance model can be maintained with the concentration of low socialpurpose functions such as purchasing, production, billing, in order to release the non-profit actors from the pressure of e ciency. These changes are expected to allow them to concentrate on their social purposes and provide the municipality authorities with more control over the system.
The role of the third sector, which is constituted by voluntary or non-profit organizations, has also been discussed by many authors (Crampton, Woodward, & Dowell, 2001;Waltzer, 1988), as its involvement in policy implementation is not new and goes back to the nineteenth century. The evolution of the implication of the third sector into network governance has often been addressed from the co-production or co-management angle (Considine, 2003). Multiple studies focus on the complexity of the decision-making process in network governance and on the tensions that exist between public institutions and other non-profit or for-profit organizations, or address the conditions that make a strategy of network governance e ective (Considine, 2003;Kenis & Provan, 2006;Kim, 2006).
Within this context of evolution from governance arrangements characterized by centralization and hierarchy toward more interactive and participative decision-making and services co-creation (Considine, 2003), the present study contributes to the public governance literature as it describes an evolution in the opposite direction and investigates its causes. Our research question (RQ) is as follows: What are the drivers for change from a completely decentralized, network-oriented model to a more centralized and procedural public governance model?
The aim of this paper is to investigate the reasons why the municipality authorities launched a centralization process of management of school restaurants in a Swiss city, moving from a completely decentralized network governance model, which involved parent associations among other stakeholders in the delivery of public services, to a more procedural and centralized model. This study is based on 25 semi-structured interviews and extensive documentation analysis that address production and distribution of meals as well as finance and administrative activities.
In this study, we found that increased demand for school meals from families, the perception of increasing exposure to insu ciently managed risks associated with growing accountability requirements constitute the main drivers for change to the centralization of certain risk-sensitive, costly, and low social purpose activities, thus providing the municipality authorities with more control over the system while preserving the associative function. This paper begins with a brief review of the literature on public governance models. Then, we present our case study, focusing on the actual challenges and pressures that the system faces. We continue our analysis of the governance of school restaurants from the perspective of risk management and accountability and present our findings. The final section provides our conclusion and our hypothesis that highlights the main contributions of our work and its related limitations, and suggests potential areas for future research.

Public governance models
Traditional public administration assumes passive clients with bureaucracy in a central role of policy-making and the implementation cycle, which implies the hegemony of the professional workforce in the service delivery system. Thus, the focus is put on the set of administrative rules and guidelines, with the dominance of the rule of law implying procedural governance (Considine, 2001). This conceptualization is based on the assumption that the welfare state of the post-1945 era will meet all the social and economic needs "from the cradle to the grave" (Osborne, 2006). With the model of the welfare state declining and the increasing complexity of public goods to provide, the organization of the traditional public administration has been increasingly criticized and has shifted focus to the quest for new public management models (Ostrom & Ostrom, 1971).
The traditional public administration moved in the late 1970s to what is known as new public management, based on the assumption that the application of private-sector managerial techniques to the production of public goods would lead to improvements in the e ciency and e ectiveness of these services (Thatcher, 1995). This assumption led to the consideration of an active-service consumer and to contracting out through several public-private partnerships, with an emphasis upon input and output control concerning public services. The public services or public goods were disaggregated to their basic units, and an increased focus was put on cost management (Hood, 1991). Several critics have also addressed the new public management benefits, as the impact upon service performance seems questionable (Pollitt & Bouckaert, 2005). By focusing mostly on the intra-organizational aspects of public management, the new public management is being perceived as limited in its ability to understand and explain the complexity of public governance in a fragmented and pluralist world (Rhodes, 1997).
In criticizing the traditional public-administration paradigm and the new public management, several authors proposed a shift in public management theory toward a more holistic theory with broad consideration of multiple stakeholders' implications, such as the third-sector co-production of public services. The role of citizens and the third sector in the provision of public good, especially of welfare services, has been broadly discussed. Many authors argued for an increased involvement of citizens in the distribution and production of public services in more participative and decentralized forms of service provision (Waltzer, 1988) at a stage of associative democracy where the state devolves as many functions as possible to civil society, thus providing the funds needed (Hirst, 1996). The citizens then contribute to the provision of public services together with public agents through voluntary e orts, in order to enhance the quality and/or the quantity of the services they receive. This process is also known as co-production (Pesto , 2006).
Coproduction is the central element in network governance, providing a direct channel of communication with users. Communication is indeed one argument for enhancing the quality of public services within a coproduction system (Vamstad, 2012). Hirschman (1980) even identifies a "voice" function in the communication channel through which users can communicate their dissatisfaction to service providers. Furthermore, network governance aims at enhancing the collaboration of the stakeholders, thus increasing the community's social capital by investing in relationships (Moran, 2005).
It seems that several actors are to be specifically included in the implementation process through more interactive decision-making and stakeholders' involvement. The concept of governance is changing, and several authors have linked this with the concept of network in the literature (Kickert, 1997;Rhodes, 2003), focusing on the complex processes of interaction and negotiation in a network of public, private, and voluntary or non-profit organizations (Klijn, 2008). Building on this approach and based on this context, Osborne (2006) introduced the new public governance theory, which intended to capture these realities and complexities by focusing on inter-organizational networks in the plural and pluralist nature of the state.
In accordance with the above models of traditional public administration, new public management, and new public governance, Considine (2001) uses a four-governance model framework (procedural, corporate, market, and network governance) to describe characteristics of transition from centralized to decentralized models in outsourcing of public services.
The procedural model reflects a public administration with highly centralized activities and hierarchical organization (Considine & Lewis, 1999). In this type of model, rules, good practices, and protocols are clearly defined. This model is aging, o ers little flexibility to its stakeholders, and is not suited for outsourcing.
The corporate model has been developed in response to the lack flexibility of the first model (Considine & Lewis, 1999). Outsourcing is assimilated to a loss of control, in particular over public spending. In consequence, new control practices and indicators are introduced in planning, budgeting, financing, and reporting activities in order to provide support for decision-making.
The third model, called the market model of governance, is di erent from the two first models (Considine & Lewis, 1999). It is characterized by a participative dimension and establishes the relationship based on a contract. One rationale behind this model is that announcement of public contracts creates greater pressure on costs on the supply side, which is in the end beneficial for the organization. Indeed, suppliers will be willing to provide more e orts and attractive conditions in order to get the contract. The public administration is then expected to optimize its costs and improve the quality of services provided.
The network model emphasizes participative and interactive organization and co-production of services (Considine & Lewis, 1999). Network governance appears as a distinct form of coordinating economic activity (Powell, 1990) characterized by interactive decision-making and stakeholders' involvement. This model operates with more flexibility and a low degree of standardization (Considine & Lewis, 1999).
The implicit assumption behind this framework is that management of expanding public activities should evolve from a centralized and hierarchical model (procedural) to a more sophisticated model with an emphasis on co-creation and a robust network of service providers. As we stated above, several authors call for this kind of evolution toward network governance in order to serve the citizenry more e ectively, mainly because of the linkages between customers and suppliers of public goods (Mandell, 2001).
Nevertheless, interactive governance such as network governance does not systematically imply e cient cooperation among stakeholders nor increased quality of provided solutions (Edelenbos & Klijn, 2006). There is indeed a considerable management e ort that goes into the management of this network, which requires skills, knowledge, and technology that public entities do not always possess (Goldsmith & Eggers, 2004). A large number of studies focus on the complexity of decision-making in network governance and on the tensions that exist between traditional public institutions and other non-profit or for-profit organizations (see Klijn, 2008;Sørensen & Torfing, 2003). Some studies are concerned with the conditions that make a strategy of network governance a success or a failure.
Provan and Milward propose a three levels' framework to evaluate public-sector organizational networks (Provan & Milward, 2001). They provide criteria to evaluate e ectiveness at each of the three levels: community, network, and organization/participant levels, recognizing that assessing network's e ectiveness is a di cult enterprise as stakeholders' interests are diverse. Satisfaction of clients' needs, service quality, and value creation as well as building social capital is at the heart of the community level's evaluation. The organizational network would then benefit from strong external support and legitimacy. According to institutional theory, the concept of legitimacy is a focal point. The fact that the organization complies with the key stakeholder's expectations is a means of obtaining and maintaining resources (Oliver, 1991). While the cost of the service is also one criterion at the community level, it becomes a critical element at the network level, in other words, for the funders and administrators of the network (Provan & Milward, 2001). The network's attractiveness, its range of services and quality of services as well as the quality of the relationship among the network's actors are key elements of the e ectiveness assessment. At the organization/participant level, the following four main criteria are considered for e ectiveness evaluation: client outcomes, legitimacy, resource acquisition, and cost (Provan & Milward, 2001). To sum up, e ectiveness at one level is not enough to declare the network's organization e ective as a whole and interactions among levels should also be taken into account in the comprehensive assessment.
Building on their previous work, Milward and Provan argue that network's e ectiveness depends on the network structure (that should be integrated and collaborative) and the network context (that should be characterized by the system's stability and resource munificence). They measure the degree of satisfaction of clients in relation to cooperation and stability of the network (Milward & Provan, 2003). Milward and Provan insist on the necessity to establish clear principal-agent relationship and relational contracts rather than competition in order to ensure accountability for performance.
It appears that the literature also recognizes accountability issues in relation to network actors' risk ownership when it comes to ensuring enhanced service quality in the co-production of public services (Bovaird, 2007). Brandsen and Hout (2006) argued that there is a natural conflict between the e ciency goals and social goals of an organization and the accountability mechanisms that are defined accordingly. Accountability is also linked to the legitimacy of third-sector associations, as an association is considered to improve its legitimacy if it is proven accountable (Mason, 2010). Mulgan (2000) captures this evolution, stating that accountability is perceived as a political value which places more trust in the independent judgment of public servants.
In light of the context described above, the aim of the present study is thus to contribute to the body of literature on public governance models within a post-new-public-management context by investigating the reasons for the municipality's launch of a centralization process of school restaurants in a Swiss city, moving from a completely decentralized network governance model that involved parent associations among other stakeholders in the delivery of public services to a more procedural and centralized network governance model. We will focus on the pressures and challenges that this association faces in terms of risks and accountability in order to identify the conditions in which network governance can be successful. Finally, we stress the limited nature of our study, whose purpose was to serve as an introductory exploration of the factors that influence changes in public governance of school restaurants. Our findings cannot be extrapolated to all Swiss municipalities, but they allow us to devise a stricter and more rigorous methodology for further studies based on the combinations of interviews, questionnaires, or quantitative surveys.

Case of a Swiss city
The department of education of a Swiss city of about 200,000 inhabitants launched a study of the governance system of school restaurants to evaluate the current system and envisage alternative models. We chose a case study approach to allow an in-depth analysis of this public governance system. This project included about 25 interviews with the responsible associations and stakeholders and an extensive documentation analysis.
Thirteen associations currently provide school meals and manage school restaurants, which amounts to about 680,000 meals served yearly for about CHF 5.4 million in revenue. The municipality covers the financial deficit of these associations, which employ about 470 volunteers and 100 employees to provide this service for families. The system relies on these 13 associations to produce meals from 8 sites and also deliver to about 40 restaurants and 1 public organization that provides childcare personnel to take care of children during the lunch break. Additionally, in some cases, associations rely on private suppliers that produce the meals and deliver them to schools or serve children. Furthermore, the associations are grouped in a federation for the purpose of collaborating through sharing information on their practices. Finally, the social services of the municipality take care of families in di cult conditions that cannot pay for the meals. The department in charge of the municipality schools is ultimately responsible for managing this system.
The subsequent evolution challenged this system and put municipality personnel and the associations, volunteers, and employees under pressure. First, demand for the service has increased constantly over the past 10 years, by about 4-10% per year. Second, families' preferences regarding school restaurants have changed. For example, families in which both parents work and earn comfortable revenues use school restaurants continuously from kindergarten, and mono-parental families increasingly adopt this service as well. The first school restaurants were created at the end of the nineteenth century to take care of workers' children who were left in the streets, but they are now extensively used by all categories of the population. Accordingly, demands for the service quality improvements have increased as citizens have also become more sensitive to services linked with childcare. Finally, foreseen changes in school schedules-a reduction of the lunch break, for example-would render impossible the current method of providing lunch service in two rounds, and necessitate the search for other infrastructures. These developments also have financial implications, as the municipality has to pay about CHF 3.5 million per year in order to cover the associations' deficits, which represent about 70% of the associations' revenues.
In terms of public governance, the current system can be associated with a highly decentralized network governance system, with the production and management of school meals ensured by the associations and the financial means granted by the municipality. The associations issue invoices to parents, collect receivables, and pay their suppliers and employees (i.e. cooks, administrative personnel). The municipality provides the infrastructure (school restaurants located within schools, kitchen equipment) and energy and pays for the transportation of food. The associations provide the municipality's authorities with financial statements (audited since 2010-2011) and budgets, and on this basis, the municipality covers the financial deficits of the associations. Some of these associations have a long history and were created at the end of the nineteenth century as charity organizations. At that time, they were funded by private donations. Later, other parent associations were created to provide lunches to children. The municipality inherited this history and the di erences in structure and functioning of the associations when it decided to finance the associations by paying first for an amount of each meal served, and then changing to subsidize losses in 2005-2006.
The system is heterogeneous and complex in terms of information and financial flows. It involves a significant number of actors with di erent missions, organizational constraints, and objectives. Old charity associations and parent associations do not function similarly, associations with production capabilities have more room to absorb costs than associations buying meals, and not all associations use the same invoicing and collection procedures. Also, the associations work closely with the public organization that hires and provides the childcare personnel responsible for the children during the lunch break. Relations between childcare personnel, volunteers, association personnel, and even school personnel are key to the process and subject to various interpretations in the di erent school restaurants. In addition, the municipality contracts directly with the transportation company to ensure the transportation of meals from production sites to restaurants and with an external expert who follows up on unpaid invoices. Finally, adequate and timely information about families encountering financial di culties and requiring assistance is not systematically shared between the associations, the education department, and the social services.

Research design
We applied a qualitative approach that combined studying documentation, conducting in-depth interviews, and collecting and validating key figures through a pre-defined grid of analysis. We chose this approach because our goal was to derive meaning from the data obtained from these di erent sources of information (Roller & Lavrakas, 2015) to explain the drivers for change in the schools restaurants governance model of the city under study. We focused on the organization, production, and distribution of meals; financing system and invoicing process; and administrative activities.
In-depth interviews have been chosen as the main data collection method of our qualitative research design. We prepared an interview guide that covered all relevant issues with all interviewees and also allowed flexibility to integrate particular responses. Our qualitative research design take took into account human observations, interactions, and discussions with the interviewer (Leavy, 2014) as well as documentation provided by the interviewees and the authorities.
The scope of our in-depth interviews research study included interviews with representatives of all actors involved in the network: municipality authorities, school restaurant associations, childcare organization, social services, and the collection agency for unpaid invoices. We interviewed the presidents of the 13 school restaurants associations, acting in fact as executive directors, and the heads of kitchen of the 8 cooking facilities that deliver the meals to the school restaurants. A standard interview protocol was established and systematically used in the interviews of the associations' committee members. In total, we conducted 25 face-to-face semi-directed interviews with 40 respondents as we interviewed in some instances two persons together, for example, the association president and the kitchen head or a kitchen head and his/her assistant. We also interviewed representatives of two other municipality authorities who were involved in the same activities for comparative purposes. This approach allowed us to cover the entire process and gather as much information as possible that we could confront and cross-check in our analysis. Figure 1 gives the breakdown of the interviewees by function in the network. The majority of interviewees can be found among presidents of associations, and secondly heads of kitchen, which can be explained by the fact that there are 13 associations in the network with 8 kitchens to ensure the service to school restaurants. The representatives of the municipality authorities in charge of the subsidizing and control activities of the school restaurants included the school services executive director, the head of finance, and the head of operations, as well as the dietitian.  Data collected through interviews were documented and validated by associations using a standard table of information that used identical metrics. Information obtained during the interviews was cross-checked by the study of documentation, mainly the validated 2009-2010 financial statements, and, when available, audit reports as well as information obtained during the interviews of other stakeholders. We also served as moderators in two sessions of working groups representing main stakeholders in order to work on the future model of governance. One group focused on organizational, administrative, and financial activities, while the second group worked on the childcare aspect and interactions among di erent stakeholders during the meals.

Assessing and managing risks
An important element of network governance is cooperation among government, non-profit, and sometimes private organizations to help ensure a high level of service coordination. Communication is considered as a factor for enhancing the quality of public services within the coproduction system according to Vamstad (2012). In our case, school restaurants' users have the opportunity to express their "voices" (Hirschman, 1980) to the members of the committees of the parent associations that represent nearly half of the meals produced. The feedback should then be shared among all the different actors in the network. However, communication through the network in the case of the Swiss city seems to be very limited and not managed through a unified protocol. Communication mostly relies on informal channels and is not institutionalized. While investment in relationships (Moran, 2005) and coproduction are considered as central elements of network governance, there seem to be very few meetings or discussions between the old charitable associations, the parent associations, the childcare organization that is in charge of the personnel accompanying the children during the break, and the for-profit producers that in some cases produce and sell the meals to the associations.
A committee is set up and meets once a year to represent a federation of the associations, the public organization that provides childcare personnel to take care of children during the lunch break, and the boroughs (communes) that include the Swiss city municipality that we studied. Nevertheless, the federation representing the associations has encountered several di culties in completing its mission; partly because the associations that are represented are heterogeneous and do not systematically rely on the federation in order to address matters of concern to them. In addition, there is no clear designation in the network as to who should be ultimately in charge of optimizing the procedures and solving the apparent problems, leaving the overall responsibility of the system to the municipality. This lack of communication and coordination generates supplementary costs in a system that is evolving toward a complicated and more heterogeneous form with multiple actors, each of whom use di erent working methods.
Provan and Milward argue that transaction costs may be assumed directly by network members but also indicate that in formally and taxpayer-funded public-sector networks, however, network growth and maintenance is often led, coordinated, and governed by a central, local administrative entity (Provan & Milward, 2001). In our case, the city authorities incur transaction costs such as communication, coaching, and training of associations' personnel, control and validation of budgets and accounts. The associations consider human resources management and in particular hiring and managing 470 volunteers and 100 employees as significant transaction costs. Collaboration in sanitary and financial audits as well as contributions to obtain quality labels is also time-consuming for the associations' personnel. While these transaction costs remain di cult to estimate on both sides, they result in an increase of the service cost, impacting the network's e ectiveness.
In addition, and based on our case, we observe that communication is not considered as a factor for enhancing the quality of public services within the coproduction system, which contradicts the argument developed by Vamstad (2012). In order to make communications more e cient and provide relevant information to all parties, a common framework should be provided. Positioning the discussions in a more risk management-oriented framework could provide a common ground for understanding the concerns of the di erent stakeholders and network players, especially the one that outsources and manages the network, such as the Swiss city under study in our case. There is also a significant body of literature concerned with outsourcing risks, most of which focuses on financial risks and that has recently dealt with uncertainty related to the tasks to be performed and the capacity of the producer to deliver the requisite service (Ball, Heafey, & King, 2003). For example, Padovani and Young (2006) proposed a three-dimensional model to assess the risks of outsourcing the municipality, arguing that the governance model (procedural, corporate, market, network) should be chosen over this risk assessment. The three dimensions concern the sensitivity of the service to the citizenry, the competitiveness of the market for the producer, and the level of di culty of switching vendors if a given service provider's performance is unsatisfactory. The assumption behind this model is that the more risky the service is to provide, the more sophisticated the nature of the governance model that must be adopted and the more complicated it will be to manage it.
This risk approach cannot only be used to choose between a procedural and more complicated governance model, such as network governance, but also to better communicate responsibilities and concerns of delivering a public service within the network in terms of e ciency and e ectiveness. For instance, all the actors in the network could take into account that services linked to childcare are perceived by society as risk-sensitive. When it comes to producing food in school restaurants, sanitary risk is perceived as very high, and the current system does not guarantee the same quality of meals among all of the associations that prepare them. The municipality mandated a diagnostic to assess the e ciency of the production process and quality of the food produced by the associations. Furthermore, in this context, several private actors are exhibiting good performance, making the market competitive both in the quality of meals and their cost. Switching to private vendors for this Swiss municipality does not seem to generate high costs, as examples in the region already exist and are proven successful. This should give consideration to the associations when reviewing the actual financing arrangements and the terms of addressing the financial risk.
Financial risk is also perceived as significant and has to be considered at three levels: cost control, liquidity risk, and fraud risk. City authorities have subsidized associations to produce, deliver, and serve meals at school restaurants, first using a pay-per-meal system that was replaced in the year 2005-2006 with the subvention payments to cover the associations' deficits. The actual system does not favor cost control, which however represents a crucial element of network e ectiveness as advocated by Provan and Milward (2001). The associations do not have incentives to control their costs (i.e., cost price, salaries), as the city will ultimately cover their deficits. According to Provan and Milward, stronger central control should favor cooperation as network members have more incentives to cooperate among themselves and with the authorities (Provan & Milward, 2001).
In addition, the absence of a comparative perspective does not incite associations to improve cost control; as a result, cost structures of associations vary significantly. A comparison of key financial and qualitative performance indicators between the di erent associations could incite them to collaborate and provide the municipality with relevant information for decision-making. The network e ectiveness argument of Provan and Milward is confirmed by our findings (Provan & Milward, 2001). With the current budget restrictions, the network e ectiveness may come at a cost that may be too high to sustain involvement of all network members as argued by Provan and Milward. Finally, cash collection and bad debts are an issue for most associations. Associations deliver the service and issue invoices to families, but they do not have any means of enforcement to collect the bills. The municipality authorities are ultimately in charge of monitoring bad debts and litigations, while they are not in direct relationship with the client. This renders the administrative organization of the network more complicated and less e ective (Milward & Provan, 2003;Provan & Milward, 2001).
Finally, fraud risk has to be considered. One case was detected in one association that resulted in a change of committee and suppliers. An audit was performed in order to evaluate the damages and risks. This event contributed to the municipality requiring audit reports on financial statements for 2011-2012. These audits represent additional requirements to improve accountability, but they will generate supplementary annual costs for associations and ultimately for the municipality.

Accountability
Better accountability can be achieved by assessing the several risks associated with the service produced, thus having a consensus of all stakeholders on the di erent responsibilities and risk ownership in the network. Indeed, the perceived lack of accountability in public o cials is one of the important reasons behind the desire of the Swiss municipality under study to change its governance model. The service quality in the co-production of public services is considered as a key element of accountability (Bovaird, 2007), and could harm the municipality's legitimacy but not the associations. This contradicts the position according to which an association is considered to improve its legitimacy if it is proven accountable (Mason, 2010). Mulgan (2000) captures this evolution, stating that accountability is perceived as a political value which places more trust in the independent judgment of public servants.
The Swiss municipality is accountable for the quality of the school restaurants services, but at the same time, has limited oversight on the associations and their personnel's activities. The municipality authorities provide the subsidies and require audited financial statements as a means of financial control. The heads of finance and operations also follow an estimation of the meal cost price. They also control the menus established by the kitchen heads as they have to comply with the label "green fork" a healthy balanced diet. Hygiene norms for food safety are regularly controlled. Finally, the yearly regular growth in the demand and the results of a satisfaction study among parents constitute performance indicators that municipality authorities follow closely.
Legitimacy is also perceived as a means of obtaining and maintaining resources (Oliver, 1991) when the organization satisfies the key stakeholder's expectations, in particular the clients' expectations for Milward and Provan (2003). Accordingly, it is important for the associations participating in the network model to prove accountable to the key risks assessed and discussed or to transfer these risks to other partners, such as the municipality, another association, or a private organization. For instance, in our case, the municipality has the overall responsibility in cases of sanitary and quality issues when it comes to meal preparation, without being able to enforce the associations' observations of the appropriate measures in respecting dietetic or other quality standards.
Furthermore, the indicators have grown rapidly over the past years when it comes to co-management or co-production. There is increasing pressure for more integration and more specialization in order to attain e ciency and e ectiveness, creating a conflict between the e ciency goals and social goals of an organization and the accountability mechanisms that are defined accordingly (Brandsen & Hout, 2006). This is indeed perceived as a great pressure, and several of the associations that produce the meals changed into more market-based, service-delivering organizations by recruiting professionals both in the producing and managing functions of school restaurants. One could even argue that this competition may discourage collaboration and hinders performance at the network's level (Milward & Provan, 2003).
Nevertheless, attention should be paid to the overly pervasive faith that is placed in market-and business-based solutions, as several authors criticize the pro-business ideology that has followed the decline of the welfare state ideology (Kuttner, 1999;Zimmerman & Dart, 1998). It is then important to focus on the social benefits and the core purpose of the associations and beware of the shift toward managing a "business enterprise" (an evolution that is also captured by several authors, for example, Young (2002) and Dart (2004). Social capital creation is also one of the valuable network's e ectiveness criteria (Provan & Milward, 2001). In our case, social capital creation and maintenance depends on the associative life and the relationships with the di erent stakeholders to produce and deliver a service corresponding to parents' expectations.

Conclusions
Within this context of evolution from governance arrangements characterized by centralization and hierarchy toward more interactive and participative decision-making and services co-creation, this article has examined the evolution of a Swiss public governance network of school restaurants that goes into the opposite direction. This network has been run with a constellation of "agents of the state", mostly non-profit providers that conduct the primary work of this public service provision (Considine, 2003;Heinrich et al., 2009). Due to increasing demands, the growing number of children in schools, and societal changes in today's modern societies, the system has come under pressure that has highlighted several problems within the system. We investigated the reasons for the launch by the municipality of a centralization process of school restaurants in a Swiss city, moving from this completely decentralized network governance model that involved associations among other stakeholders in the delivery of public services to a more procedural and centralized network governance model.
We found that pressures and challenges of risks and accountability faced by the associations, and ultimately the municipality, were the main drivers for change.
Our findings highlighted the fact that communication of financial and other information is very poor in the system. This lack of communication and coordination generates supplementary costs in a system that is evolving toward a complicated and more heterogeneous form, with di erent actors each using di erent working methods.
Furthermore, there is no clear designation in the network for who should be in charge of optimizing the procedures and solving the apparent problems, leaving the overall responsibility of the system to the municipality authorities. We argue that a risk approach can be used to better communicate the responsibilities and concerns in terms of e ciency and e ectiveness in delivering a public service in within the network by clearly identifying the risk owners, their responsibilities, their resources, and their means of control. A clarification of roles and responsibilities between the authorities and the associations would also reduce transaction costs in the system.
As accountability is an important issue, it is therefore essential for the associations participating in the network model to prove accountable to the key risks assessed and discussed or to transfer these risks to other partners, such as the municipality. Several meeting points should be put in place, and the existing management committee should define a clear agenda and have more responsibilities in terms of coordination of the whole system in order to have a "constant nurturing" of the network, as advocated by Klijn (2008).
This approach should reassess the roles of the actors in the system with di erent configurations and organizations of tasks and activities to be completed. The network governance model can then be maintained with the concentration of certain low social purpose functions that require specific professional skills such as purchasing, production, billing, and human resources management in order to release the associations from the pressure of e ciency. These changes are expected to allow the associations to concentrate on their social purposes and provide the municipality authorities with more control over the system. While centralizing the highly risk-sensitive and costly activities under the umbrella of the municipality authorities, this network governance model would preserve the associative life and the relationships with the di erent stakeholders (the social capital) to produce and deliver a service corresponding to parents' expectations. The system of financing the associations would then be adapted to the level of activities maintained within the associations.
This study contributes to the public governance literature because it describes and explores the causes of the evolution of a network governance arrangement toward more centralization while many scholars of public-sector governance call for more participative and interactive organization and co-production. The literature about governance transformation highlights the evolution of governance models toward network governance perceived as the primacy model. We observed a reverse movement toward more surveillance and accountability after having reached the limits of the completely decentralized network governance model.
Finally, we stress the limited nature of our study, whose purpose was to serve as an introductory exploration of the factors that influence changes in public governance of school restaurants. Our findings cannot be extrapolated to all Swiss municipalities, but they allow us to devise a stricter and more rigorous methodology for further studies based on the combinations of interviews, questionnaires, or quantitative surveys.