Conjuring ‘Win-World’ – Resilient Development Narratives in a Large-Scale Agro-Investment in Tanzania

Abstract Through a case study of a public-private partnership (PPP) for development in which the Swedish development agency Sida supported a Swedish company trying to implement a large-scale agro-investment in sugarcane in Tanzania, we unpack the underpinnings of what we call ‘win-world’, a resilient development narrative maintained by actors promoting the investment. Rich empirical descriptions show that this narrative was highly resilient to accumulated academic knowledge and current real-world problems. We found that the privatisation of development increased the resilience of the narrative to evidence of its own shortcomings.


Introduction
To identify how the actors behind a development project manage to produce and sustain a narrative that is counter-proven by history and current real-world problems, and how close collaboration with private actors influences this narrative, we performed a case study of a public-private partnership (PPP) for development between the Tanzanian government and a Swedish company. The PPP, which was supported by the Swedish International Development Cooperation Agency (Sida), planned to invest in sugarcane production in Bagamoyo, Tanzania, a so-called large-scale agro-investment (LSAI). A Memorandum of Understanding between the government and the company was signed in 2006, and ever since, the company's plans were strongly supported by key Tanzanian ministers. In total, 1374 local people were identified as affected by the project (AfDB, 2012, p. 85). Most of these were smallholders, residing or working on the 20,374 hectares of land targeted by the investment and scheduled for resettlement in line with international best practice. Moreover, approximately 70 Barabaig pastoralists would be compensated with land within the estate.
In May 2012, Sida received an application from the company for a 94 million USD credit enhancement guarantee (940 million SEK 1 ), which was Sida's largest ever. Sida's decision on this guarantee was still pending when the Tanzanian government revoked the company's land title in 2016. However, in 2014, Sida granted a smaller guarantee of 16.2 million USD from its regular development budget (of which 6.2 million USD [54 mSEK] was disbursed) (Sida, 2015). That decision was based on various appraisals, which were important documents for our analysis. The investment planned to bring 2000 direct and 10,000-12,000 indirect jobs to the area and engage 1500-2000 local farmers in outgrower schemes that would make an annual income of 13 million USD (Sida, 2014b). The initial aim of the investment, biofuel export to Europe, was later shifted to sugar production for the domestic market. From 2006, the company struggled to secure funding. It did not plant a single sugarcane on the designated land in the decade before the Tanzanian government revoked the land rights.
Events in 2008-2017 were studied by analysing documents and conducting interviews with three of the main proponents of the project and with local community members, in order to compare the proponents' narrative with social realities on the ground.
The present case study is interesting for three reasons: 1) it contributes to an emerging body of literature on how constructions of success are 'produced, distributed and consumed' (Büscher, 2014, p. 79), an issue rarely addressed in ethnographies of development, and examines development narratives in close collaboration with a private company, another under-researched issue.
2) It provides novel insights on contemporary land deals, since insufficient attention 'has been paid to ways in which different forms of knowledge are mobilised and circulated in defence of large scale land allocations or in support of alternatives' (Wolford, 2015, p. 226). 3) The project received unprecedented support from powerful stakeholders, despite limited progress, permitting interesting discussions about dominant win-win narratives. In summary, this case demonstrates how resilient narratives are produced and sustained in privatised development and LSAI.
In the following sections, we introduce the case and review relevant literature on development studies and historical investments in Tanzania (what we call 'accumulated academic knowledge'). 2 We then provide an empirical description of how accumulated academic knowledge and knowledge about current realities were lacking in project design and the LSAI narrative and how other types of knowledge were mobilised and circulated instead. We analyse the resilience of dominant narratives to criticism, including how various practices are used to deflect challenging counter-narratives to sustain its dominance. In a final section, we discuss how privatisation of development led to an especially resilient development narrative ('win-world') in the case and provide some brief reflections on why failed policies are repeated.

Privatisation of development
The sugar investment in Bagamoyo falls within a 'business for development' agenda re-emerging among development bodies in which private sector collaboration is viewed as a way to increase 'development impact' (Sida, 2012). Private sector support was heavily promoted in Africa as part of Structural Adjustment Programmes back in the 1980s (Stein, 2010), but now development agencies are actively partnering with private actors to achieve poverty reduction (Kindornay & Reilly-King, 2013).
Privatisation can be seen as a process whereby activities and resources are transferred to private ownership or control (Swyngedouw, 2005). The current privatisation agenda in development is a core element of neoliberal ideology aiming for a 'reduction of the state', making more space available for private actors to influence policy (Corson, 2010). Involvement of market actors is assumed to promote more efficient, flexible management, counteracting the inefficiencies of state operation (Swyngedouw, 2005). However, PPP is based on the assumption that it is a win-win-win-win solution for recipient states, the private sector, donors and civil society (Kindornay & Reilly-King, 2013). In the Bagamoyo case, the state was expected to gain, for example, dividends through piloting a new government 'Land for Equity' policy to increase state revenues from LSAIs. 3 In the socialist period following independence, the Tanzania government prohibited state involvement in private enterprise, keeping public and private sectors segregated (Coulson, 2013). In recent years, however, privatisation of development has had a major influence on agricultural policy in Tanzania, not least the PPP 'Southern Agricultural Growth Corridor of Tanzania' (SAGCOT) involving development agencies and multinational corporations (Bergius, Benjaminsen, & Widgren, 2017). The SAGCOT development model builds on ideas that more profitable economies of scale must be achieved, for example through PPPs with funding from development agencies (Maganga, Askew, Odgaard, & Stein, 2016). While privatisation of development is proposed as a means to accelerate the fight against poverty, critics point to the risk of further marginalisation of local communities through consolidation of land and other assets by powerful private and public actors (Peluso & Lund, 2011).
2. Background to the case: a decade of delays and failure of an LSAI Our case LSAI was conceptualised as a development project with three key proponents: a Swedish company, the Tanzanian state and Sida. Other important actors included the African Development Bank (AfDB), the International Fund for Agricultural Development (IFAD) and local businessmen. The three key actors supported the project from the outset and consistently presented a positive vision in which different beneficiaries (investors, local communities in and around the investment site and the Tanzanian state) would all 'win' from the project. Moreover, supporting this PPP was in line with aid politics in Sweden at the time, which favoured collaboration with the private sector. Table 1 summarises the benefits proposed to accrue to the three beneficiaries, compared with real-world outcomes.
From the outset, the project was criticised in both Sweden and Tanzania. One serious criticism was that the company had altered the Environmental and Social Impact Assessment (ESIA), to the extent that the Swedish consultant in charge of the process distanced herself from it (Helpdesk, 2009). She claimed that conclusions in the report had been 'substantially altered' and 'wrote the conclusions in the best light' (Usher, 2009). She also claimed that her signature had been forged on the final version submitted to the National Environmental Management Council (NEMC) in order to obtain the environmental permit. The international NGO Action Aid accused the company of land grabbing, chasing people away and failing to obtain free, prior, informed consent from affected communities (Curtis & Mbunda, 2015). In an opinion piece published in Sweden and Tanzania, the company was accused of, for example, dispossessing smallholders and threatening valuable forests, risking increased carbon dioxide emissions (DN, 2009).
Nonetheless, the project retained strong support from key actors, even when the grand visions failed to materialise (see Table 1). On 15 March 2016, the government revoked the company's land title, stating in a letter to the company that it had 'different plans for utilizing the land' (BEE, 2017). 4 At that time, no sugar cane had yet been planted. Nevertheless, the project had many consequences. A 200-hectare seed cane farm was established, but the cane plants it produced had to be burnt each year (from 2007), since the land on the estate could not be prepared for planting and no processing plant had been built. Moreover, local people had been waiting for over five years to be resettled, a delay which caused great loss of livelihood opportunities. At the time of writing, people are still waiting to see what will happen. However, around 300-400 villagers benefited from working on the seed cane plantation, some people were trained in a pilot irrigation project and some received livelihood training and health screening. Overall, the company achieved few of the proposed benefits and suffered severe financial losses (although three company executives were paid between 25,500 to 27,800 USD monthly) (Sida, 2015). The Tanzanian state as a whole gained little benefit and in September 2017 the company filed for arbitration at the International Centre for Settlement for Investment Disputes, a World Bank organ based in Washington, D.C. The company wants compensation from the Tanzanian state for 52 million USD in costs since the project onset (BEE, 2017). development intervention is described as repeatedly disregarding historical experiences (Ferguson, 1990), or, as argued by Roe (1991, p. 287), 'development narratives persist through time and frequently in spite of evidence learned in the field'. More recently, resilient development and conservation narratives have been framed as being about 'constructing value' (Büscher, 2014) and where the promise of 'change' becomes a 'discursive commodity that is constantly reproduced' (Lund et al., 2017, p. 125). Two previous studies use the concept 'resilient narrative' in a way that is relevant for the present analysis. Cairns, Sallu, and Goodman (2014, p. 14) describe resilient narratives as promising 'winwin-outcomes' backed by powerful actors and persisting over time. They claim that the 'resilient narrative' about the Galapagos Islands being pristine and uninhabited, although people have lived there for centuries, is applied in order to justify conservation activities (2014, p. 14). In a study on 'surplus populations', Li (2010) notes 'the repetition of some remarkably resilient narratives about agrarian transition that assume a linear pathway'. She argues that the ability to sustain the image of producing policy for protection, while simultaneously causing dispossession, stems from 'quotidian practices of compromise that enable [. . .] a monstrous disavowal' (p. 80). Thus, both these studies portray resilient narratives as simplified, even inaccurate, truth-claims that disregard empirical realities.
Development narratives build on knowledge mobilised in order to define development 'problems' and justify interventions to solve them (Ferguson, 1990;Li, 2007). The fact that knowledge is inseparable from power and politics is now widely acknowledged in the social sciences (Sumberg, 2017). Thus, individuals or groups selectively produce or use knowledge to 'establish, maintain or enhance their vested interests' (2017, p. 10). Consequently, conflicting perspectives on development reflect the different types of knowledge underpinning them or lacking in them.
The following sections summarise crucial background knowledge to the case study project obtained from development studies, previous LSAIs in Tanzania and the current context. We then examine how this knowledge was addressed in the LSAI narrative and in project design.

Literature review: relevant findings within development studies
There have been some key advances in development studies since the first development projects were launched after the Bretton Woods conference in 1944. Below, we summarise some findings of relevance for the case project.

Solutions based on evolutionary development thinking do not work
A key lesson in ethnographies of aid concerns the 'disjuncture' between 'the "ideal worlds" that development actors aim to bring about through the execution of proper policy and project design' and 'the social reality they have to relate to' (Lewis & Mosse, 2006, p. 2). This lack of attention to context emanates from a lingering Eurocentrism and cultural imperialism where experts tend to view the people and practices they aim to 'develop' as inferior in various ways (Escobar, 1993;Scott, 1998). Li (2007) defines development workers as trustees, 'a position defined by the claim to know how others should live, to know what is best for them, to know what they need' (p. 4). Interventions therefore often start with Western 'experts' assuming they know how to improve life in poor countries through referring to their own country's development path, implementing various schemes out of a 'will to improve' (Li, 2007). This often leads to problematisations of the local context that are inadequate (or even incorrect) and ineffective solutions, ultimately causing interventions to fail to deliver expected outcomes (Ferguson, 1990;Scott, 1998). Eriksson Baaz (2005) describes the history behind the idea of 'evolutionary development', which holds that every society goes through development stages in a linear way, climbing up a ladder of predetermined developments, eventually ending up on top with a 'developed', rich, capitalist system like that in the 'West'. She shows that, although long since deconstructed in development studies, this belief still lingers in many conceptions about development in Tanzania. Indeed, this simplified idea is a core element in an agrarian transition narrative which assumes that small-scale farming in Africa today is similar to that which prevailed in Europe before the Industrial Revolution and that measures which led to 'effectivisation' of European agriculture can be successfully applied to modernise African agriculture today, with the same result. This narrative misunderstands or disregards factors such as radically different soils, geology, climate, infrastructure and social, cultural and agricultural practices of various African regions, not to mention the radically different global context today, with international trade relations, tariffs and so forth (Patnaik, Moyo, & Shivji, 2011). The step on the ladder whereby farmers leave their land and get jobs in industry has also been repeatedly questioned (Li, 2011). Indeed, industry in Tanzania offers few employment opportunities for dispossessed farmers (Wuyts & Kilama, 2016).
Moreover, the idea that small-scale agriculture is implicitly inferior to large-scale agriculture has been repeatedly questioned. As pointed out by Coulson (2015), small-scale agriculture can outcompete large-scale agriculture in many contexts. Coulson refers to the seminal work by Chayanov (1986) describing important advantages small, relatively resource-poor farmers can have over large-scale agriculture. These include the smallholder practice of 'exploiting' the whole family as free labour in tasks such as building stone walls or irrigation canals, instead of having to hire expensive manpower (Coulson, 2015). Moreover, small-scale farmers can work in cooperatives, utilising technologies that are difficult to implement at large scale and avoiding costs of supervisory labour. However, the issue for smallholder farmers is not only efficiency and profit, but also the many other roles agriculture plays in their diversified livelihoods and overall lives.

Top-down interventions rarely work
As mentioned, the agrarian transition narrative is permeated by a lingering notion that 'underdeveloped' states and people are somehow inferior. Consequently, aid recipients are seen as 'passive', 'lazy', 'backward' and not sufficiently 'entrepreneurial'. However, smallholders have been found to be strategic and innovative in dealing with risk and unpredictability in their socio-economic and environmental contexts, which means they do not always opt for maximising profit (Hajdu, Jacobson, Salomonsson, & Friman, 2012). According to Eriksson Baaz (2005), the often-discussed passivity on the part of recipients is usually a sign of resistance to interventions that are rarely attuned to their most pressing needs. It can also be a sign of unequal relationships between development workers and local communities, with the latter risking more if projects fail.
Although the idea of participation in development has a long history, there was a significant shift in development policy in the 1990s as funding bodies increasingly started requiring participatory processes (Groves & Hinton, 2013). The idea that development interventions should be based on the needs expressed by local people and should ideally be 'bottom-up' and designed in a participatory manner has proven more difficult to implement than initially expected. One of the difficulties is finding out what people really want and need, rather than making assumptions. A basic understanding of local contexts and local practical knowledge is also critical to success (Kragelund, 2004;Scott, 1998). In order to find out what people need and to understand local contexts, project designers would need to repeatedly visit project sites and develop close, trustful dialogue with local people.

Literature review: lessons learnt from historical LSAIs in Tanzania
The current wave of LSAIs in Tanzania is not the first. Support has shifted between plantation agriculture and smallholder farming in Africa, including Tanzania, since colonial times (Coulson, 2013;Iliffe, 1979). In Tanzania, three different waves of LSAI can be discerned.
In the early twentieth century, sisal, rubber and cotton were promoted as plantation crops by the German colonial power (Iliffe, 1979). Of these three, rubber and cotton failed within a few years due to, for example, falling global prices, while sisal proved highly suitable for the soils and precipitation patterns in the target regions and, despite fluctuations, is still a top agricultural export commodity (URT, 2016).
The second wave of LSAIs took place around 30 years later, under British colonial rule. It included the 8000-hectare Northern Provinces Wheat Scheme, 'a relative success' (Coulson, 2013, p. 79), and the notorious Groundnut Scheme of 1946, which was 'an abject failure in all three locations, technically and economically' (Coulson, 2013, pp. 79-80). This failure was mainly due to uncritical faith in large-scale agriculture and gross simplifications in project planning, for example, no thorough soil or rainfall analyses were performed and untried machinery was used.
A third, more extended wave of LSAI took place after independence, when the World Bank and the Tanganyikan government promoted a 'Transformation Approach' entailing various initiatives for rural development, with irrigation farming being the model most similar to contemporary LSAI initiatives. These irrigation schemes had little success due to delays, lack of interest from donors and highly saline soils (Coulson, 2013). In the 1970s, the Tanzanian state decided to nationalise large farms and set up parastatal organisations such as NAFCOs (National Agriculture and Food Corporations) to produce maize, wheat and rice (Bernstein, 1981). For decades, the Canadian-supported scheme in Hanang produced much of Tanzania's domestic wheat demand, but was heavily criticised for displacing Barabaig people from their prime grazing land, with severe negative social and ecological impacts (Lane, 1994). It has been concluded that this wave of LSAIs in sub-Saharan Africa 'largely failed' (Deininger & Byerlee, 2012).
To conclude, the track record of large-scale agriculture in Tanzania is 'to say the least, mixed' (Coulson, 2015, p. 63). Many projects have had difficulties achieving expected outcomes and the majority of plantations have contributed little to the local economy beyond the wages paid (Hall, Scoones, & Tsikata, 2017) and have caused serious displacement of local communities. Our case study encountered several of the problems listed, so knowledge about these previous investments would clearly have been valuable for the actors involved.

Data collection and analysis
We analysed the narratives of three main actors (the investing company, the Tanzanian state 5 and Sida) in the case study investment, who together constructed the LSAI narrative. We then compared the results with observations on the project site and the narratives of affected local communities. Information about the main actors' viewpoints was obtained through analysis of policy documents and plans, repeated interviews between 2012 and 2017 (36 interviews), media articles and public statements. Two company executives and several company employees were interviewed, as were influential, high-level state officials in Dar-es-Salaam. Sida programme officials based in Stockholm and at the Swedish Embassy in Dar-es-Salaam were also interviewed. Information about real-world problems was gained through five field visits (2012-2016) with repeated interviews with people on Razaba ranch who were to be resettled, people in surrounding villages targeted for outgrower schemes and who worked on the seed cane plantation, village and sub-village leaders and other villagers employed by the company (20 interviews).
In the work, we treated narratives as stories within the wider discourse. Through analysis of these more specific narratives, we can observe how overarching discourses 'filter through to the everyday' (Doughty & Murray, 2016, p. 304). Using different concepts from discourse analysis and ethnographies of aid, we analysed patterns in texts and statements from LSAI proponents. We also performed a critical evaluation of certain statements.
As shown by Li (2007) and Ferguson (1990), a core element of constructing development discourse includes the process of 'problem description'. By critical analysis of our data, we investigated the 'problem representations' (Bacchi, 2009) that underpinned the narrative and whether these matched 'problems' described in academic literature and by local people.
We also examined assumptions underpinning the LSAI narrative, that is, 'meanings which are shared and can be taken as given' (Fairclough, 2003, p. 55). Social communication by powerful actors is geared towards framing a certain 'common ground' (p. 55), based on such assumptions. Fairclough outlines different types of assumptions, for example 'value assumptions' as 'particular meanings about what is good or desirable' (p. 55). A particular narrative can be supported by various actors, in order to 'universalize particular meanings in the service of achieving and maintaining dominance' (p. 58). In the study case, the view that agrarian transition is about economic progress and should proceed in predetermined steps was based on certain value assumptions about development shared by the actors promoting the LSAI development strategy. Our analysis examined how well academic knowledge was reflected in project design, the LSAI narrative and the underlying assumptions.
We also analysed the data for discursive and material practices applied by the proponents to deflect certain knowledge and create 'resilience [. . .] in [the] face of contrary evidence' (Mosse, 2008, p. 121), what we call 'deflection practices'. We systematically identified criticisms in published articles and reports and the practices applied by the three main actors in their responses to these criticisms in interviews and official statements. Finally, we examined how proponents gained and shared knowledge about the project, by analysing how they referred to each other and other sources of information.

Findings: little consideration of academic knowledge and lessons learnt
Below, we examine whether the knowledge outlined above, that is, lessons learnt in development studies, knowledge about previous LSAI in Tanzania and current realities, were applied in the LSAI narrative and project design for the Bagamoyo investment.

Lacking: knowledge about development studies
We found that lessons learnt in development studies over the past 50+ years were ignored. The actors seemed to have no knowledge about these lessons or dismissed them using various deflection practices.
The analysis revealed several problematic problem representations within the LSAI narrative and in the knowledge produced about the Bagamoyo investment. Indeed, a policy proposing LSAI as a modernisation of agriculture implies the 'problem' that smallholders' land use practices are out-dated. In a podcast, a Sida official framed the 'problem' as being smallholders' 'ancient' agricultural methods and equipment (Omvärlden, 2015). Moreover, the foreword to the SAGCOT Blueprint stated that: 'Tanzania's agriculture is predominantly small holder, characterised with very low productivity due to very limited use of modern technology and techniques of production.' (SAGCOT, 2011, p. 4).
Another, interlinked, problem representation was that of local people being undeveloped, passive or lazy, with the wrong attitude and clearly not in a position to know what was best for them, which was repeated more frequently and in more extreme forms by company representatives. In one interview, a company employee seeking to explain the company's 'way of conceptualizing things' stated that it wanted to help local farmers 'rise up' and that farmers in the planned outgrower villages currently live in a 'black hole' from where 'you cannot see straight'. He added: You can look up and you might see an object moving, and we know it's an airplane, but if you were just in that hole you wouldn't know it's an airplane, you would just see pshht! You know what I mean? Cause you can't fill in the gaps (personal interview, 10 April 2014). Therefore, he continued, one has to start from the basic things, and 'in the village context, simple things are food production, building a shelter, making children'. In a similar vein, one company executive stated that 'we need to change the mentality of the Tanzanian farmer, from subsistence to business man'. The same company executive said in another interview: The people down there, it is actually in general, but especially coastal people, they are often more lazy, because they work with fishing. They go out to fish and then they live on fishing and they don't need to work every day. and 'I experience difficulty making people want to work -I know, on our seed cane farm, in the surrounding villages, they were not interested, it was too hard work' (personal interview, 2 July 2014). However, our interviews with local employees revealed that the 'problem' as perceived by the farmers was that they earned more working for their neighbours in the village than on the seed cane farm, which was one reason why they refused jobs on the farm. In this case, the problem representation of passive farmers was based on what the company representatives perceived as a problem, reflecting the top-down approach in project design.
Sida officials did not as bluntly describe people as lazy, but several statements indicated a perception of people needing to be moved up, in line with the evolutionary pathway: 'To move from subsistence to being part of a company in a relatively short time period is of course a great challenge. If you want to lift them to another level'.
If it had been based on existing knowledge about top-down interventions, the project would have considered the perceived problems and needs of local people, their local economy and livelihood strategies from the outset. This would have recognised findings emphasising the importance of paying attention to local context. In the absence of such a bottom-up approach, the next best thing would have been a participatory approach, where the actors visited the project site independently to discuss people's needs and how these matched the project plans. Instead, the intervention was designed by the company, based on its own perceptions. The outgrower scheme was a clear example of a top-down design, since taking 1.2 million dollar loans and risking their land was not the idea of local people. Asked whether local people actually want to run companies, one company representative said: 'Yes, I believe so. Some, I think. Well, we'll see. I think that they want to. I think many people [. . .] many today maybe don't even know what a company is [. . .]' (personal interview, 2 July 2014). Furthermore, officials at Sida and the AfDB visited the area infrequently, if at all, and generally had little or no experience of rural areas. Moreover, when Sida representatives did visit, they were often accompanied by company representatives, reducing the likelihood of them forming an unbiased opinion.
During an interview with a farmer in one of the villages targeted for outgrower schemes, it emerged that he did not understand the size of the loan he needed. The interpreter translated the sum involved (2 billion Tsh) into a much smaller number and, when corrected, explained that the farmer would not understand 'billion', so he said 'million' instead. It was therefore evident that the farmer did not understand the financial deal. Thus, local people were portrayed as 'backward', but the project plan still assumed that they could operate in the same way as a Swedish businessman. In an interview, a company representative revealed a very distinct 'business is business' perspective:  [. . .] it is a risk they are taking. That is a fact. And the question is then: do they place all eggs in one basket, or what else do they do? How do they survivebut it is a business they are doing! It's not a grant. This provides a good example of why privatisation in development is problematic. When smallholders become outgrowers, they are suddenly perceived and treated as equal partners to the company, when it is obvious that they are far from equal. This might be unproblematic from a 'business perspective', but is highly problematic from a development perspective where the aim is improving people's lives, not introducing great financial risk and unequal partner arrangements.
Reference to other 'modernised' countries was made by all actors, especially company representatives. One company executive compared the case with land consolidation in Sweden: They just have to get modernised agriculture going. Part of that is what we did in Sweden in the 1860s. You need to create larger units. If you have one or two acres per person, you are locked in structural poverty. 6 This view was also expressed at Sida, but not as often or as extremely. One Sida official stated: It's hard to imagine Tanzania in 10 years remaining in a situation where every family lives on 0.5 acres. Just like Sweden changed with land consolidation. It goes hand-in-hand with natural development. Of course it cannot be forced.
As indicated above, this reasoning assumes an evolutionary, linear pathway, disregarding the radically different real-world environmental, social, political and global context of African agriculture today. Moreover, it is underpinned by assumptions on what is inferior and superior. Lastly, there are serious real-world environmental and social challenges as a consequence of large-scale agricultural systems, including in Sweden.
In summary, we found that all actors were lacking knowledge about, or were unwilling to adhere to, relevant findings in development studies.

Lacking: knowledge about previous LSAI in Tanzania
The actors also appeared to lack knowledge of lessons learnt in previous LSAIs in Tanzania, where many of the problems encountered in the project had already arisen. For example, late discovery of saline soils forced the company to reduce the planned plantation size and seek more land outside the estate, which delayed the investment, following similar trajectories as the Ground Nut Scheme. Plans were also repeatedly delayed because rainy seasons were not always considered in the time-line. From previous LSAIs, the company could have learnt about basic agronomy and poor commercial viability. Sida could have considered the overwhelming evidence that the LSAI development strategy has repeatedly failed to achieve expected outcomes and poses serious challenges for displaced pastoralists and smallholder farmers.
None of the company's core project documents mentions learning from previous LSAIs in Tanzania, apart from stating that sugarcane has been cultivated in Tanzania since colonial times (PIM, 2010, p. 19) (although not in the area targeted for investment). In interviews, Sida representatives mentioned their own lack of experience in Tanzania as a problem in handling the company's application. One Sida official repeatedly mentioned lack of previous experience of Tanzania and lack of time to gain insights into the complex case. Thus, in addition to lack of experience, there was a perceived lack of time to gather sufficient knowledge. When asked if there had been any discussion at Sida about past LSAIs, the reply was 'No, I guess we haven't [. . .] no, I can't say that we have' (personal interview, 12 February 2015).
The Sida official mentioned that a colleague had raised the issue of displaced Barabaigs in the previous Hanang scheme and a Sida document states that the experiences from Hanang should be taken into account (Sida, 2013, p. 3). However, the issue was not addressed in Sida's decision to support the company, despite the Barabaigs again being at risk of losing access to grazing land and water.
In contrast, representatives of the Tanzanian state were well aware of previous LSAIs in Tanzania. In fact, one key SAGCOT documents begins: The large-scale mechanization schemes developed in some African countries in the 1960s and 1970s have been substantial failures; often only the roads survive to show that something was done. And the massive failure of the colonial Tanganyika Groundnut Scheme provides an immediate and eloquent example of large-scale agricultural development gone seriously wrong. (SAGCOT, n.d., p. 7) The same document justifies the re-ignited belief in LSAI by claiming that supporting smallholder agriculture in Tanzania has also had very limited success. Interviews revealed that the SAGCOT design was perceived to provide 'very different guiding principles' with which investors had agreed to comply. In essence, there were two main, loosely defined, SAGCOT guidelines: i) Inclusivenessby including smallholders as contract farmers and partners in PPPs, Tanzania can achieve equal growth; and ii) attracting private investment and the right kind of private actors, 'in a way that maximises social gains and allows smallholder farmers to become profitable producers and entrepreneurs' (SAGCOT, 2011, p. 12).
While the availability of large amounts of land for immediate agricultural development is described as 'not true' in a SAGCOT strategy document (SAGCOT, n.d., p. 2), it remained a core assumption, as shown by this quote from a high-level state official: For us in Tanzania we don't know anything about land grabbing. Because we've got so much land. If you try to grab our land, that's a big task. There's so much land which is unutilised. So we don't have any problem of land grabbing, as far as we are concerned.
Indeed, reports of vast tracts of land available for exploitation persist in the agrarian transition narrative (Baglioni & Gibbon, 2013). This 'truth' stems from colonial times and assumes that land with no people is unused, disregarding real-world practices such as rotation cropping, seasonal grazing or villages' reservation of land for future generations. However, this 'truth' has frequently been challenged (Bergius et al., 2017;Maganga et al., 2016). In fact, there is evidence of increasing land scarcity (Börjeson, Hodgson, & Yanda, 2008;Tsikata, 2003), conflicts over land (Peters, 2004) and, in some areas, widespread landlessness (Stein & Maganga, 2017). Nevertheless, a company executive had strong convictions about the availability of land: There is no lack of land, dear Linda! If you have travelled so much in this country now, you have been here so many times and will be here this time for a month. If you travel around by car everywhere and fly over the country, you can see that there is no lack of land! (personal interview, 16 March 2016) While deciding to support the initial loan with a guarantee, Sida took a rather critical stance on this issue, stating that 'there is no such thing as unused land', and suggesting in its final appraisal that the purported availability of land should be assessed later (for the larger credit guarantee) (Sida, 2014a, p. 1). Many studies of LSAI, both early (Lane, 1994;Shivji, 2006) and recent (Anseeuw, Alden, Cotula, & Taylor, 2012;Locher & Müller-Böker, 2014), have shown that conflicts over land are one of the major impacts of LSAIs. In our study case too, land conflicts became a major source of delays and frustration.
To conclude, more knowledge about, or willingness to seriously consider, past experiences of LSAI in Tanzania could have helped at two different levels, by prompting the questions 'Is the LSAI development strategy appropriate, given its historical difficulties?' and 'What is important to consider in order to avoid past mistakes?' The latter question was indeed posed, by SAGCOT. It was also touched upon by Sida, but not elaborated upon in the decision to support the company. However, despite having different starting points, with the state showing more awareness of past experiences of LSAI and Sida showing near complete lack of knowledge, they both ended up with the same vision of LSAI as the solution. Where relevant knowledge did exist, mainly among state representatives, it was deflected, for example, by referring to privatisation in agriculture as novel.

Lacking: consideration of current project realities
While the LSAI narrative disregarded 'accumulated academic knowledge', it also failed to consider current realities at the project site. Such social realities included conflicts over land resulting in two court cases (one against the government and the other against both the government and the company) initiated by the local communities, conflicts over water and severe impacts of delay on people's livelihoods, such as losing out on planned state agricultural services because of planned resettlement (Jimwaga, 2017). Where knowledge of these problems existed, there was a strong tendency to deflect it through different practices and produce conclusions that maintained the image of future success. In particular, company representatives were given space to select the knowledge produced and circulated to other actors, giving them great influence over this image.
One leading character in the project was the charismatic CEO of the company, a former car salesman from northern Sweden. As the former CEO of the Swedish company SEKAB (until 2009), one of Europe's leading ethanol players, he had been a strong promoter of Swedish ethanol production and import. On initiation of the Bagamoyo project, he made spirited presentations in Sweden and Tanzania about the need for LSAI in a world plagued by climate change, poverty and population growth. Indeed, one Sida official stated that he had 'charmed Sida'. His persuasiveness may partly explain why Sida decided to trust the company's knowledge to a large extent. As pointed out by Büscher (2014), capturing the interest of donors and policy-makers is about 'selling success' (p. 79), portraying projects and ideas as valuable. Naturally, the company representatives needed to produce strong justification for the project and to depict the implementation process as smooth, which they did in interviews we held before visiting local communities. However, the contrast between the image presented at the company's office in Dar-es-Salaam and that observed on site was quite striking. For example, in March 2014, company representatives claimed that land use planning in one of the outgrower villages would be finished in July 2014 and that the land earmarked for outgrower schemes was currently lying idle. In fact, land use planning to prepare for the schemes was still not finalised by March 2016, while the village chairperson reported that the land intended for outgrower schemes was currently used for food crops.
Sida officials handling the company's application showed a striking lack of knowledge about the Tanzanian context. They did not know Tanzania had been a German colony, or basic facts about village and land governance. It also became clear that, despite the project being perceived by many Sida officials as 'complex and time consuming' (Sida, 2015, p. 7), officials allocated little time to visit the communities affected. Consequently, Sida largely lacked first-hand information about current project realities, and instead relied mainly on information produced by the company. For example, regarding the current use of land for food production, Sida had received the same information from the company as we did, but without cross-checking it with villagers. Moreover, Sida appeared to rely heavily on information from the company in both its socio-economic and financial assessments. Indeed, when asked about their main sources of information on practical realities for local communities, a Sida official listed the company itself, the government and IFAD. Regarding the financial dimensions of the project, an internal audit concluded that Sida had relied on the company's information about its own financial status and that of other companies in the corporate group, without making an independent assessment (Sida, 2015, p. 2). Other actors placed similar trust in the company, for example, when we asked the AfDB for their basic data supporting a contested statement about the company in the ESIA, they referred us back to the company for information. This provides evidence of the close relationship between them, with AfDB trusting the company to provide information.
Most importantly, the vision that the investment would create 2000 direct jobs and 10,000-12,000 indirect jobs for local communities was produced by the company, in cooperation with AfDB and IFAD. These figures were subjected to independent verification by Sida and deemed 'justified' compared with 'similar projects in the region', although with no details about these projects or any assessment of the quality of the jobs to be provided (MED, 2013). The envisioned influx of 13 million USD to the area was questioned in the consultancy report, but this figure was still used in Sida's decision to justify funding for the company (Sida, 2014b).
The project was designed to reduce the risk to the company, while there were huge risks for smallholder farmers, in particular those targeted for the outgrower schemes. Sida was aware of these risks and, in line with standard procedure, made a risk assessment: 7 'We made the assessment that of course there were a lot of risks in many different areas, but still the potential benefits outweighed the perceived risks' (personal interview, 12 February 2015).
One of the more striking risks mentioned in the assessment was that households engaged in outgrower schemes could become indebted and lose their land: Every business unit of approx. 30-50 farmers will need to borrow 1.25 million USD, that is 20-30,000 USD per person/household. The loans are intended to be taken at local banks under commercial terms. The farmers will not have a guarantee if they can't pay. If their businesses for any reason would go bankrupt, or if the company for any reason is not able to buy their sugarcane, then these people will be severely indebted. USD 20,000-30,000 is very far from the assets a normal family in the area has today. Given the costs of loans and interest, it can be several years before there is a net income. Since their land is going to be used as collateral for this loan, there is also an uncertainty about the farmers' land ownership should the loans not be paid back. Weighed together, the planned outgrower model entails considerable risk-taking for the farmers. There is a risk that the capacity of the farmers is insufficient for forming and running their own companies. (Sida, 2014a, p. 35).
These serious risks were 'accepted' by Sida, with reference to two mitigation measures: i) clear instructions to the company that these risks must not materialise since 'a sound outgrower approach is key to Sida's support' and ii) close monitoring by Sida (with external support) to ensure that these risks were 'managed in an adequate manner' (Sida, 2014a, p. 36). Thus, a purely technical approach was taken, with the assessment appearing to be a checklist exercise to fulfil bureaucratic regulations rather than a true engagement to avoid harmful impacts on local people.
So far, our findings show that the LSAI development narrative was based on skewed problem representations, for example about 'backward' agricultural production and passive farmers living in a 'black hole'. They also reveal that the narrative largely excluded knowledge of 50 years of development studies, previous LSAIs in Tanzania and current realities. In general, the main actors deflected such knowledge, using various practices, underpinned by certain assumptions, to produce a highly resilient narrative.

Producing a resilient narrative
The design of the project and the narrative underpinning it did not reflect what we call 'accumulated academic knowledge' or current realities. In fact, the narrative proved stunningly resilient to counterevidence. For over a decade, the three proponents, including a development agency aiming 'to create opportunities for people living in poverty [. . .] to improve their living conditions', 8 maintained an intact vision of the project as an imminent success.
Disregard for accumulated academic knowledge has played a key role in enabling regular reemergence of the LSAI development strategy since colonialism. However, the problem extends far beyond lack of knowledge. Below, we outline value assumptions and deflection practices that were crucial for producing and maintaining the resilience of the narrative.
The resilience to counter-evidence in the case was underpinned by actors sharing strong value assumptions, particularly regarding how agrarian transition should take place, with implicit perceptions of smallholder farmers needing to 'move up the ladder' and of LSAI being more efficient than subsistence farming. Company employees in particular very explicitly expressed the notion of smallholders not knowing what is best for them. Moreover, the proponents portrayed themselves as knowing how to remedy this, that is, they showed a strong sense of trusteeship. Some of the value assumptions implicit in the LSAI narrative stem from neoclassical economics, for example that of equal access to, and understanding of, information and that of the rational, self-maximising person (Kanbur, 2002). These value assumptions were implicit in the world view of the three LSAI proponents who, instead of being receptive to academic knowledge reproduced their own 'selflegitimating knowledge' (Goldman 2005in Wolford, 2009. Such assumptions are particularly inadequate in a cultural context that emphasises the collective over the individual and where poverty entails lack of information and choices. Poverty makes people want to optimise use of their scarce resources, rather than taking large risks to maximise incomes as required in the 'effectivisation' of agriculture. Value assumptions render certain types of knowledge marginalised, while promoting other types of knowledge. They thereby facilitate various quotidian practices applied to deflect counter-evidence and help maintain the narrative. Below, we outline the deflection practices applied by the LSAI proponents in relation to the main criticism mentioned above. In the Bagamoyo case, the high risks for targeted smallholders were frequently deflected by the actors using technical standard procedures. A practice of rendering problems technical (non-political), excluding political-economic structures (Li, 2007, p. 7) was strongly pursued by all actors. Through performing risk assessments, and referring to guidelines and best practice, the actors depoliticised the issue of risk, rather than addressing it. As the debate around the investment was highly politicised, the deflection practice of 'rendering technical' proved highly important in keeping the narrative intact.
In risk assessments, the practice of rendering technical was combined with the deflection practice of transferring responsibility to business in general, and the company in particular, as a key implementer of best practice and other risk mitigation measures. For example, Sida's risk assessment stated that the company should be given clear instructions to prevent the high risks for outgrowers, but without specifying the type of instructions. Furthermore, state representatives claimed that previous LSAIs in Tanzania should not be confused with the current wave of land deals. They described PPP as something new, creating opportunities to avoid past mistakes, and took the sustainability guidelines signed by partner companies as a guarantee for their practices. These are two examples of how privatisation of development influenced the deflection practices used to shield the LSAI narrative.
When accused of having downplayed potential negative impacts in the ESIA (and even forging the Swedish consultant's signature) to obtain one of the key permits for investment, the company rendered the issue technical. In a public statement, it deflected these accusations by stating that it had simply followed the regulations in letting a Tanzanian consultant make changes to the document in line with comments from NEMC. While that may be partly true, compliance with regulations is a complex issue, and often a matter of interpretation. In this specific case, judging from the nature of the changes made in the final ESIA, which comprised systematically downplaying or even deleting critical points and forging the signature of the Swedish consultant, it seems highly unlikely that all regulations were followed. Reference to compliance with regulations was nevertheless one of the most frequent deflecting practices, especially in interviews with company executives and state officials.
Sida acted quite differently in relation to the ESIA criticisms, which influenced the decision by Sida to refuse the first application for a credit guarantee (in 2009). However, an application from the company a few years later was granted. In the interim, a few key changes had taken place. Most importantly, the company had changed its main end-product from ethanol to sugar, which removed it from the 'biofuel versus food' debate. Moreover, the geographical focus of the investment was shifted away from ecologically highly sensitive areas. When asked about the criticisms of the ESIA, one Sida official said: 'Sida judges every project on its own merit, what it can do today. History is not so important' (personal interview, 16 May 2013). Thus, Sida combined responding to criticism with rendering history irrelevant.
When challenged by local people on Razaba ranch (via our questions, the media or reports), one frequent deflection practice was to divert the focus to the vision, the bigger picture. This practice effectively rendered perfectly legitimate concerns irrelevant. For example, one Sida official said that some local people will always be affected when land is allocated for investment but, given the low population density in general, there was room for effectivisation of agriculture, which the project would achieve. The same practice was used by the company in a public statement written in response to the opinion piece. Instead of discussing the issue, that is the risk of dispossessing farmers, the company referred to 'biased' criticisms which they saw 'no reasons to discuss' and instead described the win-win vision they shared with the Tanzanian government. This effectively shifted the focus away from the real-world problems. This statement also exemplifies yet another deflection practice, namely discrediting the message, or the messenger. The deflection practice of 'discrediting' was most frequently applied by state and company representatives. For example, criticism from NGOs about the LSAI development strategy threatening food security was deflected by a state official as 'disturbance', communicated with evil intent, or as 'misinformation' (The Daily News, 2015). Similarly, questions we raised in interviews were dismissed as 'not professional'. When the critical Action Aid report was published in 2015, accusing the investor of land grabbing among other things, this report was also grossly discredited by the government as 'highly flawed' (The Daily News, 2015). Another practice used by the government in this instance was a more explicit show of power, where Action Aid staff were called to a meeting with more than 20 state officials and strongly encouraged to change their conclusions through talking to a range of government-related authorities and ministries. Moreover, Action Aid staff were banned from visiting the Bagamoyo site.
In summary, the three main actors used a remarkable range of deflection practices to nullify criticisms from other actors, while at the same time sustaining the LSAI narrative and the vision of the Bagamoyo project as an imminent success. There were some instances where critical voices had an impact, especially at Sida, but the vast majority of criticisms were deflected in some way by all three actors.
The ultimate proof of the resilience of the LSAI narrative is perhaps provided by how the three actors reasoned when the company's land rights were revoked by the Tanzanian government. The new Tanzanian president offered the land to another (domestic) investor, so the national strategy to support LSAI remains. In interview, a Sida official was reluctant to call the project a failure and instead used the deflection practice to transfer the responsibility to someone else, in this case the government for withdrawing the land title. This implied that the company and the project had actually not failed. In the same interview, the official stated that Sida still supports a combination of LSAI and support for smallholder productivity to alleviate poverty. A company executive had no problems with us calling the project a failure. However, when it came to the future of LSAI as a development strategy, he immediately mentioned the vision and claimed that future LSAIs are still needed to 'lift up' smallholders from subsistence farming and provide them with a market. Thus all three main actors sustain their faith in LSAI as a development strategy, thereby deflecting important experiences gained in the past decade from the case project.

Conjuring win-world and shielding it from contestation
On examining how a development narrative can be produced and sustained in the face of very strong counter-evidence and criticism, and the influence of private sector collaboration, we found three major mechanisms at work: 1) lack of certain types of knowledge among project proponents or inability to take such knowledge seriously; 2) deflection of criticisms using certain value assumptions and material practices and a wide range of deflection practices; and 3) 'accepted knowledge' about current realities, largely produced by a company seeking to portray the investment in a profitable light, was circulated and consumed by the other proponents of the project. Together, these three mechanisms produced a highly resilient development narrative that we call 'win-world'.
The idea of a development narrative becoming a world in itself has been observed by others. For example, Ferguson (1990) notes that 'development discourse seems to form a world unto itself' (p. 259) and Lewis and Mosse (2006) describe the 'disjuncture' between the 'ideal worlds' that policymakers want to create and 'the social reality they have to relate to' (p. 2).
Win-world was co-produced by the main actors in our case and strongly influenced by the 'win-win' vision. This bubble of co-produced 'truth' was shielded from outside contestation through a range of everyday practices effectively applied to deflect criticism and real-world outcomes such as court cases and livelihood impacts. The company's influence on risk-taking by smallholders was especially evident. In win-world, the assumption that smallholders can take on a 1.2 million dollar loan in order to join an outgrower scheme and operate in the same environment as the company seems reasonable, or even necessary. In the real world, there is a striking difference in context between well-paid executives (regardless of success) and smallholders with limited resources who would bear the main risks. Moreover, misleading problem representations, for example of farmers as passive, produced inadequate solutions. For example, the company blamed a poor work ethic among local people, rather than uncompetitive wages, for labour problems. Together, biased value assumptions, skewed problem representations and persistent deflection practices created a project largely detached from real-world problems, which resulted in failure of the project to materialise and deliver proposed benefits.
What sets this case apart from many other ethnographies of aid is the privatisation of development context whereby development agencies closely collaborated with private actors. This resulted in specific new problems. While private business interests cannot be assumed to automatically coincide with the best interests of rural smallholders in a development perspective, Sida did not question much of the information provided by the company. Indeed, a Sida official listed the company and the Tanzanian government as Sida's main collaboration partners, together with IFAD and AfDB. The company identified the Tanzanian government as its closest partner. Thus rural smallholders were clearly not a first-hand source of information or seen as close collaboration partners. Instead, the project proponents formed an 'epistemic community' where sharing of values encouraged them to take one another's knowledge at face value (Büscher, 2014, p. 80). This, in combination with lack of time for investigations and lack of academic knowledge among staff, made Sida susceptible to selected and biased knowledge, filtered mainly through company representatives.
It is important to note that the disjuncture between ideal world and real-world problems is not unique to this project. Rather, project designers have repeatedly failed over time to integrate even the most basic facts about local context (Coulson, 2013;Li, 2007). However, the strong influence of the company on the narrative and project design strengthened this disjuncture and was particularly problematic in two ways. First, it allowed company representatives to sustain persistent elements of evolutionary thinking, which influenced the project's top-down design and the knowledge produced. Second, this knowledge contained more extreme expressions of these elements that were often based on Eurocentrism, a top-down attitude to local communities and a strong notion of trusteeship. As a private actor, the company had an implicit interest in, and talent for, producing figures proposing huge benefits and a smooth implementation process and in maintaining the image of success until the very end, in order to maintain political support and the interest of funders. Most importantly, this idealised knowledge was circulated by the company to other actors. A private actor has a different type of visioning than a development agency, since: 'in speculative enterprises, profit must be imagined before it can be extracted; the possibility of economic performance must be conjured like a spirit to draw an audience of potential investors' (Tsing, 2000, p. 84). Exaggeration or outright deception may be necessary, as 'conjuring is supposed to call up a world more dreamlike and sweeter than anything that exists' (p. 85). In our case, this dreamlike world was conjured up through relying on economic models and the company's knowledge produced to 'selling success', rather than knowledge about real-world problems.
Plans launched from within the win-world bubble do not take real-world complexities into account, but are instead based on the models and values that underpin win-world, where processes are simplified and predictable. An alternative is possible, for example the actors can portray a vision in which this unproblematic investment will succeed, but must simultaneously be aware that they will need to navigate many different problematic hurdles in order to achieve it (van Ufford, 1993). Our impression is that Sida officials and the company strongly believed in the simplified narrative, but we also concluded that at some level, all three actors were aware that giving problems and criticisms too much space would question their legitimacy. This leads us from discussing how the resilient narrative was produced and sustained to advancing some reflections on why the narrative was so resilient. The overall rationale behind a development narrative is of course that there are winners when a development strategy is being launched or reintroduced. Finding common ground on a joint vision was facilitated by the three key actors all standing to gain from the project and to lose if the project was perceived as a failure. One Sida official stated that Sida did not envision gaining anything from the project, but it is reasonable to assume it would in fact gain from supporting such a 'state-of-the-art' investment, not least through enhanced diplomatic connections. In addition to legitimacy, prestige and money also risked being lost, not least for Sida officials, while the investor was devoted to selling success to maintain financial and political support. By rendering political issues technical and furnishing the image of a smooth implementation process, all actors gained from creating a sense of manageability (van Ufford, 1993).
One strong reason for the Tanzanian state to re-ignite the LSAI development strategy was the potential to attract financial resources. Similarly to Lund et al. (2017), we observed that particular kinds of activities where state actors are able to attract significant external financial support had been promoted. For example, evaluations, feasibility studies and task forces to develop policies have been funded by development agencies such as Sida, the Norwegian Development Cooperation Agency and AfDB. Our findings are therefore much in line with Mosse (2004), who argues that the 'gap between policy and practice is constantly negotiated away' in order to sustain policy models and the resources they attract (pp. 663-664). As Fairclough (2003) states, a particular narrative can serve to 'achieve and maintain dominance' (p. 58). Authority is produced and reproduced when actors confirm each other's positions. One other crucial outcome of the LSAI development strategy is the consolidation of land under state management, regardless of whether the investment materialises or not. In other words, there are a range of 'wins' for proponents of the LSAI development strategy when it re-emerges on the political agenda. However, detailed data on the 'true' wins for different proponents of the LSAIs are elusive and further research is needed.
There are two general implications of our findings. First, if partnering private actors create a stronger resilience to accumulated academic knowledge and real-world problems, and thus a greater disjuncture between this knowledge and project design, it could increase the risk of development projects failing to materialise and to deliver proposed outcomes. Second, the privatisation of development has implications not only for the character of the narrative and risk of failure, but clearly also for how these investments are implemented. For example, smallholders are given exaggerated promises and involved in more serious risk taking, potentially devastating for their livelihoods, and their training is more permeated by a business agenda. Thus, the resilience of win-world is not only a theoretical technicality, but has real-world consequences for the people targeted by privatised development interventions.