Why do policies change? Institutions, interests, ideas and networks in three cases of policy reform

Abstract

Policy researchers have used various categories of variables to explain why policies change, including those related to institutions, interests and ideas. Recent research has paid growing attention to the role of policy networks—the actors involved in policy-making, their relationships with each other, and the structure formed by those relationships—in policy reform across settings and issues; however, this literature has largely ignored the theoretical integration of networks with other policy theories, including the ‘3Is’ of institutions, interests and ideas. This article proposes a conceptual framework integrating these variables and tests it on three cases of policy change in Burkina Faso, addressing the need for theoretical integration with networks as well as the broader aim of theory-driven health policy analysis research in low- and middle-income countries. We use historical process tracing, a type of comparative case study, to interpret and compare documents and in-depth interview data within and between cases. We found that while network changes were indeed associated with policy reform, this relationship was mediated by one or more of institutions, interests and ideas. In a context of high donor dependency, new donor rules affected the composition and structure of actors in the networks, which enabled the entry and dissemination of new ideas and shifts in the overall balance of interest power ultimately leading to policy change. The case of strategic networking occurred in only one case, by civil society actors, suggesting that network change is rarely the spark that initiates the process towards policy change. This analysis highlights the important role of changes in institutions and ideas to drive policymaking, but hints that network change is a necessary intermediate step in these processes.

Key Messages

• This article integrates theories of policy change based on institutions, interests, ideas and networks.

• Change in policy networks are often associated with policy change, but rarely change on their own. Instead, they follow changes in institutions, interests and ideas.

• Health donors (interests) have power to change institutions—and thus networks—leading to policy change.

• Civil society actors were able to instrumentally use their networks in one case to eventually achieve policy change.

Introduction

Understanding the drivers of policy change is a pursuit that has captured the imaginations of researchers and practitioners alike ( Walt 1994 ; Walt and Gilson 1994 ; Gilson and Raphaely 2008) . While theory-driven health policy analysis continues to grow in low- and middle-income countries (LMIC), the complexity of the contexts and issues studied calls for greater integration of multiple policy theories for a given case of policy change ( Agyepong and Adjei 2008 ; Smith 2014 ; Walt and Gilson 2014) . Our field does not suffer from the same disagreements that prevent theoretical integration in the industrialized world—namely, the stalemate between behaviouralist and stucturalist paradigms ( Skocpol 1985) .

In this article, we hope to shed light on the relative influence and temporal ordering of various factors from a range of theoretical perspectives to understand why policies change. Existing policy change frameworks can be distilled into three key elements, or explanatory variables: institutions (processes, context); interests (actors, power) and ideas (content, evidence, values), known as the ‘3Is’ ( Walt 1994 ; Palier and Surel 2005) . More recently, a new explanatory variable has entered the fray. The empirical study of policy networks–defined as policy actors and the relationships between them—reflects the state of contemporary policy-making ( Jones et al. 1997 ; Gruening 2001) characterized by diverse sets of actors, fluidity of issues and institutions and power distributed in networks instead of hierarchies ( Gilson et al. 2003 ; Hyden 2006 ; Bratton 2007 ; Woelk et al. 2009 ). Indeed, much of the policy network literature from high-income countries has focused on the impact of new actors and their ideas on policy-making, and how network structure influences policy outcomes ( Coleman and Skogstad 1990 ; Atkinson and Coleman 1992 ; Thatcher 1998 ; Marsh and Smith 2000 ; Kriesi et al. 2006 ; Sandstrom and Carlsson 2008) . Related concepts include policy communities ( Coleman and Skogstad 1990) , advocacy coalitions ( Sabatier and Weible 2007) and issue networks ( Heclo 1978) and draw on organizational, policy and management sciences.

Network theory and tools have been applied to LMIC health policy and systems research ( Wonodi et al. 2012 ; Blanchet and James 2013) , but as is the case in high-income countries ( McClurg and Young 2011 ; Lubell et al. 2012 ) these studies have largely considered networks in a theoretical vacuum. This article aims to develop and test a conceptual framework integrating theories of institutions, interests and ideas, as well as networks to explain policy change for three cases in Burkina Faso.

A 3I + N conceptual framework of policy change

The ‘3I’ framework offers a comprehensive set of variables to consider in the analysis of policy change ( Lavis et al. 2002 ; Lavis c2004 ; Palier and Surel 2005 ; Waddell et al. 2005 ). We reviewed the literature on the 3Is as well as networks in order to propose an integrated conceptual framework of these four potential independent variables. We hypothesize that these variables may interact to shape policy change (see Figure 1 ); the rarer case in the complex system of policymaking is that one or more of the variables act independently to shape policy change.

Figure 1.

Conceptual framework of 3I and network variables on policy change

Networks are defined in the conceptual framework as both empirically measurable sets of actors and their relationships (i.e. the measured sets of individuals who participated in human immunodeficiency virus (HIV) policy-making, and how they were connected vis-à-vis working relationships), and as intentional governance or management structures with agency to act strategically (i.e. an HIV issue network with defined members and goals). A key problematic of this study is whether networks—either empirical or intentional—exert independent influence on policy change, or whether they are mere bystanders, mediators or moderators, of the causal relationship between any of the Is and policy change. As in other studies of complex phenomena, this framework aims to shed some light on the mechanisms underlying observed phenomena.

Institutions are the ‘rules of the game’ ( North 1990) that structure policymaking in ways that favour some outcomes over others. Institutions, which include policies themselves, shape policy change primarily through the ways in which they create and distribute incentives and learning ( Pierson 1993 ; Hall and Taylor 1996 ; Ostrom 2005 ). While a central premise of political institutionalism is the rarity of policy change, the presence of change is explained primarily by external events that alter institutional rules enough to provide a window for public action. Institutions may be formal or informal, but informal institutions are particularly pertinent in the study of LMIC policy change ( Helmke and Levitsky 2004 ; Hyden 2006 ; Bratton 2007) .

Although networks are sometimes considered a form of institution, we consider them a variable in their own right. Networks, like institutions, impose structural constraints on policymaking by mediating the pattern of relations among actors ( Marsh and Smith 2000 ; Sandstrom and Carlsson 2008 ), but are far more fluid than institutions. Networks are likely to change in response to institutional pressures: ‘a change in institutional rules directly affects network structure by creating new opportunities and incentives for policy interactions’ (2012, p. 355). Conversely, networks can create, reinforce or challenge institutions by facilitating interactions among actors in ways that might lead to shifts in norms, preferences and power, a possibility which is most likely in contexts of highly informal or weak institutions ( Hall and Taylor 1996 ; Helmke and Levitsky 2004) .

Interests describe the preferences and power embedded in policy actors (see Figure 1 ). Behavioural schools of policy change assume policy outcomes are largely determined by actor interests and behaviours ( Hall and Taylor 1996) . The ability of actors to attain and exercise their interests depends on the distribution of resources and power in a policy domain, as well as individual capacity and skills. Further, most interest-based theories acknowledge the structural constraints on individual agency. Pertinent, but not exclusive to LMIC settings, scholars describe a growing authority of private and non-state actors in government policymaking processes ( Mathews 1997 ; Litvack et al. 1998 ; Buse and Walt 2002) adding to the power of international organizations ( Kahler and Lake 2004 ; Dobbin et al. 2007 ). The distribution of power is largely driven by access to resources in these settings but has an equally important normative element ( Shiffman 2014) .

Networks can help visualize how interests, embedded in nodes, are structured in the policy process and how network structure changes as actors form and dissolve relationships. Like institutions, networks can influence the balance of power by choosing to include or exclude participants ( Marsh and Smith 2000) . In the other direction, actors embedded in nodes have agency to decide who they will interact with, conscious and strategic decisions that can shape networks to advance certain interests ( Marsh and Smith 2000 ; Howlett 2002) and control information exchange (Shearer 2014).

The concept of ‘ideas’ in policy sciences is broad and relates to the content and strength of actors’ values and knowledge in the policy process ( Hall 1993 ; Surel 2000 ; Lavis et al. 2004 ). Ideas shape agenda-setting, policy formulation and implementation by determining which representations of the problem and potential solutions will be heard and understood by policy-makers ( Hall 1993 ; Surel 2000 ; Sabatier and Weible 2007) . The importance of technical information and knowledge as a variable has risen as policy-making has become increasingly driven by expertise in some settings ( Radaelli 1999 ; Beland 2010) ; the use of research evidence to inform health policy has achieved international normative status in the past decade ( World Health Organization 2005) .

Networks play an important role in the creation, dissemination and reinforcement of ideas ( Owen-Smith and Powell 2008) . Ideas are exchanged over network ties and thus the pattern and structure of ties affects the rate and reach of information dissemination ( Reagans and McEvily 2003) . Networks can reinforce or shift paradigms and values ( Sabatier and Weible 2007) , and some network structures are more exposed to new ideas than others. There is an immediate feedback effect in that the entrance of new ideas may be highly disruptive to networks. Values, preferences, experiences and knowledge determine actors’ propensity to form ties. The clustering of actors around certain ideas may thus influence network shape.

Figure 1 suggests these variables are interdependent; specifically, that institutions provide the scaffolding for network structure, interests are embedded in actor nodes and ideas are exchanged along ties between actors. Changes in any one factor can be initiated by endogenous or exogenous factors and will likely spill over to other factors, involving feedback loops. Identifying the spark and its consequences requires detailed knowledge of the issue, context and history.

Methodology

This is a comparative case study of three policy processes in Burkina Faso, using historical process-tracing and within-case analysis methods to draw inferences about the role of each of the variables on policy change ( Gerring 2004 ; Yin 2009 ; Collier 2011) . We take a qualitative approach to describing changes in the variables, including network change, which is defined as any major shifts in the composition of actors or how they are linked to each other. Case study and process tracing techniques are well-suited to the in-depth investigation of complex phenomenon such as policy-making. This study is concerned primarily with the policy formulation phase of the policy process, although the necessary historical perspective cannot ignore agenda-setting where it has an effect on the process of and prospects for policy change.

Case selection

We define the study unit as Burkina Faso and the embedded cases as policy processes, defined as series of events leading to a government statement of intent to act on a policy issue, and for which there are clear plans to implement their decision. This definition was created in order to enable historical process tracing ( Collier 2011) and to orient interview respondents to a specific outcome of those processes. Policy cases were selected in part for pragmatic considerations, including the availability of relevant documents on the cases and their projected network sizes adequate to enable statistical analyses for linked studies. Cases were also selected according to their diversity on independent variables of interest, namely hypothesized network structure, in order to explore and confirm theoretical propositions about the factors that lead to policy change ( Seawright and Gerring 2008) . Three policy cases were chosen: community integrated management of childhood illnesses, which we imagined to be a more closed network; home management of malaria, which we knew to involve a wider range of stakeholders due to Global Fund rules; and the removal of user fees for antiretroviral treatment, which we understood to be a long process involving a range of diverse stakeholders with competing perspectives (see Table 1 for a full background of each case). Cases will be referred to by their substantive themes: ‘child health’; ‘malaria’ and ‘HIV’.

Table 1.

The cases and how the 3I and network variables led to policy change

	Child health	Malaria	HIV
Health issue(s) addressed	Child deaths at home or in the community	Malaria deaths in the home or in the community	Financial access to antiretroviral treatment for people living with HIV
Proposed policy change	Train CHWs to manage childhood pneumonia, malaria, diarrhoea and malnutrition in the community.	Train CHWs to manage malaria in the community using ACTs.	Remove user fees for antiretroviral treatment, rendering it free to patients.
Status quo at start of policy development process	At a national scale, only health workers in health facilities were trained to treat sick children.A small number of non-government organizations had programmes treating malaria, diarrhoea and malnutrition by CHWs, as well as CHWs providing health education.	Previous malaria home management programme cancelled due to chloroquine resistance; new drugs only available from health facilities or private pharmacies.	National policy required patients to pay for treatment: patient cost decreased from US$27 in 2002 to $3 in 2008 where 45% of population lives below poverty line of $1.25 PPP per day ( United Nations Development Programme 2013 ). In practice, few providers charged user fees.
Policy decision (date) and reference documents	Introduce community case management in two regions, and pilot the pneumonia component in two districts (2008) as part of a grant proposal.Written into policy in ‘Strategic Plan for the Community Integrated Management of Childhood Illness’ (2010)	Introduce home management of malaria nation-wide, implemented by National Malaria Control Programme and non-government organizations (2008).Written into policy in ‘Strategic Plan for the Fight Against Malaria: 2006–10’ (2007); funded through ‘Burkina Faso’s Proposal to Global Fund Round 7’ (2007) and ‘Burkina Faso’s Proposal to Global Fund Round 8’ (2008)	Removal of user fees for antiretroviral treatment (2009).Announced by President in public and in press release: ‘Burkina Opts for Removal of User Fees’ (2010)
Institutions and networks	Donor rules encouraged the entry of new actors and their ideas, reshaping the policy network.New actors and ideas were essential in initiating and pushing through policy change.	Donor rules encouraged the entry of new actors and their ideas, reshaping the policy network.Policy legacies of malaria home management favoured its reinstatement.New actors, as well as favourable policy legacies, enabled the adoption of the policy proposal.	Policy legacies of civil society participation in treatment provision created organizational niches, which reinforced their cohesive network.Cohesive civil society network encouraged adoption and diffusion of informal institutions, which influenced formal policy change.
Interests and networks	Introduction of new interests, embedded in new actors, changed network structure and shifted balance of interest power. New actors directly advocated for policy change.New balance of power favoured policy change.	Introduction of new interests, embedded in new actors, changed network structure and shifted balance of interest power.New balance of power favoured policy process and policy instruments that implicated civil society.	Strength of civil society network increased its power and influence.Change in leadership created new opportunities and incentives for ties between actors, changing network structure.New balance of power favoured policy change.
Ideas and networks	New evidence demonstrated failure of facility-based paradigm and successful community case management experiences. Its exchange created new ties, altering network structure.The spread of new ideas moved the policy process from agenda-setting to formulation, resulting in policy change.	New actors had experience and knowledge necessary to implement policy proposal.	Evidence demonstrated failure of existing paradigm. Its exchange created new ties, altering network structure.The spread of new evidence influenced policy change.
External events	External funding opportunity	External funding opportunity	Internal funding opportunity

Data collection and analysis

Data were collected through document review and interviews during May 2011 to March 2012. The primary researcher searched for published research evidence, reports, policy documents, news media, meeting minutes and presentations in local and international databases in order to create a timeline of events and to inform the interview guide.

In-depth semi-structured interviews were carried out in-person in the capital city of Burkina Faso, with some interviews done in other locales. Initial respondents were identified through the document review followed by respondent-driven sampling. Interviews were based on a pre-established question guide which was pilot tested before finalizing, and most interviews began with the question, ‘Tell me the story of this policy issue’, leading to narrative histories of the events leading to policy change. Interviews were conducted in French, audio recorded, and notes taken. Interviews lasted 45 min on average. Audio recordings were transcribed in French and then coded in English by the bilingual lead researcher according to a pre-established codebook based on the 3I conceptual framework with additional codes related to networks. Cases were analysed for within- and across-case variation, as well as for variation over time ( Gerring 2004) . Efforts were made during analysis to identify emergent themes as well as negative data. NVivo 10 software was used to manage and code interview data ( QSR International 2012 ).

Ethical approval was received from the authors’ institute [McMaster University] and the Burkina Faso Ministry of Health’s National Health Research Ethics Council (Council National d’Ethique de la recherche en santé). Signed consent was requested from respondents prior to beginning interviews.

Results

Table 1 describes the background and context for each case and the high-level findings for each of the 3I and networks. Figure 2 depicts a timeline of events. The following data are based on review of nearly 100 documents and 81 interviews amongst 62 policy actors (many respondents yielded interviews on multiple cases; see Table 2 ). Figure 3a–c incorporate network graphs for each case as measured in linked studies (AUTHOR).

Figure 2.

Timeline of events

Table 2.

Interview respondents.

	Child health	Malaria	HIV
Interview respondents	n = 20	n = 33	n = 24
	Government: 14	Government: 19	Government: 9
	Civil society/non-governmental	Civil society/non-governmental	Civil society/non-governmental
	Organizations: 0	Organizations: 10	Organizations: 11
	Development partners: 5	Development partners: 2	Development partners: 2
	Other: 1	Other: 2	Other: 2

Figure 3.

(a–c) illustrate main findings on network graphs. These graphs represent cross-sections of each network (measured empirically during a linked sub-study; Shearer 2014. These figures are meant to be illustrative; locations of each finding, for example, are not related to specific nodes or relationships in the network.

Community integrated management of childhood illnesses (IMCI)

Community health policy in Burkina Faso originated in 1985 when president Thomas Sankara announced his vision for ‘one village, one primary health post’ ( Seck and Valea 2011) . This policy led to the selection and training of community health workers (CHW) chosen by their communities across Burkina Faso. While this policy was essentially abandoned after the president’s death in 1987, it created a cadre of CHWs who were later used by other government and non-government programmes. Their existence led communities to expect them, termed ‘lock-in effects’ in the political science literature ( Pierson 1993) , and to policy learning amongst policy-makers (institutions).

In 1999, Burkina Faso adopted the World Health Organization’s IMCI policy to train facility-based health workers to correctly identify and treat childhood malaria, pneumonia and diarrhoea ( Direction de la santé de la famille 2005 ). Although a component of IMCI intended to train community workers to provide education and referral activities, it was never fully developed or implemented. Instead, IMCI reinforced the dominant facility-based paradigm (ideas) of the 1990s and the administrative capacities (institutions) of the health system to plan and deliver facility-based care. Nurses benefitted from assured salaries and viewed their new roles as entrenched in the health system (institutions); they later opposed community-based approaches for fear of losing power (Government, 1). The office created to manage IMCI in the Division of Family Health, staffed by former clinicians, further entrenched the facility-based paradigm to the detriment of a community-based approach (institutions and ideas). Funding and technical assistance from WHO, who was also very profession-based, reinforced what seemed to be a small but cohesive network of policy elites (networks and ideas). The network’s cohesion and actors’ shared paradigms were mutually reinforcing and the institutional context allowed it to remain largely closed to new actors or ideas (institutions, ideas and networks). Its prospects for policy change were limited.

Consistent with theories of why policies change ( Hall 1993 ), the paradigm shifted towards community approaches only following observed weaknesses in facility-based results (ideas). An evaluation of the country’s national health plan showed slow progress towards—MDG-4 and survey data ( UNICEF and INSD 2006) reinforced the tacit knowledge that large numbers of children continued to die at home (Government official). Despite the scope of resources allocated to IMCI, programme coverage and utilization were low ( Ministère de la santé 2007 ), leading to a growing sense in the health bureaucracy that coverage and access must be addressed.

But we recognized that since the implementation of these policies, we had not attained the expected results. It is because of this that the rapid gain interventions were identified… all those are the activities that we included as part of rapid gain activities in order to meet the MDGs. But, we realized that with these implementation strategies, we had not succeeded in attaining the Goals… We realized that we must use other strategies for reinforcing the role of social mobilization. Developing community-based services. (Government, 2)

The joining of these factors allowed community-based management of childhood illnesses to emerge as a policy option. Policy actors at the global level were having similar conversations, leading to the adoption of integrated community case management (iCCM) as a policy solution by global epistemic communities (ideas, interests and networks) (Dalglish 2015).

Meanwhile, state- and non-state actors were experimenting with the community-level management of major childhood illnesses, including the treatment of pneumonia with antibiotics administered by CHWs. A 2005 visit by a Senegalese team attempted to persuade the government that iCCM was feasible and effective, but policy elites were uninterested and ‘not ready’ for ‘old and illiterate’ CHWs to administer antibiotics (Government, 3). This, combined with the dominant facility-based paradigm (ideas) and a health law forbidding the use of antibiotics by lay persons ( Burkina Faso 1994 ) made policy reform seem unlikely.

A policy window opened in 2008 when the Partnership for Maternal, Newborn and Child Health (PMNCH) along with the Bill and Melinda Gates Foundation (Gates Foundation) presented the Government of Burkina Faso with the opportunity to apply for US$6 million in funding over 3 years to accelerate progress towards MDG-4. Funding rules (institutions) required that the proposed interventions must reduce under-five mortality by 25% during the project period; it was later determined by researchers hired by the funders that a community-based package including pneumonia, diarrhoea and malaria management; insecticide treated nets; and Vitamin A supplementation would meet this goal ( Bryce et al. 2010 ). Donor rules (institutions) opened the policy network to new actors by mandating the co-leadership of the proposal process by UNICEF and the Division of Family Health, as well as the full participation of other multilateral agencies, the participation of evaluation teams from a local research institution and an American school of public health, and finally, input from the funders (Researcher, 4). The inclusion of these actors changed the structure of the network ( Figure 3a , Finding 1). Respondents noted that many health bureaucrats and policy elites in this network opposed the inclusion of pneumonia, in line with the existing facility-based paradigm (ideas and interests), but UNICEF’s policy preferences aligned with PMNCH and the Gates Foundation ( Figure 3a , Finding 2). The debate over pneumonia became the focus of the process with efforts by UNICEF to overcome opposition using persuasion and the dissemination of research evidence from other countries which led to the formation of new knowledge exchange ties in the network (networks and ideas; Figure 3a , Finding 3). To this end, numerous respondents cited the normative influence of the 2003 Lancet series on child survival during meetings ( Black et al. 2003 ); peculiarly, the series did not actually report the effectiveness of iCCM-based approaches ( Bryce et al. 2003 ). It is not clear to what extent UNICEF used the publication strategically to support their pre-determined policy position but it certainly attained a symbolic status during the policy development process.

Nevertheless, the original submission did not include the community management of pneumonia, but this national autonomy was short-lived—the funders exercised their veto power (institutions) in their comments on the submission, stating: ‘Community IMCI must focus on the community management of pneumonia’ ( Direction de la santé de la famille 2008 ). Around this time, in what many respondents considered to be one of the most important events for community case management in Burkina Faso, UNICEF supported a government health official to attend a 2008 meeting in Madagascar, where 20 countries shared community-case management experiences ( Figure 3a , Finding 3). Respondents described this meeting as part of a specific advocacy strategy that strategically invited certain Ministry of Health (MOH) staff ( USAID 2009 ). Burkina’s attendee was convinced by what he saw and returned to share his experiences with colleagues in support of fully iCCM, including pneumonia, in the funding proposal (ideas).

Yes, me for example, I presented, every time I presented the experiences I had seen in Madagascar. With… the experiences of Senegal, of Malawi, of Rwanda… All those countries. In any case I made these presentations and that helped people, to convince people that if we do it with agents well motivated, of a certain level, it can help manage, to decrease mortality. (Government, 5)

Exposure to a larger supra-national network seemed particularly important in this case. A change in the information capital of one network actor led to the formation of new ties, thus altering overall network structure to shift the balance of power and ultimately facilitating policy change (networks and ideas; Figure 3a , Finding 4). The government’s next submission in the same year included a pneumonia management pilot in two health districts and the existing drug prescribing laws were bypassed by the introduction of an MOH strategic plan for community management of childhood illnesses ( Direction de la santé de la famille 2010 ). In this case, a change in institutions seems to have brought about changes in the composition and structure of the network, as well as in interests and ideas, ultimately leading to policy change.

Home management of malaria

Home management of malaria was first introduced in Burkina Faso in 1997, using existing CHWs to dispense chloroquine and paracetemol for fever. Although this programme faced difficulties in scaling-up due to lack of funding ( Programme national de lutte contre le paludisme 2004 ), it was generally perceived to be successful in expanding malaria treatment in districts where it was implemented (institutions and ideas) ( Kouyate et al. 2007 ). The original programme created a number of policy legacies (institutions) that favoured its reinstatement in later years. It created administrative capacities within the health bureaucracy and increased the influence of the National Malaria Control Programme, and positive experiences with the original programme caused many health bureaucrats to support its future iterations (Government, 6). The initial reliance on CHWs, who remained central to the programme, further cemented their role in the community, thus strengthening population lock-in effects (institutions).

Home management of malaria was an important component of the country’s overall malaria strategy and was supported and encouraged by WHO, Roll Back Malaria and other development partners (interests). It also existed within a supportive research environment where studies frequently demonstrated its effectiveness, feasibility and acceptability (ideas) ( Pagnoni et al. 1997 ; Sirima et al. 2003 ; Kouyate et al. 2007 ; Tiono et al. 2008 ), generating policy learning among government elites. All of these factors reinforced a political economy of malaria research and aid that favoured the expansion of the home management programme (institutions, interests, ideas) and which was supported by policy elites. These factors worked together to entrench a small, cohesive and self-reinforcing network of largely technical actors in the National Malaria Control Programme and national research centres (network).

However, the original programme was cancelled in 2005 based on evidence of chloroquine resistance and new global guidelines for the use of artemisinin combination therapy (ACT) as a first-line therapy for uncomplicated malaria ( Programme national de lutte contre le paludisme 2006) . Concerns over ACT cost and supply kept home management of malaria off the policy agenda for a number of years—the government was barely able to supply ACTs to health facilities ( Tipke et al. 2009 ). The emergence of chloroquine resistance created fears among policy-makers that the same thing could happen to ACTs, considered to be the last line of defence against malaria, if they were mismanaged by CHWs (ideas); yet, informal drug use was simultaneously becoming a growing problem as the original home management programme had created treatment expectations within communities (institutions). Without ACTs available at the community level, patients continued to seek chloroquine for self-treatment of malaria through private pharmacies and counterfeit drug vendors ( Ouedraogo et al. 2008 ; Tipke et al. 2009 ). The ongoing use of an ineffective drug posed a public health and policy problem ( Kouyate and Nana 2010) .

Even if we refuse the treatment in the community the people will do it. Because they are going to buy products and medicines from the street or elsewhere. Better to formalize this community treatment than leave people to their choice of drug, which at this time was very dangerous. (International organization, 7)

As with the child health case, a policy window for reform was opened by external funders. Around 2006, home management of malaria became a strategic focus of the Global Fund to Fight AIDS, Tuberculosis and Malaria (Global Fund) and the Government applied in 2007 for funding to pilot home management with ACTs in two districts. This decision was informed by the growing body of studies from Burkina Faso showing the feasibility and effectiveness of CHWs to use ACTs in communities (ideas) ( Tiono et al. 2008 ; Sirima et al. 2009 ), which were widely exchanged within the small, technical malaria policy network. The application for pilot funding was accepted, but never fully implemented. Burkina Faso learned in 2008 that the Global Fund was willing to make enough funding available to scale-up the programme nation-wide; 2008 was also the first funding round for which the Global Fund changed the rules of the game involving civil society organizations—mandating their near equal participation as co-recipients of the grant (institutions; Figure 3B , Finding 1). Nearly all respondents remarked that without this condition, civil society would not have been implicated to the extent that they were, and the national grant may not have been possible.

When they sent us the directions for Round 8, they encouraged us to involve the community sector and that would count as a ‘plus’ in the proposal, it’s true. If it had not been said like that, maybe the community sector would not have been involved at this level… (Non-government organization, 8)

Decisions that had once been made by the relatively small but powerful National Malaria Control Programme were now party to influence from a much larger network of actors (networks, ideas, interests). Civil society beneficiaries were selected and participated actively, subsequently becoming a powerful voice in the process. In any case, the expanded network of civil society actors enabled the government to move forward with a decision to apply for national scale-up of home management of malaria, led by the fact that these partners had the capacity to reach populations throughout the country that the public service could not ( Figure 3b , Finding 2). In contrast to the iCCM case, the funder’s mandate of who participated was less about influencing policy change, per se , as it was about influencing the implementation feasibility and financial risk of their investment.

The explicit application of research evidence took a back seat during the Round 8 process but continued to inform many influential members of the network (ideas). The opening of the network to civil society actors, who had less of a technical focus than their government counterparts, reduced the exchange of research evidence; overall, few civil society respondents reported awareness of existing, local studies on home management of malaria. As well, the incentive to use research evidence to directly inform policy (i.e. the results of the pilot project) was reduced upon the offer of funding through Round 8. The sole use of research evidence appeared to be by external consultants who were hired to write the final proposal, as the funders required it include a certain amount of data and evidence.

Removal of user fees for HIV treatment

Healthcare user fees in Burkina Faso can be traced to the Bamako Initiative, a WHO/UNICEF initiative to improve access to primary health care through decentralization and cost-recovery ( Ridde 2003) . Ratified in 1987 and launched in 1992 in Burkina Faso ( Seck and Valea 2011) , the Bamako Initiative (institutions) has left a legacy of user fees across Africa and has had particularly strong effects in West Africa. Indeed, the normative effects of the Bamako Initiative may have posed the most significant barrier to user fee removal (ideas). Its legacy was apparent in national policy documents which presented the Bamako Initiative as an overarching orientation for the health system ( Ministère de la santé 2000 , 2007 , 2011 ). Interviews with government elites demonstrated their support for the Bamako Initiative and associated ideologies of individual responsibility. Thus, many of these elites, including the Minister of Health prior to 2008, were opposed to the removal of user fees (ideas, interests). In the HIV/AIDS decision-making structure, the Minister of Health held veto power, as did high-level government staff in the National AIDS Council, and finally, the President of Burkina Faso (institutions). Thus, until a change in ministers in 2008, technical arguments from the MOH could not progress up the decision-making chain.

Together with the coordinator of [Ministry of Health HIV/AIDS office], we had prepared during this time a document to explain why we were able to move towards the removal of user fees but the political authority had other motivations than the advice of technicians. (Government, 9)

Civil society actors in Burkina Faso’s HIV sector held the opposite philosophical perspective, and thus policy preferences, driven by an ideology of social justice and universal access to treatment (ideas).Their voice was ultimately heard due to multiple policy legacies entrenching their influence on the debate (institutions). First, civil society emerged to occupy an organizational niche in HIV service provision beginning with the government’s ineffective response to HIV treatment in the early 2000s ( Peschi 2004) . Recognizing its limitations, the government allowed and encouraged these developments, resulting in a quasi-formalized network of non-governmental treatment providers with growing organizational, programmatic and normative influence (institutions, networks, ideas) and leading to bifurcation between the civil society and government sides of this policy network. In 2003, civil society organizations formalized their organic network into a coalition with the explicit purpose of advocating for the removal of antiretroviral treatment user fees (interests, networks). This is the only case where respondents explicitly referred to their ‘network’ in the formalized sense of the definition. Respondents described this coalition as well-organized, sharing a common vision and highly active in policy discussions. This power of the coalition in policy debates was reinforced by the second major policy legacy, the rules of the National AIDS Council which required representation of CSOs. Civil society actors benefited from dense national networks, as well as connections to regional and international actors, and thus information (networks, ideas; Figure 3C , Finding 1). Their capacity to work cohesively improved during 2008 when they were required to participate in the HIV/AIDS funding application to the Global Fund, and they shared the sentiment that, ‘Pretty much all the civil society leaders, we were in collaboration during this period of the issue’. (Civil society organization, 10)

The way in which civil society actors had actively networked themselves reinforced their power in the debate, and marks the only instance of strategic network management, or endogenous network change, across these cases. As a unified voice with strong connections throughout the country, at all levels of advocacy and service delivery, their network strengthened and legitimated their treatment paradigm. Meanwhile, when confronted with patients who could not afford treatment, but who were not poor enough to be considered ‘indigent’, service providers who were members of the national network began ignoring the official user fee policy (institutions). The network structure enabled the practice to diffuse ( Figure 3c , Finding 2) and by 2006, a survey showed that 79% of patients attending HIV clinics did not pay for antiretroviral treatment received ( Kouanda et al. 2010 ). Ultimately, most respondents agreed that the growing evidence on the failure of cost recovery had acted to break down the dominant paradigm supporting user fees (ideas) and to build bridges between civil society and supportive government staff (ideas and networks).

A change in the minister of health in 2008 opened a window for a wider debate on user fees ( Figure 3c , Finding 3). This initiated the technical work necessary for the decision, including an important technical note prepared by the HIV/AIDS office in the MOH which demonstrated that at most, 2% of total costs could be recovered in the current system ( CMLS-Santé 2009 ). Growing availability of data to demonstrate the failure of the policy forced the government to re-consider its position (ideas). The exchange of ideas within and across civil society and government sub-networks resulted changed the structure and balance of power in the network ( Figure 3c , Finding 4).

Also in 2008, the government introduced a new social safety net account (‘filets sociaux’) in response to the 2008 global economic crisis, allowing new government and donor funding to target specific programmes or social needs (institutions) ( World Bank 2011 ). This opportunity to create a dedicated funding envelope for antiretroviral treatment may have satisfied normative constraints more than operational ones, as financial analyses had shown that cost-recovery was playing a very limited role, if any, in financing treatment (ideas). However, the funds did assuage decision-makers’ fears that they might make a promise to citizens that they could not keep.

In December of 2009, the President announced the removal of user fees for antiretroviral treatment ( CNLS-IST 2010 ). Those familiar to the events stated that the decision was made without direct input from the National AIDS Council or health ministry, suggesting an example of ‘Big Man rule’ (institutions) ( Hyden 2006) . Civil society respondents suggested their meetings with him might have persuaded him, and some respondents reflected that he might have felt social pressure considering that neighbours Mali and Niger had already made the decision (ideas). The majority of respondents stated that political or electoral motivations were unlikely (interests).

Discussion

Our study builds on existing literature by integrating concepts of policy networks with institutional, interest- and idea-based theories of policy change. Consistent with those analytic approaches, we observed the important role of institutions, interests and ideas—and particularly changes therein—in influencing prospects for policy change. What this study adds is the synthesis and analysis of the interactions between these two approaches, demonstrating their joint, interdependent influences on the policy process while suggesting that change processes are generally initiated by changes in one of institutions, interests or ideas ( Table 1 ). In periods of stability, networks and the 3Is were mutually reinforcing. Typical of complex systems, alterations in one could set off changes in the others, ultimately resulting in opportunities for policy change.

In general, we observed directionality that moved from a change in the 3Is, to networks, to policy processes and outcomes, but in the HIV network we observed instances where endogenous changes in the network led to changes in the 3Is, and then policy change. The HIV network came the closest to what is described as ‘networked governance’ ( Provan and Kenis 2008) or ‘strategic network management’ ( Klijn and Koppenjan 2000) in the public management literature—that networks are a strategic tool for structuring the policy process and influencing outcomes. Coalition-building in the civil society sub-group was used to advance that networks’ policy goals, resulting in a network structure that was decentralized but dense. Decentralized network structures are more conducive to informal, responsive, and innovative governance and institutional behaviours ( Provan and Milward 1995 ; Howlett 2002 ; Sandstrom and Carlsson 2008) ; indeed, the structure of the HIV network enabled policy experimentation and its eventual diffusion, which ultimately informed policy change. Further research is required to understand whether, why and how low-income country policy networks are managed strategically, and what influence this has on policy processes and outcomes.

Interestingly, donors and development partners had the lowest involvement in the HIV policy case, and the other two cases certainly highlighted the influential role of external actors both directly (interests), and indirectly through changing rules (institutions). Donor rules in the child health and malaria cases initiated change processes moving from the 3Is to networks to policy change. In mandating who must be involved in funding processes, external funders altered the composition of existing policy networks, thus allowing the influence of new actors and their ideas. Two points emerge from this observation. First, funding opportunities and their related processes have become an integral part of health policymaking—if not the principle impetus for policy change in low-income countries—often restructuring existing national policy networks in ways that open a door to policy change. Second, in bringing these processes to countries, external funders play an important, if indirect, role in shaping national policymaking networks, and thus policy outcomes. We heard no evidence during our study that donors or other external actors knowingly aimed to reshape whole policy networks, although further research should examine the extent to which external actors or other elites capture networks—intentionally or not ( Hanefeld and Walt 2015 ). The intentions of donors between these two cases differed: for child health donors had a specific programmatic goal that required policy change, and likely mandated certain actors’ participation with this goal in mind. Notably, policy change is an explicit goal of the global iCCM movement ( Bennett et al. 2014 ). Also relevant is the fact that actors held conflicting policy preferences in this network, perhaps necessitating stronger tactics by the donor and development partners; for example the evidence that development partners intentionally played a ‘brokering’ role, targeting specific individuals in the MOH to gain exposure to new information and to have conversations with peers in other countries (USAID 2009). The question of legitimacy also exists; to what extent must donors and other external actors rely on changing the rules as a means of overcoming their weaker legitimacy in these networks?

The Global Fund, on the other hand, does not explicitly aim to change policy, conventionally defined, but certainly reshapes programmes and then policies through its soft power of targeted resources. Their rationale for changing the composition of stakeholders—particularly for this case—was driven by implementation feasibility more than by the need to overcome conflict. Their interests in the network were represented far less directly than for those of the Gates Foundation. At face value, the Global Fund rules around civil society participation have greater legitimacy than those of the Gates Foundation, but in reality the Global Fund’s decision to shift financing away from the public sector has potentially broader implications for country ownership and democratic legitimacy.

Strengths and limitations

This study has been successful in extending theory-driven policy analysis to low-income country policy processes and demonstrating its utility and applicability in this context. Our choice of diverse policy cases improves the generalizability of these results to other policy cases in Burkina Faso and the choice of a relatively typical French-speaking Sub-Saharan African country suggests some degree of external validity. Further research is needed within and outside of Burkina Faso to confirm these findings.

This study is limited by its inability to quantitatively describe changes in network structure over time. Further research should aim to collect temporal data on network structure throughout the policy process. Interview data in this study are limited by recall bias as well as difficulties in accessing development partners for interviews, which is unfortunate considering the role they played in shaping institutions, networks and policy change.

Implications for policy and practice

We hope this study will be useful to policymakers and practitioners on a number of fronts. First, we hope that those involved in policymaking, as well as researchers, will continue to take a critical and purposeful view towards the policy process and its outcomes, particularly in low-income countries where effective policy solutions are most needed. The complexity of policymaking means that there is no ‘one-size-fits-all’ theory of policy change, but our framework identifies the key variables as well as their change mechanisms. Second, we hope that our focus on networks will encourage the adoption of a network lens in everyday thinking, which we view as critical for managing policy processes in a highly fluid, diverse, policymaking environment. This lens resonated with respondents during interviews and we hope it can be applied more deliberately in the future.

Equipping civil society and other national policy actors with the skills to achieve endogenous network change, with the goal of improving the effectiveness of the policy process and its outcomes, is an important area for further research and practice. Levelling the playing field between national and supra-national interests will become increasingly important, particularly as external actors gain skills and knowledge necessary to manage networks to achieve their own aims.

This study closes the gap between a range of policy theories, including institutions, interests, ideas and networks, proposing and testing a common framework for their joint influence on policy change. Empirical data from three policy cases in Burkina Faso demonstrated that change was generally led by a shift in one of the 3Is, setting off events in the policy process that altered the other Is, changed the structure of the policy network, and led to policy change. Even more broadly, this study suggests a new research agenda that continues to define concepts and methods for exploring the integration of network variables in health policy analysis in LMICs.