How to Promote Institutional Reforms in the Agricultural Sector? A Case Study of Uganda's National Agricultural Advisory Services (NAADs)

This study provides alternative explanations to the limited success of the National Agricultural Advisory Services (NAADS) programme in Uganda, which was considered a role model for a demand&#8208;driven, decentralized and market&#8208;oriented agricultural extension reform in Africa. The analysis shows that the reform process was shaped by the interaction of two coalitions of actors: a donor&#8208;dominated coalition that pushed for radical reform and advocated the total overhaul of the existing structures and the creation of new ones, with the aim of changing the mindset of the extension workers and managers towards the adoption of a more efficient, farmer&#8208;oriented and performance&#8208;based system. The other was a domestic coalition that believed in a &#8220;gradual&#8221; or &#8220;incremental&#8221; approach to promote changes and adjustments in the existing system to make it more efficient and accountable. This group was opposed to the complete overhaul of the existing extension system&#8212;taking an apparently more conservative stance. The study shows that the exclusion of the gradual reform coalition in the design and early implementation of NAADS increased the vulnerability of the programme to political capture and governance problems.

system. The other was a domestic coalition that believed in a "gradual" or "incremental" approach to promote changes and adjustments in the existing system to make it more efficient and accountable. This group was opposed to the complete overhaul of the existing extension system-taking an apparently more conservative stance. The study shows that the exclusion of the gradual reform coalition in the design and early implementation of NAADS increased the vulnerability of the programme to political capture and governance problems.
This is an open access article under the terms of the Creative Commons Attribution License, which permits use, distribution and reproduction in any medium, provided the original work is properly cited.

| INTRODUCTION
There is a widely held view that increasing agricultural productivity is not possible without effective and efficient agricultural advisory/extension services that are responsive to smallholder farmers' needs. Yet, after many years of reform efforts across the continent, there is still a considerable lack of knowledge about how to achieve this goal (Anderson, 2008). This study aims to contribute to addressing this knowledge gap by analyzing the case of Uganda's National Agricultural Advisory Services (NAADS), which constituted one of the most far-reaching extension reform efforts in Africa.
NAADS was one of the first agricultural extension reform models in Africa to aim at developing an alternative to the Training and Visit (T&V) system (Anderson, Feder, & Ganguly, 2006), which had been criticized for its top-down supply-driven nature. NAADS adopted a decentralized, demanddriven and farmer-led system. Public sector extension agents were replaced by contracted private service providers. Farmer groups at village level participated in decision-making processes including the contracting of service providers. This model was perceived as a solution to overcome the longstanding problems in public service provision, such as lack of incentives and staff absenteeism (Chaudhuri, Hammer, Kremer, Muralidharan, & Rogers, 2006). Accordingly, many practitioners and scholars who had tried to establish an alternative to T&V placed their hopes in NAADS. The World Bank and other donors provided substantial funding for the extension reform model. The reform started in 2001 and, even though it was originally designed to be implemented over a period of 25 years, it covered the entire country by 2008.
Early reports, including the preliminary results of a major quantitative impact evaluation by the International Food Policy Research Institute (IFPRI), showed positive results (Benin et al., 2007). However, as the programme matured, enthusiasm faltered. Research evidence indicated problems such as farmers' ambivalence towards the programme (Musemakweri, 2007;Parkinson, 2009), mismanagement of public funds (MAAIF, 2009a), doubts about the capacity of private service providers (Mangheni, Mutimba, & Biryabaho, 2003;Obaa, Mutimba, & Semana, 2004) and low technology uptake by farmers (Bua, Okorio, Kataama, Mutabazi, & Okwadi, 2004). In 2007, the President of Uganda suspended NAADS-a novelty, considering that no other major donor-funded agricultural development programme in Uganda had ever been suspended by the President. Major reform elements of NAADS, such as the focus on private service providers and on working with groups, were reversed, representing a departure from the original principles. In 2014, the reform process was terminated and the function of agricultural extension was transferred back to the Ministry of Agriculture, Animal Industry and Fisheries (MAAIF) with the re-establishment of the Directorate of Agricultural Extension Services. Considering that NAADS was one of the most promising agricultural extension reform initiatives in Africa, its failure offers an important opportunity to derive lessons for future efforts to promote agricultural extension reforms on the continent.
In view of the controversy about NAADS, the first question to consider is the extent to which this extension model, in its original form, succeeded in achieving its objectives. The best source to answer K E Y W O R D S agricultural extension, agricultural institutions, development aid, governance, policy reform, Uganda this question is probably an impact evaluation study conducted by IFPRI (Benin et al., 2011(Benin et al., , 2012. 1 However, the study could not establish a statistically significant increase in agricultural revenue (Benin et al., 2011).
Another source of evidence regarding the performance of NAADS is the World Bank's own assessment of the programme. The Review of the Implementation Completion Report of the NAADS project by the World Bank's Independent Evaluation Group (IEG) (World Bank, 2011) rated the outcome of the project as "moderately unsuccessful." The IEG identified problems such as "increased elite capture, instances of misuse and waste of resources, misdirected inputs, depressed farmer morale, and the weakening of FGs [farmer groups] and farmer fora" (World Bank, 2011,).
In view of these facts, the question arises as to why NAADS, the most celebrated extension reform model in Africa, was ultimately unsuccessful? Joughin and Kjaer (2010) argued that the programme was subject to political capture due to increased political competition. Political pressures led to conflicts between donors and governments (Harrigan, 2003). However, this argument does not explain why the NAADS programme had limited success even before the major political interference started in 2007. A later study (Kjaer & Joughin, 2012) identified lack of ownership of the reform by major actors. This has long been identified as a major obstacle to the success of development interventions and has stimulated far-reaching donor efforts, such as the Paris Declaration on Aid Effectiveness of 2005 (OECD, 2005).
Against this background, the reasons why NAADS did not succeed remain a puzzle. This study aims to contribute to solving this puzzle by providing an in-depth analysis of the evolution of NAADS from the events that led to its establishment up to the point of termination.
The study comprises two components. The first is a careful reconstruction of the events that led to the establishment of NAADS and its change over time, based on objectively verifiable sources, such as documents produced during the process. The second component is an analysis of the subjective perceptions and views of the different actors involved and of their strategies to influence the reform process.
The study uses the Advocacy Coalition Framework (ACF) developed by Sabatier and Jenkins-Smith (1993) in combination with discourse analysis (Hajer, 1995(Hajer, , 2006Roe, 1994;van Dijk, 1998) as a conceptual basis. The ACF framework was selected because it is widely used in the political science literature (Weible, Sabatier, & McQueen, 2009) and it offers the potential to reveal some essential features about NAADS that had been neglected in the earlier literature. The analysis presented here identified two advocacy coalitions. One coalition promoted NAADS and is referred to as "NAADS Coalition" here. The second coalition promoted a more gradual reform process and advocated a greater role of the public sector and referred to here as "Public Sector Coalition." The study shows that a clash in core policy beliefs constituted an unresolved problem that affected NAADS since its beginnings and remained an obstacle throughout implementation. The study concludes that efforts to foster policy learning across these two coalitions (Jenkins-Smith & Sabatier, 1993) will be a promising approach to resolve the longstanding problems of agricultural extension reform in Uganda and other countries with similar governance challenges (Pritchett & Woolcock, 2004). 1 The study combined matching and panel regression techniques to assess the impact of NAADS on farm household income and other outcome indicators. On the positive side, the authors found that "NAADS service providers, compared to others, were rated very highly by a majority of the groups in terms of the methods they used in the training and their performance" (Benin et al., 2011, p. xx). The authors also found that individuals' participation in the farmers' groups was high, but the data indicate that participation declined over time (Benin et al., 2011). Even though NAADS was expected to increase agricultural productivity, the authors found that "direct participation did not have any statistically significant effect on adoption of new crop and livestock enterprises and the improved agricultural technologies and practices considered here, except in the case of recommended planting and spacing practices" (Benin et al., 2011, p. xxii).

| METHODOLOGY
The article is based on three types of data sources: (1) published information relating to the NAADS process (policy documents, reports, academic literature), (2) participant observations of the Ugandan authors through their involvement in the process, and (3) interviews with actors involved in the process which were held both to collect information (a) regarding the facts of the process, and (b) their views, perceptions and strategies vis-à-vis NAADS.
The documents reviewed cover the period since the 1996 elections, the starting point for the analysis. The participant observations of the principal investigator started in 2006. As an active participant, observations were made on how the different policy actors articulated their policy beliefs, how they interacted, how decisions were made and disseminated within the established bureaucracy, and how the different policy actors responded to decisions made in various fora.
The interviews for this study started in 2007. Based on the review of documents and on initial interviews with officials in MAAIF, a map for the policy landscape was drawn in order to get an overview of the process that led to the creation of NAADS. Based on this initial mapping and a snow-ball sampling approach, 56 interviews were conducted for this study. The interviews were conducted in a semi-structured form and the questions captured both information from the respondents regarding the facts about the process as well as their opinions, perceptions and strategies. Table 1 displays the institutional affiliation of the 56 respondents, from all major organizations involved in the NAADS reform.
For each interview, detailed notes were made, including verbatim quotes on key issues. The analysis was based on the ACF and methods of discourse analysis/narrative policy analysis (Hajer, 1995(Hajer, , 2006Roe, 1994;van Dijk, 1998). The interviews were analyzed with a view to identifying the beliefs that the interviewees hold on key issues. Sabatier (1993) distinguishes three types of beliefs: deep (normative) core beliefs, near (policy) core beliefs and beliefs about secondary policy aspects. To simplify the analysis, this study distinguishes only two types: (1) core beliefs, representing fundamental normative positions (such as those relating to the role of the public and private sector in the economy), and (2) policy beliefs (beliefs on specific aspects of a programme, e.g. the mode of procurement).
To identify the core and policy beliefs of the actors involved, methods of discourse analysis as developed by Hajer (1995Hajer ( , 2006 and narrative policy analysis by Roe (1994) were applied. 2 Based on this analysis, the actors were grouped into two major "advocacy coalitions," which shared common belief systems and engaged in similar strategies. Applying van Dijk's (1998) insights, the "self-representation" and "other-representation" of the two coalitions were also identified. It is important to note that the grouping of the actors into two advocacy discourse coalitions is an analytical tool aiming to help in understanding the policy process that took place. Of course, two actors never have exactly the same opinions; there is always some variation within each advocacy coalition. The goal of the analysis was to identify common patterns across different actors, an established approach in qualitative policy analysis.
To ensure confidentiality, the findings from both participant observations and interviews are presented anonymously. The respondents were randomly assigned numbers, following an established approach used in political science studies (Connors, 1999) and they are quoted here as R1, R2, R3, etc. Four strategies were used to ensure credibility in line with the principles of quality assurance in 2 The main approach was to identify recurring patterns of speech that can be characterized as "story-lines" (Hajer, 2006), that is narratives which reflect the ways in which actors perceive and explain the phenomena of interest for this study. The literature on discourse analysis and narrative policy analysis is characterized by an epistemological controversy regarding the question of whether there is an "objective truth" that can be discovered using appropriate analytical methods (positivism) or whether all facts are social constructed (social constructivism) (Jones & McBeth, 2010). This study takes a positivist approach to the analysis of the NAADS process. qualitative research (Guba & Lincoln, 2001;Bitsch, 2005): (1) prolonged engagement at the site, (2) persistent observation, (3) peer debriefing, and (4) checking of interpretations with members of stakeholding audiences. The first two strategies were applied through the long-term involvement of the research team in studying NAADS. The other two strategies were applied by sharing preliminary research findings in meetings and workshops and by discussing them with peers. To control for bias, attention was paid to discussing the findings with members of both advocacy coalitions. As a further strategy for quality assurance, all respondents who are directly quoted in this article were contacted and asked to cross-check the quotes of their statements except for five respondents who could not be reached, because they had retired. To ensure the reliability of the information collected, the method of triangulation was used by combining different data sources and methods to cross-check information. 3 3 The study also applied investigator triangulation, which means "employing more than one researcher, constituting a research team to balance predispositions" (Bitsch, 2005, p. 84

| Overview of the process of designing and implementing NAADS
Figures 1 and 2 provide an overview of the NAADS reform process, which can be divided into two phases. The first phase ( Figure 1) starts with the elections in 1996, in which the "Transformation of Agriculture" was a major element of the platform of the ruling party, the National Resistance Movement (NRM). The second phase ( Figure 2) starts with the introduction of the input subsidy programme in 2005, the first departure from the original reform model, and ends with the situation at the time of submitting this article.

| Institutional and administrative context
The institutional design of NAADS can be placed in the context of the decentralization reform in Uganda, which came into effect when a new Constitution was passed in 1995. In 1997, the Government commissioned a Post-Constitutional Restructuring Report (Ministry of Public Service, 1998) which recommended transferring the agricultural extension services to the districts, a reform that was implemented in 1998. Consequently, the Ministry was downsized from over 1,300 staff members to about 300. The Directorate of Agricultural Extension in MAAIF was disbanded as a consequence of this reform (R11, R14).
In the same year, the World Bank-funded Agricultural Extension Project (Unified Agricultural Extension Project) that had started in 1993 came to an end. This project had two major reform goals: the creation of a Unified Extension Service by combining the extension programmes of different ministries and the introduction of the T&V system of management (World Bank, 2000). According to its Implementation Completion Report, these two goals reflected "mainstream thinking of the Bank at the time," but they were "from the beginning and throughout the duration of the project, contentious design issues. Critics from among the ranks of Government, development partners, and farmers themselves were increasingly vocal in their concerns about the approach as the program unfolded" (World Bank, 2000, p. 4). The ICR rated the outcome of the project as unsatisfactory, its sustainability as unlikely, its institutional development impact as modest, and both Bank and Borrower Performance as unsatisfactory (World Bank, 2000). The ICR highlights a number of governance problems that led to these unsatisfactory outcomes, such as "concentration of vehicles at the central level, late payment of operating budgets and staff allowances, the non-linking of incentives to staff performance, … very limited improvement in the efficiency of management services in MAAIF … as indicated by the poor project management and financial control systems, tardy disbursement of project funds, late procurement of goods and services, delayed civil works, establishment of inadequate monitoring and evaluation systems" (World Bank, 2000, p. 3).
The ICR reflected a change in "mainstream thinking" in the World Bank by pointing out that "project design reflected a degree of centralization in its institutional structure which, in hindsight and with the benefit of wider experience, proved to be overly bureaucratic and insufficiently responsive or accountable to the needs of farmers" (World Bank, 2000, pp. 4-5). According to interview information (R67, R69), agricultural extension was in fact intensively discussed at the same time within the World Bank in the Thematic Group on "Sustainable Agriculture Systems, Knowledge and Institutions" (SASKI). 4 This group recommended the replacement of the T&V model, which had aimed to improve extension through managerial reforms within the public sector, with a new model that aimed to replace public sector extension staff by private service providers while at the same time giving farmers a voice in the contracting of service providers. In Latin America, first experiences had already been made with this approach, and the SASKI group made efforts to establish this new model as part of the World Bank's agricultural extension policy (R69). In terms of the ACF, these events can be interpreted as a change in policy beliefs that took place within the World Bank. The new agricultural extension project for Uganda became the first World Bank extension project in Africa to be designed according to the principles of the new model.
The World Bank-led "design group" made dedicated efforts to create ownership for the new NAADS reform model in Uganda. The Minister of Agriculture at the time, was concerned with the unresolved problems of agricultural extension and was interested in getting new actors, particularly the private sector, involved in extension. He joined a trip to Latin America organized by the World Bank to see the new extension model there in practice (R65). Another important ally for the design group was the Ministry of Finance, Planning and Economic Development (MFPED). This Ministry was certainly interested in identifying options for spending funds on extension in a more effective way. Moreover, MFPED had an interest in working closely with donors to ensure the continued funding of agricultural programmes. International extension experts and scholars, who were keen to see the World Bank abandon their T&V model, were also supportive of the effort and attended NAADS design workshops (R67).
Understandably, the design group focused on the participation of these reform-minded individuals in the planning of the new programme. At the time of the design, the Directorate of Extension of MAAIF had already been disbanded due to decentralization. This denied the World Bank design group counterpart professional staff from MAAIF. The group could still have worked with the Ministry, but, as interview information indicates, they believed that the professional staff in MAAIF would not support the reform effort they had in mind. One interviewee from the group remarked: "MAAIF is the most bureaucratic Ministry and not flexible to new thinking and approaches. Officials in MAAIF continue to live in the past and are resistant to reforms. It is very difficult to advocate for agricultural extension reforms within MAAIF" (R33). Another respondent expressed a similar view: "MAAIF has frustrated efforts to reform and only keeps blaming and opposing any reform efforts" (R53).
However, professional staff within MAAIF point out that the need for reform was well recognized, not only by the Minister, but also by staff within MAAIF. In 1997, an inter-ministerial retreat was organized, which brought together more than five ministries to deliberate policy reforms needed in agriculture following the adoption of the decentralization policy. The outcome of this retreat was the development of the Medium Term Agricultural Modernization Plan (R30, R51, R54). The plan proposed an increase in resource allocation to agricultural extension and research, and to strengthen the link between research, extension and the farmers. Similar to the World Bank's assessment in the ICR, MAAIF staff found that past agricultural extension projects were characterized by weak monitoring and accountability systems and that reform efforts had to address these inherent weaknesses. Contrary to the policy beliefs of the SASKI group, MAAIF staff intended to improve the existing public-sector model rather than to replace it with a different system.
Obviously, the focus on this reform option may have been linked to interests within the Ministry, such as defending employment opportunities in the public sector (Niskanen, 1971). However, interview information suggests that the focus on a public sector reform option was also based on the policy belief that the private sector at the time was still underdeveloped and lacking the necessary skills. MAAIF envisioned that the public extension system should be further developed to form the core of a future pluralistic extension system (R11, R14). Importantly, professional staff of MAAIF were of the view that the reforms needed to proceed gradually to be effective. One respondent explained: "The radical agricultural extension reforms, as being advocated by the NAADS program, are institutionally disruptive and destabilizing; technically destructive; and may not be sustainable in the long run" (R44). MAAIF staff felt that the donors took over the process and that their own reform proposals were neglected. As one interviewee remarked: "Suppose the World Bank was not involved, would these policies be in existence? And if it pulled out now, would these policies hold?" (R14).

| The emergence of two advocacy coalitions
Applying the ACF, one can define two advocacy coalitions that emerged during the design of NAADS. The first coalition, which promoted the NAADS project, is referred to here as the "NAADS Coalition." This coalition was supported by development partners led by the World Bank that included among others, the International Fund for Agricultural Development and the Danish International Development Agency. It also included the MFPED. The second coalition, which was led by the professional staff of MAAIF, is referred to here as the "Public Sector Coalition," as it promoted a publicsector model as an alternative to NAADS. This coalition included academia, local government, the National Agricultural Research Organization and national organizations of farmers.
The two coalitions differed in their core beliefs regarding the role of the public versus the private sector. The NAADS Coalition was convinced that there was no alternative to resolve the longstanding problems of agricultural extension than outsourcing service provision to the private service providers and empowering farmers' organizations to oversee their performance. Applying Murrell's (1992) distinction between "radical" and "gradual" reforms, the NAADS Coalition was convinced that a radical reform was required as earlier reform approaches had failed. In contrast, the Public Sector Coalition was convinced that there was no alternative to a gradual reform approach since, in its view, the private sector did not have the necessary capacity and skills. The analysis of the interviews reveals a considerable level of contempt, distrust and suspicion between the two coalitions. In line with van Dijk's (1998) theory, a positive 'self-perception' and a negative 'other-perception' were frequently observed. For example, a member of the Public Sector Coalition described the NAADS Coalition as "dominated by the so-called 'reformers' who in effect are rent-seekers and captured by reform models designed in Western capitals" (R35). 5 The attitude of the NAADS Coalition to the Public Sector Coalition is illustrated by the following quotation: "Since colonial times, agricultural extension has been a domain of the public sector, highly centralized with the traditional vertical bureaucratic structure. Public extension agents protect themselves and hide their inefficiencies behind this bureaucracy. The NAADS programme is attempting to break this tradition and it is expected to face resistance from the traditional bureaucrats" (R21).

| Divergent views on problem identification
The two coalitions held different opinions on the extent to which the NAADS reform was a response to the problems that affected agricultural extension in Uganda. The NAADS Coalition argued that the design of NAADS was derived from the past failures of the different agricultural extension approaches that have been implemented in Uganda. They also pointed out that extensive technical studies were conducted prior to the design of NAADS; and were used as Working Papers during the design of the programme (Republic of Uganda, 1999).
From the Public Sector Coalition's point of view, the activities of the design group could at best be described as an effort to adjust to the Ugandan context the principles of demand-driven extension reform developed in Latin America and in international extension circles. Alternatives to the outsourcing model, e.g. strategies to strengthen accountability within the public sector, did not have a realistic chance of being considered.
Several extended workshops with national and international participation were held to discuss institutional design issues. According to interviews with World Bank staff, the design group organized three intensive workshops in Kampala, attended by over 50 participants from different institutions in the public and private sectors (R67). Field trips were held to consult local stakeholders, such as district and sub-county governments to discuss various ideas for the design of NAADS (R67).
The Public Sector Coalition perceived this approach to be a preconceived idea by the development partners, who only used the history of extension in Uganda as a justification to gain acceptance for the design principles they had already developed previously within their internal and international fora.
Members of this coalition also criticized the fact that the technical studies that supposedly informed the design of NAADS were mostly written by international consultants, rendering them largely as being externally driven.
A review of the lists of task force members for the design of both PMA and NAADS shows that senior staff of MAAIF were excluded. Interview information from participants in the design workshops revealed limited participation of MAAIF, despite the remarkable extent of stakeholder participation that was otherwise achieved (R11, R74). It remains an open question the extent to which this exclusion was a deliberate decision of the design group based on the anticipated opposition of professional staff to the planned reform, or whether it was perhaps due to other reasons. Since the political leadership of MAAIF as well as MFPED was involved, one could argue that it should have been their responsibility, rather than that of the development partners, to make sure that professional staff were part of the process. MAAIF staff members themselves may also have decided not to participate, as a form of silent protest, or for other reasons. While the exact circumstances could not be reconstructed, the fact remains that senior officials who had long-term experience of agricultural extension in Uganda were not part of the design process.

| Divergent policy beliefs on the institutional arrangements for contracting
The contracting out of extension services was a central institutional element in the NAADS reform, one that was supposed to solve the accountability problem. To empower farmers, the contracting was designed in such a way that farmer representatives would be in charge of procurement supervised by local governments (MAAIF, 2000). According to the guidelines, only farmers' representatives were allowed to vote on the allocation of service providers' contracts in the NAADS Procurement Committee. It was decided that contracts should only be given on a three-to six-month basis, later revised to one year, depending on the enterprise selected, apparently with the goal of obtaining a high level of accountability. The Public Sector Coalition members did not believe that short-term contracting was an appropriate solution, because they felt that smallholder farmers needed to be contracted over longer periods of time if they were to adopt new technologies (R11).
One of the main governance challenges of NAADS, and which contributed to its lack of success, related to problems in contracting (World Bank, 2011). Contracting a public agency with the private sector is a well-known potential entry point for corruption and political interference, as the World Bank's own experience and publications show (Ware, Moss, Campos, & Noone, 2007). Decentralized procurement in a system with weak local capacity is well-known to be particularly vulnerable to governance challenges; and yet, this was the system introduced under NAADS.
Some provisions were made to prevent political interference, such as the exclusion of any politicians from the procurement process, the certification of service providers and the provision of strict technical support and supervision to evaluate competitive bidding (MAAIF, 2000). Interviews held with members of the NAADS Coalition indicated that they did not envisage contracting problems becoming a major issue. Potential governance challenges of procurement were largely absent from the literature on demand-driven extension services at the time. Members of the Public Sector Coalition, who had extensive experience in Government procurement, identified lack of capacity for contract management at the local government level, as a major potential constraint (R3, R11, R14).
One problem that also seemed to have been underestimated related to the transaction costs involved in making short-term contracts at the sub-county level. This problem appears to have forced the NAADS management in the early phase of implementation to decide that extension contracts at sub-county level could only be made for three priority commodities (R11, R26, R67). NAADS served approximately 1,000 farming households at sub-county level (World Bank, 2010), and these were likely to have demands relating to just three commodities. It seems ironic that a programme that placed so much emphasis on being demand driven should introduce such a far-reaching rule limiting the demand it could meet.

| Adoption of NAADS by Parliament
The design of NAADS was part of the Plan for the Modernization of Agriculture (PMA), which played a major role in the 2001 presidential elections. These were the first elections in which the sitting President had to run against an alternative candidate from within his own ruling party, indicating an increasing level of political competition. The opposition criticized the Government for its inability to deliver effectively on its 1996 electoral promise to promote agriculture (Besigye, 2000). As the NRM had made the transformation of agriculture a major topic in the 1996 elections, the political pressure to show progress in this area was substantial. The NAADS Act was passed by Parliament in June 2001 and paved the way for implementation of the NAADS programme.
The two coalitions had rather different views on the process of passing the NAADS Act. The Public Sector Coalition criticized the rush to establish the new programme, which did not allow for any extensive debates in Parliament. One member commented: "The NAADS Act 2001 was expeditiously pursued and passed by Parliament in record time with the full support of the Minister of Agriculture, even when he was aware about resistance to some provisions in the draft bill from his own court yard" (R14).
The NAADS Coalition had a different view. They viewed the NAADS Act of 2001 alongside the PMA (Republic of Uganda, 2000), the Local Government Act of 1997, the NAADS Master Document and the NAADS operational manual. They argued that these were the policy documents guiding agricultural extension. They reasoned that these documents had been systematically developed and linked to each other. Therefore, they did not accept the criticism that NAADS was rushed through Parliament under political pressure (R21, R25, R26, R27, R33).

| Expansion of NAADS
In the early phase of implementation, the NAADS Coalition continued to have the upper hand while the Public Sector Coalition remained in the background. During the NAADS mid-term review in 2004 there was not a single head of a technical department of the MAAIF among the 200 participants of the review. Interview information suggests that this was a form of passive resistance by the Public Sector Coalition to the reform (R2, R6, R14).
Though NAADS was designed as a 25-year programme, it spread rapidly, and by 2008 it covered the entire country (World Bank, 2010). 6 Commenting on this rate of expansion, one member of the Public Sector Coalition observed: "The programme was spreading so fast that any experienced development practitioner should have been able to tell that the speed is not sustainable and does not match the institutional capacity" (R11). The interviews indicate that the NAADS Coalition also considered 6 The Master Document of the NAADS Task Force and Joint Donor Group (MAAIF, 2000) foresaw the roll-out of the programme as follows: six districts in the first two years, 10 more districts in the third year, another six districts in the fourth year, and then 18 additional districts in the subsequent three years. The pace of expansion was to be determined by the experience in the initial six districts. In practice, however, the roll-out proceeded much faster. By 2004, the programme had spread to 21 districts and 153 sub-counties (NAADS, 2004). By 2006, the programme had reached 345 sub-counties in 49 districts. the rapid expansion to be a challenge, but they attributed this to processes outside the purview of NAADS. As one member remarked: "The World Bank had put a freeze on recruitment in local governments, until decentralization was in place and this created capacity gaps in local governments, yet the NAADS design had assumed existence of these capacities" (R27).
For the members of the NAADS Coalition, the main driving force for the fast expansion was the success of the programme in its early phase, which created a strong political demand for rapid expansion (R25, R26, R27, R54). Members of the Public Sector Coalition had a different view. They pointed out that at the beginning of the programme the proponents of extension reform were overzealous in trying to prove to critics that the programme model was working and exaggerated its few achievements. The exaggerated promotion of the achievements invited political demands for expansion of the programme, and since its proponents were keen to win political capital, they responded to these demands affirmatively (R1, R8, R15, R30).

| The input subsidy programme of 2005
In 2005, an input subsidy programme was introduced into NAADS under the name "Institutional Support to Farmer Groups." This can be seen as a first deviation from the original design by the SASKI group. Members of the NAADS Coalition attributed the introduction of the input subsidy programme to the Government's intention to secure votes prior to the 2006 elections ( Figure 2). The Government continued to face criticism from the opposition for lack of achievement in the agricultural sector (Joughin & Kjaer, 2010). Moreover, this was the first election since the NRM had come to power in which opposition parties were allowed to compete. The NRM developed a new programme called "Prosperity for All" (PFA), which served as their main electoral platform.
According to Joughin and Kjaer (2010), the introduction of the input subsidy took the donors by surprise; and they were not able to stop the measure. The fact that other components of the PMA, especially the rural financial services, had not been implemented provided justification for accepting the input subsidy programme. The programme was designed as a revolving fund, which made it more palatable as a substitute for a lack of access to finance as it provided substantial additional financial resources. 7 Farmers were to be provided with inputs on the condition that they pay back 75% of the costs of these inputs (R26, R32, R56). As it turned out, paying back 75% became unrealistic because of the fraudulent behaviour that occurred in the procurement of the inputs (UNFFE, 2011). 8 To what extent the leadership of NAADS expected such problems to occur is not clear. Initially, they may have had an interest in the programme as it provided substantial additional financial resources. 9 Based on the experience with the input subsidy programmes that MAAIF had implemented earlier, the members of the Public Sector Coalition had concerns that implementing the subsidy programme under the decentralized procurement system of NAADS would lead to major governance problems. However, since they had adopted a mode of silent resistance, some of them even viewed the introduction of the input subsidy programme positively, as they predicted that this approach would get NAADS into major trouble. One member commented: "The leadership of NAADS has become overambitious to accept responsibilities under PFA for which we are aware that the programme has no capacity to implement. It is a question of time for them to realize that they will not be able to satisfy the political needs" (R5).

| The suspension of NAADS in 2007
NAADS came to play a central role in the implementation of the PFA Programme, which led to further deviation from the original design principles. The events started with the suspension of NAADS by the President of Uganda in September 2007. The reasons for the suspension have remained a matter of debate. For the members of the NAADS Coalition, the suspension was simply a political ploy by the Government to capture the programme, and to remove the carefully introduced provisions in the design of the programme that aimed to protect it from political interference on the ground. 10 The President himself justified the suspension on the grounds that NAADS officials were misusing the money for the programme (New Vision, 2007).
Members of the Public Sector Coalition saw the fact that the President could suspend the programme as a consequence of NAADS' set-up as a semi-autonomous agency. The NAADS leadership, unlike the leadership of other semi-autonomous agencies in the agricultural sector, had chosen to act independently of MAAIF. In the process, the programme managers lost what might be called "bureaucratic protection," 11 and exposed themselves to unbridled political interference. Another argument of the Coalition was that permanent officials in MAAIF staff have more independence than the staff of NAADS, whose contract has to be renewed every three years. Dismissal of permanent employees from public service, in contrast, requires an independent review panel. Members of the Public Sector Coalition were aware of cases where such panels supported the technical opinion of staff members who had become subject to political pressure (R05). As Public Sector Coalition members pointed out in justification of their position, no other donor-funded agricultural programme had ever been stopped by a direct order of the President (R10, R14).

| The introduction of the model farmer approach
After the suspension of NAADS in 2007, the programme was re-introduced with an expanded mandate to implement the PFA Programme. Under the so-called "model farmer approach," substantial input subsidies were provided to two individual "model farmers" per parish (group of villages). This model farmer programme was to be implemented by NAADS in parallel with the regular NAADS programme. The committee to select the model farmers included politically elected local officials, the local chairperson of the ruling party (NRM), and the local intelligence officer (MAAIF, 2010). The members of the NAADS Coalition opposed this reform element, and the donors threatened to 10 A representative of a donor agency commented: "There is a lot of politics and we do not understand the cause. We have a legal document with the Government of Uganda which we follow. The Government has violated the agreement by suspending the NAADS program" (R53). 11 The concept of "bureaucratic protection" implies that if the President wants to intervene in a programme that is handled under MAAIF, established procedures do not allow him to contact the responsible officer in charge directly, rather he has to go through the Minister and the Permanent Secretary (the administrative head of the Ministry). Even though political pressure can, of course, be exercised through this channel, there is more opportunity for negotiation, because the Minister has, according to the members of the Public Sector Coalition, opportunities to defend the technical opinions of this staff. stop funding the programme altogether. Yet, they were not able to prevent this change and continued to perceive it as a ploy to capture the programme. The members of the Public Sector Coalition saw the lack of ability of NAADS to prevent such far-reaching political interference as the consequence of NAADS' semi-autonomous status. In earlier programmes implemented by MAAIF, the Minister of Agriculture had been confronted with similar demands, but since MAAIF follows a procedure of consultations among professional staff, the Minister had been able to moderate such extreme political demands (R5, R 11).

| Rising criticism of NAADS and the design of Agricultural
Technology and Agribusiness Advisory Services (ATAAS) (2008-2012)

| Disenchantment with the NAADS programme
The World Bank project that introduced and supported NAADS was officially closed in June 2008. An increasing number of newspaper articles and reports indicated that there was mismanagement of the programme in local government, including an increasing number of cases of corruption, lack of access to extension services by the majority of the farmers, increased costs of service delivery and declining quality of service providers. 12 The World Bank's own Implementation Completion Report Review (World Bank, 2011) came to similar conclusions. At the local level, there were also concerns about the skewed financing of NAADS towards three commodities, and the lack of an adequate budget for the other agricultural components, such as disease and pest control and regulatory services. 13 These problems led to farmers abandoning the affected enterprises, including those promoted under the NAADS programme (R11, R14, R32).
Eventually, the NAADS Coalition had to acknowledge that the programme had run into problems. Their interpretation of the reasons for this failure differed from that of the Public Sector Coalition. The NAADS Coalition continued to identify political interference as a major problem. One member remarked: "At the design of the programme, we never anticipated the level of political interference we are witnessing today and management never prepared for it" (R26). Capacity problems and the failure to lay off Government staff were seen as other factors: "We overestimated the capacities of local governments to implement the programme during the design and did not put into consideration that there were many similar Government programmes running in local governments. Yet we had also anticipated that local government staff would be laid off to provide services in the private sector. In my opinion, these assumptions were the greatest barrier to NAADS implementation" (R27).
The speed of implementation, which was also seen as the outcome of political pressure, was identified as one of the reasons for the problems, too. As one member of the NAADS Coalition remarked: "The hurried manner in which we formed the groups did not give enough time for them to be coherent and visionary; and this could have undermined the very cardinal goal of farmer empowerment" (R55). This observation was supported by a survey by the Uganda National Farmers Federation (UNFFE, 2011) that reported disintegration of farmer groups shortly after receiving inputs.
The Public Sector Coalition, in contrast, believed that the main barriers to agricultural extension policy implementation lay in the design of the programme. Apart from problems relating to its set-up as a semi-autonomous agency, they also pointed out that the lack of efforts to achieve a consensus had negative implications, too. For instance, they felt that adjustment of training needs to the new extension paradigm had been neglected: "The educational institutions have not been re-oriented and continue to train in the old fashion. Our training was basically on agronomic practices and that is what we have been practising for a long time. In spite of the new paradigm shift, we still go to the farmer in an agronomic style" (R32).

| The preparation of ATAAS and the fight for laying off public sector extension staff
During the design of the second phase of NAADS under a new investment project Agricultural Technology and Agribusiness Advisory Services (ATAAS) the NAADS Coalition continued to pursue the extension reform according to its original principles. One of the World Bank's main demands in the negotiation phase of ATAAS was that the Ministry should lay off all public extension staff in local government. The ATAAS proposal stated that public extension staff in local government had to be converted to NAADS on performance-based contracts; those staff members that remained in public service would be assigned to non-extension functions (World Bank, 2010). This proposal was contentious as one MAAIF official observed: "This decision not to utilize the remaining public agricultural staff at sub-county, in a country where only about 10 percent of the farmers have access to extension services, is ill-conceived" (R11).
In spite of the eminent resistance, the NAADS Coalition continued to push for the reform. The Ministry of Finance, through the Ministry of Public Service, directed local government to proceed with the conversion of staff. The Public Sector Coalition responded by mobilizing coalition members, who in turn lobbied authorities including the Minister of Agriculture, the Speaker of Parliament, the Courts of Law and the President of Uganda (R5, R10, R23, R38, R44). These efforts climaxed with MAAIF presenting an official policy position on conversion of public extension staff to NAADS (MAAIF, 2009b). This policy position contradicted the position already agreed upon between the MFPED and the development partners. The Prime Minister then directed the Ministry of Public Service to halt the conversion (R11). Thus, the Public Sector Coalition successfully reversed the implementation of one of the key reform elements.

| The reform of the NAADS reform-no consensus in sight
After the 2011 national elections, a new Minister of Agriculture was appointed, who was supportive of the position of the Public Sector Coalition. This change coincided with the expiry of the employment contracts of the leadership of the NAADS Secretariat. Unlike previously, when the contracts were automatically renewed, all posts at the Secretariat were advertised when the contract period came to an end, and a new leadership of NAADS was appointed in 2012. An Inter-Ministerial Technical Committee consisting of line ministries responsible for agriculture, finance, public service, local government and trade and industry was appointed to study MAAIF's reform proposal to establish a harmonized and well co-ordinated extension service delivery system termed as a "Single Spine" extension service delivery. The final report by the Committee recommended shifting the responsibility of agricultural extension back to MAAIF by re-establishing a Directorate of Agricultural Extension Services. The report also recommended merging the NAADS programme and staff into the District Production and Marketing departments at local government level to eliminate the parallel extension delivery systems that had emerged from the NAADS reform. The Committee also recommended separating procurement and distribution of inputs from the extension service delivery. The report envisaged private service delivery as the end state of the reform (Inter-Ministerial Technical Committee, 2013). Cabinet adopted the committee recommendations in June 2014 and the National Agricultural Extension Policy was developed and approved in October 2016, effectively reversing the NAADS reform.

| DISCUSSION
The current literature on the failure of NAADS concentrates on the problems of "political capture," lack of ownership, and clashes in values (Joughin & Kjaer, 2010;World Bank, 2011;Kjaer & Joughin, 2012). While the analysis conducted here broadly supports these findings, this article reveals a more nuanced picture.

| Clash in beliefs and adversarial style of policy-making
Analysis suggests that lack of consensus was one of the major reasons for the failure of NAADS reform. Leading officials among both donors and domestic political leaders believed that professional staff in MAAIF would oppose the reform. Yet, the belief that such a far-reaching institutional reform could be pursued without actively involving the MAAIF proved to be untenable. Using the concepts of the ACF, this study identified a clash in core beliefs among the two coalitions, especially regarding the role of the private sector versus the public sector, the institutional design of NAADS, organization of contracting and speed of reform. No effort was made to reach common ground. The events clearly reflected an adversarial style of policy-making, driven by the shift in political power between the two coalitions. The Public Sector Coalition had no influence on the design and early implementation of NAADS, but later gained increasing influence after the programme ran into difficulties. 14

| Radical versus gradual reform
The approach of the NAADS Coalition to pursue the reform against possible resistance within MAAIF is understandable. They saw NAADS as the main opportunity to establish an alternative to the T&V system in the developing world which had been widely criticized within the international extension community. The reform had the full political support of MAAIF and the backing of the Ministry of Finance and other stakeholders. Apparently, the NAADS Coalition did not see the value of the technical expertise of MAAIF staff and their experience on how to best deal with mitigating political pressure within the existing bureaucratic system. The international attention to the programme as a replacement of the T&V and the willingness of leading international experts to contribute their 14 While the analysis presented here focused on core and policy beliefs, it is important to acknowledge that material interests may also have played an important role on both sides. Based on Niskanen's (1971) analysis of the theory of bureaucracy, one can argue that some of the Public Sector Coalition members were critical of NAADS as they saw it as being associated with a loss of power and budgets, and a consequent loss of access to "rents" through the formation of a new agency outside their sphere of control. Likewise, some of the NAADS Coalition members within Government may have had their eyes on the well-paid employment within the autonomous new entity and on future rewards from their close association with the World Bank. Likewise, proponents of NAADS within the donor community may also have been motivated by career opportunities and international recognition associated with the achievement of a landmark reform. expertise to its design may have created an incentive to "oversell" the programme and turn a blind eye to its problems.
There has been a longstanding debate as to when a radical reform approach is more suitable than an evolutionary reform approach, which is gradual and based on consensus (Murrell, 1996). The radical nature of the NAADS reform programme may be seen as corresponding to what Murrell (1992) termed "the crystallization of the faith in reason over experience and experiment." Murrell argued that an evolutionary approach values the accumulated experimental wisdom of society-the stock of its personal knowledge. Accordingly, this position is sceptical about reforms that promote entirely new schemes derived purely from theory, particularly if they exhibit speed, irreversibility and are large in scale.
The institutional design issues identified in this study indicate that a gradual reform approach may have had unforeseen benefits within the semi-autonomous nature of NAADS. Semi-autonomous agencies have proved to be institutional structures that limit political interference (Cukierman, Webb, & Neyapti, 1992). 15 The NAADS programme, however, shows that these predictions did not hold in the Ugandan context. It is difficult to assess which of the other problems might have been avoided if MAAIF expertise had been brought to bear on the design of NAADS. The analysis presented here suggests that the experience of MAAIF could have been helpful in avoiding some of the procurement problems and design shortcomings.

| The role of development partners
The World Bank and several other donors provided close to 80% of financial support required for NAADS (Benin et al., 2011). The fact that a high level of donor influence can have negative effects is well established in the literature, especially with regard to institutional reforms (van de Walle, 2001). 16 As Shirely (2008) argues, donors have limited possibilities to promote institutional reforms from the outside. Their best chances are to support genuine home-grown initiatives. 17 The case of agricultural extension, unfortunately, has a long history of donor-driven initiatives that are uniformly promoted across many countries (Anderson et al., 2006). Though efforts were made with NAADS to create local ownership, the analysis presented here indicates that these efforts did not go far enough for the reform to succeed.

| Policy learning across advocacy coalitions
The NAADS case is in line with one of the main insights from the ACF: "When core beliefs are in dispute, the line-up of allies and opponents tends to be rather stable over periods of a decade, or so" (Sabatier, 1993, p. 27). In such a situation, it is important to move from an adversarial to a collaborative policy-making style and to promote "policy learning" across coalitions (Jenkins-Smith & Sabatier, 1993). A neutral third-party broker could play an important role in creating a forum where such policy learning could take place. In the case of Uganda, it will be challenging to identify a broker that has sufficient capacity and credibility, since many organizations have taken sides over the past decade. Facilitators who are experienced in negotiation and conflict resolution could play an important role in this context. The negotiations could focus on subject matter issues, such as the question of 15 The theoretical literature also supports this view (Voigt & Salzberger, 2002). 16 An evaluation of World Bank-supported public administration reform efforts underlines the challenges of such reforms (IEG, 2008). 17 This view is also argued by Moyo (2010), who argues that foreign aid has over the years promoted despondency and undermined local innovations and initiatives. how accountability can be established in either a public or a private sector model, and could also lead to experimentation, as a strategy of policy learning (Olken, 2007).

| CONCLUSIONS
This study makes it possible to draw some conclusions for agricultural sector reforms more generally. A gradual approach to reform can exploit windows of opportunity, which arise when the opposition to reform is not powerful, but a radical approach may only be suitable for interventions that require the "stroke of a pen" to be implemented, such as changing the exchange rate, or abolishing a subsidy or an import tax. The findings of this study indicate that for far-reaching and large-scale institutional reforms, such as agricultural extension reform, there is a need to build consensus among key actors. Building consensus makes it possible to draw on the accumulated experience of all involved and promotes ownership, which allows for policy-oriented learning. A reform built on consensus also has greater flexibility to experiment and to change approach where necessary. Finally, the study lends support to the proposition that donors might be more effective if they engage in promoting better governance and support "home-grown" reform initiatives, as highlighted in the World Bank's World Development Report 2008 (Byerlee, De Janvry, Sadoulet, Townsend, & Klytchnikova, 2008).