Conceptualising context in institutional reforms of land and natural resource management : the case of Vietnam

Research and policy debates over natural resource management in developing countries have largely focused on identifying the set of institutions that best supports resource sustainability and poverty alleviation. We argue that beyond finding the right institutional fit for a social-ecological system, it is equally important to understand how context affects the design and outcomes of institutional reforms. We propose a refined conceptualisation of context, based on a revision of the Institutional Analysis and Development framework. We defend a systematic analysis of context, distinguishing between contextual factors affecting the fitness to local socio-ecological conditions and contextual factors that mobilise power such as political-economic interests and prevailing discourses. We illustrate our argument with empirical research on land-tenure reforms that have been implemented since the 1980s in northern Vietnam. The proposed analytical framework and conceptualisation of context allows a more pervasive understanding of contextual factors, enabling the incorporation of the forms of power that give meaning and legitimacy to institutional change.


Introduction
"Getting the institutions1 right" has been advocated by a wide range of scholars and development practitioners as a crucial step towards poverty 2 alleviation and sustainable natural resource management (NRM).The importance of context in this question of institutional performance has also increasingly been acknowledged in common-pool resource studies, hereafter called "commons studies" (Edwards and Steins 1999;Agrawal 2001;Auer 2006).The evidence that the same institutional arrangements can be successful in one setting and fail in another (Acheson 2006) has led to the conclusion that there is no institutional panacea (Meinzen-Dick 2007;Ostrom 2007).In this light, understanding the contextual factors and mechanisms that create different outcomes for the same institutional reforms appears as an important endeavour to avoid the fallacy of devising new blueprint approaches.
A particularly prominent example of institutional reforms for pro-poor NRM has been that of land reform.Redistributive land reform was a core component of socialist policies across Eastern Europe, Asia and Africa in the 20th century.Other types of land reform initiatives that followed in the last decades range from land registration and titling, tenancy reform, consolidation of landholdings, to restitution of land rights to historical owners (Sikor and Müller 2009).An assumption that has dominated the latest set of initiatives is that secure land tenurethat is "the terms and conditions on which natural resources are held and used" (Bruce 1986 in Scoones 1995) -will positively affect agricultural and economic growth (Deininger et al. 2009) thereby reducing poverty.
This argument has generated heated debate regarding both its theoretical underpinnings and its validity in the field (Peters 2009;Place 2009).A salient feature of this debate is the diversity of research results with ambiguous, or even contradictory, findings.In his recent review of the economics literature on land reform in Africa, Place argues that both theoretical and empirical research has produced "what appears to be ambiguous, inconsistent and conflicting findings in terms of how land tenure variables affect agricultural productivity" (2009,1333).He concludes that such variability arises from different contexts and that contextual factors should be given a greater attention.Similarly, in her review of the contribution of anthropological research to the analysis of land tenure and land reform in Africa, Peters (2009) concludes that the main failures from state-led land policies stem from their overlook of the political, cultural and social context in which they have been implemented.
A wide range of local or distant factors influencing institutional performance have been documented in the literature.The latter, mostly based on case studies, has provided detailed understanding of causal explanations in many national, regional or local settings, but this understanding is messy and fragmented.Few frameworks allow a comparison of context across a large number of case studies.As a whole, efforts to characterise and systematically integrate context into institutional design and analysis have remained limited.
This paper aims at narrowing this gap by refining the concept of context, based on the revision of an existing institutional analysis framework (Clement 2010).Illustrating our argumentation with the case study of land reform initiatives in northern Vietnam, we defend that this analytical frame is well suited to identify the contextual factors and processes that create discrepancy between institutional reforms and their outcomes.Ultimately, we hope that its use will support a better integration of context into policy debates, facilitate comparative analyses and contribute to the design of sound recommendations for institutional design and policy formulation.
This paper is organised in five sections.The following section examines how context has been taken into account in previous land reform studies.Then we review different conceptual approaches to build our analytical framework.In the next section, we present an application of this framework to the case study of land reforms in Vietnam; we introduce the historical background of land reform initiatives in this country and explore the reasons of policy shortcomings.Lastly, we conclude on the lessons researchers and policy-makers can draw from a systematic and holistic context-based analysis of institutional performance.

Land reform, institutions and context
Before examining how context has been considered in earlier studies of land reform, we shall clarify what context means in this paper."Context" comes from the Latin "contextus", derived from Latin "textura", meaning "a weaving"a set of two threads very closely interlaced with each other.This metaphor reminds us of the intricacy of context with other factors, which makes it difficult to characterise.Cunningham said about policy: "Defining policy is rather like the elephant -you know it when you see it but you cannot easily define it" (Cunningham 1963, cited in Keeley andScoones 2003, 22).It does also apply very well to context.Importantly, what is understood by "context" basically depends on the unit of analysis.For example, an economist might define context as the legislative environment in which markets function, whereas a political scientist might rather view context as the macro-economic conditions in which policies emerge, sustain or collapse.In this paper, we define "context" as the "dynamic factors influencing the design and performance of institutions over multiple governance scales".What those dynamic factors entail is conceptualised in the next section.

Context and institutions in studies on land reform
Three major theories on institutions are said to have influenced3 the direction of land reform across countries and continents (Scoones 1995, 118): (1) the tragedy of the commons (Hardin 1968) which, by implying that natural resources cannot be managed sustainably without an external intervention, has recommended for their privatisation or nationalisation; (2) the property rights school (Demsetz 1967), which has identified population pressure on land, technological and economic conditions as key drivers in forging land-people relationships and has largely recommended private property; and (3) the "assurance problem" approach (Runge 1981), which defends that over-exploitation of resources does not arise from free-riding behaviour but rather stems from a coordination problem among resource users.Ostrom (1990) and a large number of commons scholars, including anthropologists, human ecologists and natural resource economists, further developed our understanding of collective action to manage the commons, and more broadly our understanding of social-ecological systems (SESs).They notably demonstrated the limitations of Hardin's theory through a variety of methods and evidenced that communities can and do manage natural resources sustainably under certain conditions (Ostrom et al. 2002).
Partially influenced by these theories, land reform has followed two major directions.First, most have promoted the formalisation of land property rights, whether to individuals or to communities.Second, the terms and conditions of land reform projects across countries have been almost exclusively defined by the State (Sikor and Müller 2009).Both courses of action have been challenged on different grounds but with a common reference to the importance of context.Some scholars have argued that land-policy flaws lie in a misguided analysis of the actual roots of rural poverty.Poverty, in their view, is rooted in politicaleconomic conditions rather than in the lack of formally recognised individual property rights (Peters 2009;Sjaastad and Cousins 2009).Others have emphasised the need to consider land as a means to secure rights (for shelter, food, etc) rather than as an economic asset (Assies 2009) -recognising the role of discourses in giving meaning, and particularly legitimacy, to institutions.Viewing land as a means to achieve food security is likely to give more space to customary arrangements than considering land as a commodity.
Proponents of land rights formalisation have acknowledged that the latter was not sufficient to ensure economic development, but depended on multiple contextual factors (see review in Sjaastad and Cousins 2009).To quote only a few, these range from the presence of customary authorities, population density, urban/rural environment, migration patterns, ethnic tensions (Chimhowu and Woodhouse 2008) to local power disparities, market opportunities, agro-ecological potential (Place 2009) or income level and existence of social security mechanisms (Assies 2009).
Another object of debate is whether land reform should be driven by the State (Sikor and Müller 2009).State-led land registration and titling have in several cases become a source of insecurity and have created conflicts or inequities because of their neglect of local customary arrangements (Assies 2009;Bouquet 2009).A possible explanation is a mismatch between the institutional reform imposed by the State and customary institutions.But a more fundamental issue relates to whose voice is included in decision-making, whose claims matter and who has power to make decisions.As Sikor and Lund (2009, 2) recall "struggles over property are as much about the scope and constitution of authority as about access to resources".
From this review, three types of contextual influences on the outcomes of land reforms are evidenced.First, policy design might be misguided by a poor understanding of the SESs targeted by the reform, where proposed institutions do not adequately address issues at stake or do not fit the SESs -it is a wrong diagnostic.Second, institutional reforms are as much the result of competing interests in and claims over natural resources as driven by the will to improve institutional performance.They have been commonly used to retain power, control population or keep the electoral base obedient, thereby hindering the achievement of their official objectives of poverty reduction, food security or environmental protection.Whose interests dominate policy processes is thus a key issue to analyze and predict policy outcomes.Third, the implementation of land reform initiatives is affected by multiple and diverse contextual factors, located outside of the SESs, that were not anticipated during institutional design.We examine in the next section how these three types of influence have been addressed theoretically in the broader literature on context and institutional reforms on NRM before proposing a framework for contextual analysis.

Conceptual approaches to context in institutional reforms of NRM in general
Commons studies have been largely devoted to the second type of contextual influence mentioned above, that is analyzing the factors internal to the SES that affect institutional performance (e.g., Ostrom 1990;Ostrom et al. 2002).Stemming from the assumption that there is an optimum set of institutions for each SES, a large number of scholars have aimed at identifying general principles for institutional design that can be applied to a wide range of settings (Ostrom 1990).Borrowing from various disciplines, including anthropology and economics, they have identified key factors contributing to institutional performance, such as group size (Agrawal and Goyal 2001), uncertainty of the resource (Wilson 2002) or the level of monitoring (Ostrom and Nagendra 2006).Under this approach, commons scholars have focused on the characteristics of the local SESs, sometimes with reference to larger economic or political forces, but often lacked an in-depth analysis of the latter.
Noticeably, anthropologists have offered thick understandings of the cultural and historical determinants of institutional functioning (Mosse 1997;Cleaver 2000).Political ecologists have also enlarged the 'contextual window' in two respects: first by examining the role of socio-economic change and distant political forces in NRM (Robbins 2004); second, by exploring the intricate relationship of discourses, power and institutions (Hajer 1995;Escobar 1996).These bodies of knowledge have addressed, in particular, the first and third types of contextual influence mentioned before, related with problem framing and power.However, a lack of common analytical frame has hindered comparative political ecology studies and findings have hardly been translated into constructive policy recommendations (Walker 2006).Some scholars have also criticised an a priori focus on political factors in causal explanations (Vayda and Walters 1999).
We believe that institutional analysis offers a valuable starting point to develop an analytical frame if one can overcome its limited attention to power.The Institutional Analysis and Development (IAD) framework developed by Ostrom and her colleagues (Kiser and Ostrom 1982;Ostrom et al. 1994), is one of the most extensively used and tested frameworks in studies of common-pool resource management and offers a useful tool for comparative analysis.It is characterised by a set of general variables, which makes it applicable to a wide variety of settings.These are the biophysical conditions, the community attributes and the rules-in-use (a detailed presentation of the framework can be found in Ostrom 1999;Ostrom 2005).Each variable is decomposed into a nested set of sub-variables, inducing a recognised "aptness for micro-scale analysis and operational clarity" (Auer 2006, 216).A prominent feature of the framework is its linkages across governance levels -though this feature has been under-utilised by commons scholars.It links the operational level, where decisions directly affect resource access and use, to the collective-choice level, where the rules that govern resource access and use are designed, and to the constitutional level, where decisions impact the rules that govern how decisions are taken at the collective-choice level.
Earlier, a few scholars have taken steps to integrate context in the IAD framework.Edwards and Steins (1999) added a variable called "contextual factors" and proposed to identify local and remote contextual factors through retroduction.The latter is a valuable path to capture a wide range of causal determinants but the framework does not go very far into refining our conceptualisation of context.Furthermore, the role of context is limited to situations of extractive use and contextual factors are pre-identified as "demographic changes, technological development and the integration of the resource into the market" (ibid, 209).
Another adaptation of the IAD framework was developed by Rudd (2004), who defined context as "exogenous driving forces".The latter also include: "e.g., demographic, environmental or technological change" (ibid, 119) and the role played by context is similar to that envisioned by Edwards and Steins (1999), where context poses a threat or places a pressure on the SES.
Ostrom and her colleagues also recently made a step towards a broader consideration of context by adding the "political, economic and social settings and related ecosystems" as a variable into a revised version of the IAD framework, called the SES framework (Ostrom 2007).However, in our opinion, the added variable still lacks sufficient conceptual refinement to adequately address power issues.
Commons studies have usually focused on identifying what we define as the right "fitting contextual factors".Those are the original variables of the IAD framework: the attributes of the community (including cultural determinants), biophysical conditions, and rules-in-use.These factors determine the institutional fit, that is, whether the new institutions are adapted to the characteristics of the SES.We recently proposed a revision of the framework to politicise it (Clement 2010), where we added two variables: a) the political-economic context and b) discourses.We further develop the framework in this paper by proposing to conceptualise context as, on the one hand, the "fitting contextual factors" and, on the other hand, the "mobilising contextual factors", defined as the two variables added to the framework (Figure 1).This is de facto a formalisation of the extension of the IAD framework adding a political-ecology perspective.The mobilising contextual factors operate both at the micro and macro-level and refer to the 'factors that mobilise power'.The political-economic context shapes power dynamics and power distribution.Discourses are both an expression and instrument of power and knowledge.The categorisations of the world that discourses convey apply a specific structure of knowledge and power on society, prescribing what is right and wrong, normal and abnormal, legal and criminal, etc. From this perspective, we are interested in capturing the pervasive and microforms of power described by Foucault (1975Foucault ( , 1976)).
The mobilising factors interact with institutions in a variety of ways.Institutions affect power distribution and continuously shape the political-economic structures that have supported their creation and functioning.The political-economic context affects the design, implementation and outcomes of a given set of institutions at multiple levels.Discourses shape the meaning of institutions and affect their legitimacy by shaping actors' beliefs and values.
Finally, we make full use of the often neglected IAD potential for multilevel analysis by explicitly adding the "actors" and "action situations" variables from the collective-choice level to make explicit whose interest and whose decisions are shaping institutional change and who might have influence over the mobilising contextual factors.This also adds further agency to the analytical frame.Mobilising factors similarly affect the action arena at the collective-choice and constitutional choice levels as illustrated in the case study.

Historical background of land reform in north Vietnam
The area of interest for the study is the northern mountain region of Vietnam (Figure 2), home to 11 million inhabitants (13 per cent of Vietnam's total population), among which 55 per cent belong to ethnic minority groups (GSO 2009).The latter have in the past traditionally practised shifting cultivation in the uplands4 , while in the midlands, farmers have commonly practised composite swiddening, a type of farming system which integrates both permanent wet-rice fields and rotating swidden plots.Land practices and farming systems in the midlands and uplands have changed considerably during the past decades.This study considers the land reform on land classified as "forestry land", which covers 75 per cent of the total land area in the northern mountain region.

The collapse of the centrally planned economy
In the 1950s, the communist government of northern Vietnam progressively collectivised the means of production and controlled agricultural inputs and products, first in the lowlands, then in the uplands from 1959.Land with a slope >25 degrees was classified as "forestry land", meaning land for forestry purpose.Forestry land was nationalised and managed by a system of logging companies established by the State, called state forestry enterprises.
In the early 1980s, the country reached the culmination of a major economic crisis.Due to a progressive decrease of agricultural production under the collectivised system, living conditions and food security had regressed to the standards of the late 1960s (Kerkvliet 1995) and agricultural cooperatives were facing major economic problems.State forestry enterprises also encountered serious economic difficulties resulting from a reduction in state support and a shortage of forest resources to exploit following poor management (Nguyen Quang Tan 2006).
Decollectivisation was initiated gradually as a response to the economic crisis with the introduction of contract systems with households for paddy production in 1981.Vietnam shifted more radically to a socialist market-oriented economy in 1986 with the adoption of the Đổi Mổi policies.New rules and regulations on upland management were defined to address the emerging land conflicts that had multiplied between state forestry enterprises and local communities.In 1991, the Forest Protection and Development Law encouraged the allocation of forestry land and forest resources to households for their management, protection and commercialisation.The allocation of forestry land was truly propelled at the national scale in 1994 (Decree 02/1994/CP).The land allocation process has consisted of land demarcation, registration and issuance of land-property titles 5 (also called "red book certificates").

Rights and responsibilities of households and communities for forest and forestry land
The 1991 Forest Protection and Development Law defined three sub-categories of forest and forestry land according to their intended uses 6 : (1) special-use forest/forestry land with an intended use for nature conservation (biodiversity preservation) and landscape protection (including historical and cultural heritage); (2) protection forest/forestry land with an intended use for protecting water resources and soil; and, (3) production forest/forestry land with an intended use for commercial activities: exploitation of timber or non-timber forest products.Forestry land designates land with or without forest cover planned for afforestation.Forest and forestry land classification determines the categories of recipients for forest-and land-use rights as well as their rights and responsibilities.Specialuse and protection forest and forestry land are managed by the State -but the management of protection forestry land can also be contracted to individuals and households.Production forest and forestry land is allocated to individuals, households and communities.
Communities have managed land and forest resources under three types of institutional arrangements: (1) tolerated customary management on ancestral land owned by the State [approximately 0.5 per cent of the total forestry land area in 2007 (Forest Protection Department (FPD) 2007)]; (2) formal contract arrangement with a state organisation for protecting forest, and (3) management on allocated land.Communities 7 were formally recognised as legal forest and land users in the revised Land Law 2003 and in the revised Forest Protection and Development Law 2004.However, contrarily to individuals and households, they have not been devolved in the Land Law 2003 the rights to "exchange, assign, lease or donate the land use right; or to mortgage, guarantee or contribute capital using the land use right" (article 117).
The devolution of long-term land-use rights to individuals and households has been a cornerstone of state policies for rural development and poverty 5 Allocated land superficies ranged from 2 to 30 ha per household for time periods from twenty to over fifty years, depending on the land category (agricultural/forestry), family size, intended use or land cover. 6The category does not reflect the actual land use: production forestry land can be covered by natural forest and special-use forestry land can be degraded with a poor forest cover. 7Communities are defined in the 2003 Land Law as: "Communities of citizens, comprising communities of Vietnamese residing in the same hamlet, village, mountainous village or highland village and similar residential areas with the same habits and customs or of the same line, which are allocated with land by or have land use rights recognized by the State" (article 9).alleviation.But, whereas the land distribution and titling has significantly improved livelihoods in the lowlands, it has had mixed results in the uplands (Dinh Duc Thuan 2005;Sikor and Tran Ngoc Thanh 2007).The imposition of individual property regime on formerly commonly managed land has succeeded, as targeted by the government, in halting shifting cultivation practices and fixing nomadic communities.However, it has in turn resulted in neither livelihood improvement nor better land stewardship.The recognition of the community as a legal recipient of land-use rights has been a much applauded change, but whether it will actually improve the livelihoods of the upland poor still remains uncertain.Using the contextual framework we proposed, in the next section we explore the contextual factors and processes that have created a gap between policy intentions and outcomes.

Applying the framework
Fieldwork was conducted over a 1.5-year continuous stay in Vietnam.Data collection included semi-structured interviews with 80 farmers in three villages of Hoa Binh province, 21 government officials in the provinces of Hoa Binh, Son La, Thai Nguyen and Yen Bai, and 36 national policy-makers (including civil servants, funding agencies, NGOs and researchers) in Hanoi.This long-term cultural immersion and regular interaction with researchers and policy-makers allowed acquiring a relatively accurate understanding of the context in which forest and land policies have been designed and implemented (Clement andAmezaga 2008, 2009).
We expanded the framework across multiple levels (Figure 3) to analyse the possible factors that have created a gap between the outcomes of two sets of policies.On the right hand side, the main decision-making levels (central, provincial and village) and governance levels as per Ostrom's IAD framework are indicated (operational, collective-choice and constitutional).The policy decisions of forestry land allocation to individual households and to communities were taken by central policy-makers and hold both a constitutional and collective-choice character.They have changed the existing rules-in-use from a set of rules (1) to a set of rules (2) (Figure 3).Farmers take operational and collective-choice level decisions [for instance, when the local elite negotiate the rules of land allocation with local government officers (Sowerwine 2004)].
The framework was used as a way to categorise primary and secondary data.This categorisation was helpful to explore the processes that have created gaps between official policy objectives defined at the national level and outcomes observed at the community/household level (Clement and Amezaga 2009).Fitting contextual factors also affect policy decisions at the provincial and central levels.
For instance, if we consider the biophysical conditions: the larger the area of forestry land in the province and the lower its accessibility, the more incentives provincial government officials have to encourage the allocation of forestry land to communities.Allocating larger patches of land to communities allows speeding up the allocation process compared to the lengthy process of demarcating individual plots of land.The distinct effects of fitting and mobilising contextual factors are detailed in the next section for each set of institutional reform.
First, we start with the process of land allocation to individual households and, in the sub-section following that, discuss the process of land allocation to communities.

Impacts of land reform on poverty and natural resources management
This section discusses the impacts of forestry land distribution and titling to individual households.First, it is worth noticing that the process of land registration in the uplands has faced serious difficulties and its overall progress has been slow.The latest available official forestry land allocation figure indicates that, in 2007, 75 per cent of the forestry land area was officially 'allocated' (FPD 2007), but only 31 per cent was actually devolved to households, communities and other private organisations in 2007 and 2010 (FPD 2007(FPD , 2010)).The rest still remains in the hands of diverse state organisations (people's committees, state forestry enterprises, army units and forest management boards).
Several studies have reported that forestry land allocation in the northern uplands of Vietnam has not only failed to reduce poverty but has, on the contrary, aggravated food insecurity (Dinh Duc Thuan 2005;Jakobsen et al. 2007).Land allocation, coupled with settlement schemes, has led upland farmers to replace swiddening cultivation with fixed cultivation -which was one of the major aims of the Government of Vietnam (GoV).However, fixed cultivation has not resulted in increased agricultural productivity as expected.On the contrary, the shortening or suppression of fallow periods has led to soil nutrient loss and decreased soil fertility (Le Trong Cuc and Rambo 2001;Castella et al. 2006;Jakobsen et al. 2007).As a consequence, farmers have stopped cultivation in some areas, thereby losing their major and often unique source of cash income -with no or few other livelihood options available (Jakobsen et al. 2007;Clement and Amezaga 2008).On the contrary, agricultural intensification and increased crop yields were observed in some villages which had resisted the allocation process (Sikor 2006).
Furthermore, whereas the allocation of agricultural land in the lowlands had followed an egalitarian process8 , the allocation of forestry land has been largely captured by the local elite, widening the socio-economic gaps between the poorest members of the community and the better-off (Castella et al. 2002;Sowerwine 2004;Clement and Amezaga 2008).The main criteria to select recipients of forestry land are the financial ability and the willingness of the household to afforest the bare land or to manage the forested land (Decree 02/1994/CP).Better-off households were thus usually preferred.They often received the most accessible land, close to the community settlement.Furthermore, only better-off households have had the ability to make long-term investments and the required market connections to engage in forestry activities (Clement and Amezaga 2008).In addition, land privatisation led to the reduction of common-pool resources such as non-timber forest products and grazing areas, which was of particular importance for the poor (Gomiero et al. 2000;Hager 2006).Lastly, considering non-economic aspects of poverty, land allocation was observed to have disrupted some of the social, cultural and religious relationships that ethnic minority groups had developed with land and forest (Pholsena 2003;Rambo and Jamieson 2003).

Fitting contextual factors
Findings from our study in Hoa Binh Province (Clement andAmezaga 2008, 2009) and from other studies led in the northern uplands of Vietnam (e.g., Celander and Hoang Xuan Ty 2000;Gomiero et al. 2000;Castella et al. 2006) suggest that the outcomes of land allocation have resulted from the neglect of three important fitting contextual factors.Primarily, the inadequacy of state-imposed institutions with the local biophysical conditions was a central policy flaw.Contrarily to the lowlands, upland soil is characterised by a low fertility and the addition of inputs is difficult because of low accessibility and steep slopes (notwithstanding the fact that many upland farmers do not have the financial ability to buy fertilisers).This was supported by evidence from the fieldwork conducted by the authors (Clement and Amezaga 2008) and by others' works in the northern mountain region.Land allocation has added a serious constraint to the viability of upland farming systems (Gomiero et al. 2000;Castella et al. 2006).Empirical evidence thus supports the theoretical claim that common property is generally better adapted to large land areas with low productivity and low accessibility than individual property (McKean 2000).
Second, state-imposed institutions have been in conflict with the local rulesin-use, which relied on collective arrangements.Individual property rights were not only incompatible with shifting practices, but they also clashed in some areas with the local rules for fixed paddy cultivation (Sikor 2006).Furthermore, individual property did not fit the collective rules-in-use that governed grazing and the collection of non-timber forest products (Celander and Hoang Xuan Ty 2000).In our case-study area, the rigidification of the rules for land management broke the fragile equilibrium that had made possible the cohabitation of grazing and cultivation activities (Clement and Amezaga 2008).
Lastly, land registration has not acknowledged the beliefs, cultural practices and social relations of many ethnic groups -which led to resistance and conflicts (Scott 2000, Sikor 2006).For instance, Sikor demonstrates how villagers in a commune of northwestern Vietnam have resisted land registration because it did not acknowledge the complex and dynamic social relations they had developed with land (2006).The land tenure reform also clashed with the rituals related to the choice of a particular area for cultivation of some ethnic groups (Pholsena 2003).

Mobilising contextual factors
Despite the significance of fitting contextual factors, our analysis would be incomplete if it only considered the mismatch between state-imposed institutions and the SES.Pervasive drivers have also greatly contributed to policy shortcomings.
First, the political-economic context has not encouraged the development of economic activities associated with land privatisation.Land allocation has indirectly restricted agricultural activities as it has been coupled with the process of land classification.Most upland allocated is classified as forestry land, where agricultural activities are severely restricted or forbidden.In addition, despite its stated commitment to develop "social forestry" (meaning in the Vietnamese context, "forestry let to the people"), the GoV has provided few incentives to households to engage in forestry activities.Instead, afforestation programmes have supported the provincial forest administration and state forestry enterprises by targeting central subsidies on protection and special-use forestry land rather than providing incentives to farmers for productive activities (Clement and Amezaga 2009).Upland farmers have therefore had few options available to benefit from the land for which they had received property rights.Land allocation also conformed to strong geopolitical interests, as it facilitated the control of ethnic minority population and the securitisation of border areas.
Secondly, upland allocation has failed to reduce poverty because prevailing discourses framed the problem according to biased or false beliefs.Two influential narratives favoured upland allocation.One was relating to deforestation and land degradation, two issues that had become key concerns for policy-makers in the 1990s.The heroes, villains and victims of the story were clearly identified: "In terms of the evolution of barren lands or forest land degradation, sedentary shifting cultivation9 is the most extensive cause" (World Bank 1995, ii).Since at least the French colonial regime, discourses have portrayed shifting cultivation as an environmentally destructive and inefficient practice, which is the product of the low level of education and supposed backwardness of ethnic minority people and of poverty (ADB 1997).As a result of this discursive closure, it has never been considered that shifting cultivation could be desirable and practised because of its benefits under particular biophysical, institutional and cultural conditions.
The second narrative celebrated the pursuit of efficiency.The aim of policies would be to 'rationalise' agricultural production in the uplands by promoting agricultural intensification (ADB 1997) and 'appropriate' land-use practices (Hong Sy Dong 1995).In a post-socialist context valuing individual entrepreneurship, individual property was seen as the best option to maximise land efficiency.The focus on efficiency and productivity is evident in the land classification itself: all land not yet classified as agricultural or forestry land is considered "unused" as per the Land Law 1993, implying that all land with no apparent productive function (such as fallow land) has ultimately to be converted under a productive category (agriculture or forestry).In this narrow view, bare land was perceived as degraded and useless rather than as an important component of a larger land-use system for land resting or grazing activities 10 .
Besides misguiding policy processes, the discourses that supported institutional change have also participated to reinforce poverty through the beliefs and values they carried.Ethnic minorities have recurrently heard over the last decades that they are backward and responsible for deforestation.The resulting feeling of guilt and disparagement was salient in our case study area, where some farmers said the government had allocated the land: "because villagers have too much destroyed the mountains" (Clement and Amezaga 2008).Such stigmatisation might affect farmers' feeling of wellbeing, thereby diminishing their capabilities to achieve what they value (Lamont and Small 2007) -quite the opposite of reducing poverty.
5.2.On the path of reform: land allocation to communities 5.2.1.Impacts of land reform on poverty and natural resources This section now examines the process of land distribution and titling of forestry land to communities.The recognition of communities as legal recipients of land-use rights in Vietnam is relatively recent and has been largely restricted to pilot studies supported by donors and to a less extent by the government.According to official statistics, <7 per cent of the forestry land area has been officially allocated to communities and other 'organisations' in 2010'organisations' in (FPD 2010)).This figure highly varies across provinces.For instance, Son La Province was an exception in its early devolution of land-use rights to communities in 1994-1996 11 and has allocated almost half of the forestry land to communities.However, one should also be cautious with allocation figures when assessing the extent of decentralisation, as what has been called "community" has often been de facto the commune people's committee, the local government executive body.
It is still early to assess on the ground the impact of the devolution of commonproperty rights on poverty reduction.Sunderlin (2006) identified two reasons to be optimistic about policy outcomes.The first is that there are examples of communities who manage forest sustainably.The second reason is that the GoV has been successful in alleviating poverty in the past and that much of this success has been achieved in the rural sector, which is also where community forestry operates.Our analysis of the context under which state-led community forestry is being implemented suggests that, on the contrary, policy outcomes might fall far behind their expectations.

Fitting contextual factors
As indicated previously, community-based NRM has been widely practised in the mountainous regions of Vietnam and associated farming systems have shown to be well adapted to local biophysical conditions (Do Dinh Sam 1994;Tran Duc Vien 2003) -at least as long as access to land and forest was not restricted by the State.A common-property regime thus seems to fit a large variety of conditions found in the uplands.Rapid and dramatic social and institutional changes resulting from sedentarisation policies, government migration programmes and the imposition of individual property might have reduced its present adequacy.Nevertheless, one could reasonably argue that its recognition is a positive move towards a higher institutional diversity -which would logically result in a better fit of institutions with the local context.Yet favourable fitting contextual factors are not a sufficient condition for common property rights to substantially contribute to livelihood improvement.

Mobilising contextual factors
Policy change in favour of communities has been supported by a propitious politicaleconomic context.It meets indeed a number of interests of both provincial and central governments.As explained by a Vietnamese senior cadre managing a donorfunded community-based management (CBM) project: "Before, the government spent a lot of money for forest (…).With community forestry, people can manage forest with their own fund" (formal interview 2006).It is an opportunity for the State to reduce the costs associated with forest protection, notably by reducing the number of forest protection contracts which constitute a major transfer of funds to households12 .In addition, devolving land-use rights to local communities eases the lengthy and costly process of land division, registration and titling to individuals and households (Clement and Amezaga 2009) -costs which are also affected by the biophysical conditions of the forestry land in the province (area and accessibility).As explained by a Forest Protection Department senior official in one of the visited provinces: "It is easier to allocate land to communities.We only discuss with the head of the village, no need to discuss with all villagers" (formal interview 2006).One can be dubious about the pro-poor outcomes of forest and land allocation based on the government's sole interaction with the village elite.
At first glance, a positive discursive change has occurred with the apparition of the counter-narrative "forest protection and management are better achieved if left to communities" in public official discourses, replacing the old narrative on shifting cultivation.However, interviews suggest that the rationale for promoting CBM in popular discourses is very loose.The major argument advanced in favour of CBM was that it is a traditional way to manage forest which has worked in the past.The traditional character of a practice does not per se guarantee sustainable management.The arguments which have provided the theoretical basis for CBM, for example the participation of communities in crafting institutions have been left aside.Discourses have thus 'black-boxed' CBM by depicting the concept as inherently 'good' without providing an articulated justification.
The revised 2004 Law on Forest Protection and Development is also unclear on the legal rights of communities.Despite a section on forest allocation to village communities, the law does not recognise communities as legal owners (articles 3 and 5).Furthermore, allocation to communities is not really encouraged: it is recommended either when forest is already managed or used "efficiently" by the community or when forest "cannot be assigned to organisations, households or individuals" (article 29).It tacitly suggests that the allocation to households is preferable to the allocation to communities.Thus, whereas mobilising contextual factors were favourable to CBM's legal recognition, a closer analysis of the discourses and politico-economic interests suggest that this 'mobilisation' is not only driven by the genuine will to improve the livelihoods of forest-dependent communities but also by other interests.The lack of sound arguments defending CBM in public discourses and the caution implied by the legislation for its adoption are likely to result in a poor implementation, driven by the bureaucratic and economic interests mentioned previously.Discourses indeed suggest that policy-makers have not imagined CBM as the devolution of rights to people to organise themselves but rather as a transfer of responsibilities to manage natural resources according to state rules.For instance, the chair of the Steering Committee of the Social Forestry Training Network in Vietnam, mentioned that one reason to support CBM is that: "The ethnic minority, who possess valuable knowledge (…), are not being utilised as much as they should be" (Bao Huy 2006, 47).Local people are viewed as the instruments of the State to achieve state goals -objectives are thus running counter to the anti-poverty goal of enhancing capabilities and freedoms.

Conclusion
When designing institutional reform for NRM, policy-makers and researchers have usually focused on "getting institutions right", that is identifying which particular set of institutions works better for this purpose.Commons scholars have proposed general principles to design institutions that fit the characteristics of the SES considered (Ostrom 1990).At the same time, there has been an increasing recognition that the outcomes of institutional reforms highly depend on the context in which they are embedded.
We proposed in this paper a refined conceptualisation of context where we distinguished the "fitting contextual factors" and the "mobilising contextual factors".While the former look at how well institutions are fitting the SES, the latter point towards the power and discursive processes affecting the legitimacy and functioning of institutions.
We integrated this conceptualisation of context within the IAD framework and applied this analytical frame to the case study of land reform in Vietnam.First, our analysis points towards the drivers for a poorly designed institutional reform.The blanket imposition of an institutional arrangement which was largely inadequate to the SESs of northern uplands is one of the major causes for the failure of land allocation to households in alleviating poverty.The incongruity between the biophysical characteristics of mountainous areas and how these regions have been imagined by the State has been a recurrent determinant for policy failures in the northern uplands of Vietnam (Sowerwine 2004).
Yet, more than 20 years ago, the members of the Central Executive Committee of the Communist Party of Vietnam (CPV) had identified very similar roots of policy shortcomings in the uplands, namely a lack of in-depth knowledge of the region and the neglect of upland social and biophysical conditions when transferring lowland-based models (CPV 1989).Looking beyond the fitting contextual factors allows the analyst to understand why, despite this accurate analysis, policy-makers have repeated the same mistakes.The reason is that the mobilising contextual factors were not conducive for opening space for other narratives and institutions.Discourses were characterised by a problem closure which did not allow new evidence to change the prevailing narratives on the perceived evils of shifting cultivation.Furthermore, the government had political and economic interest to keep control over forested land and forestry activities.As a result, few farmers have been able to get economic benefits from the rights over land granted to them.
Land allocation to communities has aimed at overcoming the shortcomings of the previous land reforms in respect to the fitting contextual factors.It is argued that the rights to use upland as common-property will allow upland ethnic minority groups to revert to their ancestral land management system.Whereas discourses have changed towards a politically correct narrative praising CBM, they nevertheless remain unclear about its rationale.The prevailing argument is that CBM is 'traditional' -it has little to do with devolving power to communities to manage their land.Rather the political-economic context encourages devolving rights and responsibilities but not decision-making power.
Commons scholars have recently called for a greater institutional diversity (Ostrom 2005(Ostrom , 2007) ) as a reaction against the search for institutional panaceas.Encouraging institutional diversity is a key condition for institutions to fit the multiple SESs they support.In this paper, we show that the 'right fit' is necessary but not sufficient to achieve pro-poor outcomes.Besides fitting the SES they govern, institutions also need to be supported by a favourable political, economic and discursive context.In this respect, political economy and discursive policy analysis (e.g., Roe 1994;Hajer 1995;Fischer 2003) offer useful complementary tools to guide institutional analysts in crafting policy recommendations.We strongly encourage the use of these tools as part of the institutional analysis framework of researchers and policy-makers willing to guide pro-poor institutional reforms.

Figure 1 :
Figure 1: Contextual framework for institutional analysis at the operational level (adapted from revised IAD framework, Clement 2010).

Figure 2 :
Figure 2: Administrative map of the northern mountain region of Vietnam (it includes the northeast and northwest sub-regions, which appear in pink).

Figure 3 :
Figure 3: Extended framework across governance and government levels.In grey boxes: mobilising factors; in italics font: fitting factors.