Will REDD+ Safeguards Mitigate Corruption? Qualitative Evidence from Southeast Asia

Abstract High levels of faith and finance are being invested in REDD+ as a promising global climate change mitigation policy. Since its inception in 2007, corruption has been viewed as a potential impediment to the achievement of REDD+ goals, partly motivating ‘safeguards’ rolled out as part of national REDD+ readiness activities. We compare corruption mitigation measures adopted as part of REDD+ safeguards, drawing on qualitative case evidence from three Southeast Asian countries that have recently piloted the scheme: Indonesia, the Philippines, and Vietnam. We find that while REDD+ safeguards adopt a conventional principal-agent approach to tackling corruption in the schemes, our case evidence confirms our theoretical expectation that REDD+ corruption risks are perceived to arise not only from principal-agent type problems: they are also linked to embedded pro-corruption social norms. This implies that REDD+ safeguards are likely to be at best partially effective against corruption, and at worst will not mitigate corruption at all.


Introduction
Mitigating the negative effects of climate change is an international policy goal. 1 Billions of dollars in public and private monies are financing programmes aiming to reduce or prevent the sources of greenhouse gasses (Rosenberg et al., 2014). An initially promising mitigation mechanism to have emerged internationally is the Reducing Emissions from Deforestation and Forest Degradation (REDD+) programme. REDD+'s exchange-based logic is simple: pay poorer countries to keep some of their forests standing. Specifically, it seeks to incentivise developing countries through performance-based payments to maintain rather than destroy their forests and to manage their existing forest resources in ways that reduce emissions (Angelsen, 2009;Karsenty & Ongolo, 2012).
To date, REDD+ has been piloted in several countries. 2 Pilot or 'readiness' activities are the first phase of a national REDD+ programme and encompass policy assessments and strategy adoption that feed into the second and third phases of enacting policies, enforcing measures, and quantifying forest carbon changes in exchange for payments. REDD+ readiness activities are almost exclusively publicly financed, to the tune of billions of dollars. Money is channelled through multilateral facilities like the UN-REDD Programme, through bilateral country agreements such as Norway's International Climate and Forest Initiative (NICFI), 3 and via domestic resource mobilisation in REDD+ pilot countries (Norad, 2014;Norman & Nakhooda, 2014).
Many pilot countries rank low on formal governance indicators, including measures for the control of corruption, political equality, rule of law, and efficient public goods provision (Kaufmann, Kraay, & Mastruzzi, 2011). In this article, we examine the first of these risks, corruption, which may threaten the effectiveness and equity of REDD+ as a major component of global climate change mitigation efforts. We have a dual focus: first, we examine the corruption risks that are hypothetically likely when implementing REDD+; second, we consider the efficacy of corruption mitigation measures that have been put in place as part of REDD+ safeguards. Drawing on evidence gathered through a set of structured qualitative case studies of three Southeast Asian countries to have piloted REDD+ (Indonesia, the Philippines, and Vietnam), we focus in particular on publicly financed REDD+ readiness activities in these three countries, since the majority of REDD+ financing to date comes from these sources (Norman & Nakhooda, 2014).
The article is structured as follows. In the first two sections we outline background information about REDD+ and review the relevant literatures on corruption and the forestry sector. In the third section, we describe the methodology of our study. In the fourth section, we explore how the evidence generated through the case studies answers our research questions. We argue that, despite principal-agent and normative anti-corruption approaches pursued as part of REDD+ readiness activities, forms of corruption could still derail country REDD+ programmes and, with them, some national-level climate mitigation plans. We conclude with reflections on future research directions.

REDD+'s evolution and structure
Reducing Emissions from Deforestation and Forest Degradation (REDD+) is an institutional approach for cutting carbon emissions from changes in forestland (Angelsen, 2009). It was formalised at the 13 th UNFCCC conference in Bali, 2007, although its roots go back further to the Noel Kempff Mercardo Climate Action Project of 1997 (Holloway & Giandomenico, 2009;Karsenty & Ongolo, 2012). Initially focused only on measures aimed at countering deforestation and the degradation of forests, negotiations at the 2009 UNFCCC meeting in Poznan resulted in the inclusion of conservation and the enhancement of forest carbon stocks, and REDD became known as REDD+. An umbrella term for local, national, subnational, and global activities, the programme encompasses specific forest enclosure-conservation projects meeting particular governance criteria while also referring to policy, legal, and institutional frameworks that make these projects feasible (Angelsen, 2009).
A variety of institutional actors are involved with REDD+ readiness activities. Among the more prominent are the United Nations REDD programme (which assists developing countries to participate in REDD+) and the World Bank's Forest Carbon Partnership Facility (which provides funding to states to prepare for implementing REDD+). Implementing countries have put in place specific REDD+ institutional arrangements to manage REDD+, including steering groups, government commissions, multi-stakeholder councils, and national strategies, guidelines, and regulations. Local and national government agencies, especially those related to environmental and forest management as well as civil society actors, play key roles in laying the groundwork for REDD+ at national and sub-national levels. 5

Corruption and forestry: agents and institutions
We have a dual focus in this article. First, we ask what types of corruption risks are hypothetically likely to emerge in publicly financed national REDD+ programmes. We assume that the exact types of corruption risks and behaviours that will emerge in each national setting will depend on the political, social, and economic context, but we also expect broad patterns in these risks and behaviours to conform to our theoretical perspective on corruption. We map the corruption risks reported via our three case studies against theories of corruption and the hypothetical risks identified in early literature on REDD+.
Second, we ask what probable impact the anti-corruption measures that have been put in place as part of REDD+ safeguards work will have on mitigating identified and hypothetical corruption risks. We reflect on this issue of likely impacts by examining how mitigation measures map onto theoretical explanations for corruption. We cannot examine whether anti-corruption measures have actually prevented or reduced corruption in REDD+ since the main part of financing for the initiative has yet to flow and direct payments to stakeholders are still exceptional (Ickowitz, Eills, & De Sassi, 2017;Norad, 2014). However, it is important to understand the types and probable impacts of anti-corruption measures put in place during the pilot phase, given pressures to disburse REDD+ payments and that accompanying corruption risks may increase in future.

Theories of corruption
Though debates continue (Heywood, 2017;Rose-Ackerman & Palifka, 2016;Marquette & Peiffer, 2015;Persson, Rothstein, & Teorell, 2012), two broad explanations for corruption have emerged, defined as the abuse of entrusted authority for personal gain (Kolstad & Søreide, 2009). First, corruption has been traditionally modelled as the consequence of basic principal-agent challenges. Asymmetries of information and individual interests between a principal (that is an employer) and an agent (that is an employee) grant the agent both the motive and the discretionary opportunities to pursue his or her interests at the expense of the principal (Klitgaard, 1998;Marquette & Peiffer, 2015;Rose-Ackerman, 1978;Sappington, 1991). In the case of the public sector, government authorities (agents) often enjoy high levels of monopoly and discretionary authority over state resources and decision-making functions. They thus have a significant informational advantage over citizens (principals) and ample opportunities to abuse their entrusted power for private gain. Assuming individuals are rational actors, opportunities for profitable rent-seeking behaviours are unlikely to be dismissed, especially as principals can never fully monitor the behaviour of agents.
Second, scholars and practitioners have recently argued that corruption should be seen as a function of prevailing norms and functions. That is, corruption may be systemic because of the existence of pro-corruption social norms such as nepotism and neo-patrimonialism that shape preferences towards corruption rather than away from it (Marquette & Peiffer, 2015;Persson, Rothstein, & Teorell, 2012;Berninghaus et al., 2013;Andvig & Moene, 1990). In this type of context, corruption is the expected form of behaviour, the accepted way of doing things that serves certain social functions such as organisational structure. Rather than being a form of 'rule breaking', corruption is instead a form of 'rule following' that depends on the behaviour of other individuals (Berninghaus et al., 2013). Where individuals expect others to default to corrupt interactions, including those typically referred to as 'principles' and 'agents' in a given setting, corruption should be the expected form of behaviour since there are so few incentives to abstain from it (Persson, Rothstein, & Teorell, 2012). Dong, Dulleck, and Torgler (2012) furnish panel data indicating that a willingness to be corrupt is influenced by the perceived activities of peers and other individuals, while also noting that a critical mass of cooperative individuals is required to induce a positive dynamic of societal cooperation. We tend to agree with Marquette and Peiffer (2015) that while principal-agent models explain many corruption dilemmas, viewing corruption as also tied to social norms and functions helps us appreciate why this phenomenon is so stubbornly persistent.

Corruption and aid
Much of the funding for implementing REDD+ currently entails large fiscal transfers from rich to poor (er) countries. Political elites may use REDD+ monies in a similar fashion to foreign aid or the revenues from high-value natural resources such as oil, gas, mining, and forests. The relatively quick appearance of large amounts of money could trigger rent-seeking behaviour and worsen institutional quality (Djankov, Montalvo, & Reynal-Querol, 2008;Svensson, 2000). In the context of REDD+, increases in funding flows could enhance incentives to engage in corruption on the part of those receiving and benefitting from funds. REDD+ could thus become a type of 'political aid curse' (Altincekic & Bearce, 2014). Greater amounts of funding will increase the opportunities for corrupt behaviour by impeding the ability of donors and others to screen and monitor funding and projects, and the resources to distribute within a normatively corrupt context. Countries in which there is widespread impunity for corrupt behaviour, and in which there are few checks and balances on those responsible for using REDD+ funds, are even more likely to see cases of corruption in REDD+. Higher funding flows provide more resources to distribute within social relationships, and may even reinforce the idea that a biased distribution of revenues is the preferred way of doing things since more people will benefit from redistribution under conditions of greater resource availability.
Whether a resource-curse based logic can explain corruption in REDD+ funds relies on a few assumptions: that REDD+ funding is as fungible, non-conditional, and constant as resource rents are (Altincekic & Bearce, 2014). Rather, while REDD+ funding may eventually be constant since the need for climate mitigation is not likely to disappear within a short time-frame (unlike poverty), such funding is likely to come with strict conditions attached for its intended uses. Donors also have monitoring and evaluation mechanisms in place to try to ensure good use of funds, and thus the institutional context of REDD+ funding flows is likely to be somewhat better than for resource rents.

Corruption prevalence in the forest sector
Natural resources are in general often plagued by corruption, and corruption is an important if not the most important reason why resource-rich countries suffer from the so-called 'resource curse' (see Kolstad & Søreide, 2009;Leite & Weidmann, 1999;Robbins, 2000;Williams & Le Billon, 2017). For reasons of national economic and political security, governments often purposively limit public scrutiny of the way in which they manage resources and maintain a high level of discretionary power over key stages of resource extraction and production processes. Government entities normally have the exclusive authority to grant permits, licenses, and contracts for exploration and extraction, to collect revenues such as royalties and taxes, and to monitor the compliance of operators with environmental and other regulations. Such large monopoly and discretionary powers combined with the large revenue flows that resources generate create perfect conditions for corruption. Not only is REDD+ being implemented in countries characterised as having weak governance institutions, but it is also being implemented in a sector that has been highly affected by corruption, the forest sector (Brockington, 2008;Burgess, Olken, Hansen, Potapov, & Sieber, 2011;Remy, 2017;Sundström, 2016). Corruption is a major facilitator of increased deforestation and poor forestland management practices (Sundström, 2016). Transparency International's 2011 Bribe Payers Index ranked forestry as the fifth most corrupt sector out of 19 sectors.

Corruption risks and safeguards in REDD+
Since its inception in 2007, scholars and policy-makers flagged corruption as a potential impediment to the achievement of REDD+ goals (see Bofin et al., 2011;Brown, 2010;Dermawan et al., 2011;Larmour, 2011;Tacconi et al., 2009). Corruption is, in theory, likely to affect REDD+ in two ways. First, existing broad patterns of corruption in implementing countries are likely to shape individual behaviour within REDD+ programmes. Second, REDD+ will involve sizable fiscal transfers under relatively under-tested aid modalities to countries scoring poorly on formal quantitative indicators of corruption. REDD+ will also enhance the value of forestland. The combination of large amounts of money, weak institutions for channelling financing, and a broader context of poor governance have been viewed as likely to enhance both incentives and motivations for corrupt behaviour (Brown, 2010).
Scholars identified hypothetical corruption risks for REDD+, that is corruption-related factors that could undermine the achievement of REDD+ goals during the first years of its existence. As summarised in Table 1, these hypothetical risks span a range of corrupt practices (that is bribery, collusion, fraud, embezzlement, extortion, kickbacks, nepotism) at a variety of scales (that is within projects, institutions, funding modalities, and polities). Table 2 compares these hypothetical risks with corruption risks for REDD+ reported via our three Southeast Asian cases. Brown (2010), for instance, noted that opportunities for manipulating baseline forest and carbon data, double-counting project benefits, and verification and monitoring problems were the most likely trio of corruption challenges. Data quality as well as benefit sharing are likely to be areas of high corruption risk, given that REDD+ is ultimately intended to result in performance-based payments that change individual-level behaviours. Tacconi et al. (2009) argued that actors with economic interests in continued deforestation (for example agricultural or timber conglomerates) could pursue bribery to undermine REDD+ schemes. Pham et al. (2013) identified a number of corruption risks in REDD+ programmes, including in the collection and distribution of revenues and financial benefits; the existence of unclear, overlapping, and weakly enforced regulations; and non-transparent, non-accountable decision-making over forestland licenses and titles.
In addition to focusing on hypothetical corruption risks in REDD+, a few scholars have attempted investigation of empirical cases of corruption in similar forest conservation schemes (Cavanagh, 2012(Cavanagh, , 2017Mrema, 2017). Cavanagh (2012) uses a case involving the World Bank/Global Environment Facility, the Ugandan Ministry of Trade, Tourism and Industry (MTTI), and the Uganda Wildlife Authority (UWA) to illuminate how corrupt processes unfolded in the context of Uganda's Protected Areas Management and Sustainable Use (PAMSU) project, from 2002-2010. An inquiry led by a retired Supreme Court Judge found widespread irregularities in UWA's use of PAMSU funds, including 'ghost employees' paid for services and international travel that were never undertaken (Cavanagh, 2012). Although an independent audit intended to safeguard UWA's use of PAMSU funds, the auditor Table 1. Main hypothetical corruption risk areas and behaviours identified in early literature on REDD+ • Manipulation of baseline forest and carbon data (fraud) • Double-counting of project benefits (fraud, embezzlement, nepotism) • Misuse of benefit sharing and other revenue flows (embezzlement, kickbacks, bribery, extortion) • Manipulation of weak monitoring and verification systems (fraud, collusion) • Undue influence to continue activities that involve deforestation (bribery, nepotism, collusion) • Unclear, overlapping, and weakly enforced regulations (embezzlement, kickbacks, bribery, extortion) • Non-transparent, non-accountable decision-making about forest resources Sources: Tacconi et al. (2009), Brown (2010, Larmour (2011), Pham et al. (2013). testified that field-visits to project sites were not undertaken and that audit reports were instead collated using internal MTTI-UWA-World Bank reports (Cavanagh, 2012). In light of the corruption risks, seven safeguards for REDD+ were developed under the auspices of the UN-REDD Programme, three of which relate specifically to issues of governance, including corruption. These are: (i) ensuring transparent and effective national forest governance structures, (ii) ensuring respect for the knowledge and rights of Indigenous Peoples and members of local communities, and (iii) ensuring full and effective participation of relevant stakeholders (Program, 2010). UN-REDD has also established guidelines for Measurement, Reporting, and Verification (MRV) of REDD+ activities, offering technical support and advice to countries in setting-up their MRV systems, although the precise nature of these systems remains the prerogative of national governments and a recent review notes progress may not be promising (Fischer, Hargita, & Günter, 2016). These safeguards touch on standard concerns in public aid donors' operational interventions for anti-corruption, which have mostly adopted a principal-agent perspective for corruption's existence (Marquette & Peiffer, 2015). The problem with such principal-agent interventions is that they often do not sufficiently take into account the role of pro-corruption social norms in undermining principal-agent accountability regimes. Capacity building, technical training, and institutional reforms intended to strengthen REDD+ safeguards may offer principals rent-seeking opportunities (for example per diem abuse) and enable them to present a façade of good governance while, in reality, they are still part of the corruption challenge (Moene & Søreide, 2015;Søreide, Tostensen, & Skage, 2012). The core question we address is how effective are REDD+ safeguards likely to be when it comes to reported corruption risks in specific pilots given our theoretical perspective that corruption may exist both because it is difficult for principals to fully monitor agents and because corruption is underpinned by prevailing norms that serve certain societal functions.

Methodology
We adopted a qualitative case study approach to answer our questions of corruption risk types and mitigation measure efficacy for several reasons. Our method of conducting structured, focused, comparative case studies allows us to identify corruption risks that are common across contexts as well as those that are particular to a given setting. Such an approach allows us to identify the causal pathways by which corrupt behaviours are likely to emerge, and to examine the likelihood that proposed or actual mitigation measures will be able to prevent or reduce corruption. Case analysis allows us to analyse the interests and incentives of various actors in the REDD+ ecosystem, and the relationships between these actors that can sustain unethical behaviours.
We selected our three Southeast Asian cases, Indonesia, the Philippines, and Vietnam, based on the following criteria. First, these three countries are located in the world region (East Asia and the Pacific) with some of the largest producers of greenhouse gas emissions and which receive the highest amounts of climate finance globally (Rosenberg et al., 2014). This means that 'getting it right' in terms of tackling corruption in REDD+ within these countries is critical for the overall effectiveness of global climate actions. Second, REDD+ readiness activities had existed in each country for two or more years by the time we started our research, meaning we could draw on evidence from actual pilot projects and activities. Third, existing policy (and in some cases academic) literature on national or subnational corruption risks for REDD+ was available, allowing our research to depart from already identified concerns and identify emerging risks. Fourth, REDD+ readiness activities were part of an official national domestic framework supported, at least in part, by bilateral or multilateral aid financing, implying both the potential for significant future financing and supposed adherence to a common set of 'safeguards'.
Data collection occurred in the first half of 2014 in Indonesia and the Philippines, and in 2015 in Vietnam. For each country, this consisted of three elements: (i) a literature review; (ii) qualitative, semi-structured interviews with key informants at national and sub-national levels as well as focus group discussions with relevant stakeholders; and (iii) primary and secondary document collection and analysis. For interviews with national-level key informants, we identified individuals through the literature review and a mapping of relevant REDD+ institutions in each country, including government officials, academics, and civil society groups. We approached these identified informants directly via email, telephone, or through an office visit. We relied on snowball sampling to reach additional informants, using information provided by key informants to contact newly identified ones. REDD+ pilot areas were identified through country documentation, and we targeted involved communities, local government officials, and locally based academics and civil society groups for participation in interviews primarily through snowball sampling methods. Aware of the methodological challenges in reporting on stakeholder perspectives of REDD+ (see Atmadja & Sills, 2016), we applied a common research protocol across the three countries. The protocol contained five central and 22 subsidiary research questions that were to be answered, as far as possible, through data collection for each country. A copy of this research protocol is found in Appendix Table A1, while an anonymised list of interview and focus group participants is found in Appendix Table A2.
Interviewees received a printed project information sheet in advance of interviews specifying that their remarks would not be attributable to them as individuals. Interviews and focus group discussions proceeded only with the express consent of participants. Triangulation occurred for each country case through comparing existing policy and academic literature with interview findings and via focus group discussions or online surveys.
The literature review for each case study consisted of systematically searching online databases for all relevant academic works, reports, and policy documents that have been published on three issues: REDD+ policies and institutional arrangements in each country at both national and sub-national levels; government reforms; and governance issues in forestry, natural resource management, and REDD+. The literature reviews were additionally used to inform further data collection via the individual and focus group interviews.
Several adjustments were made while implementing fieldwork. The Philippines case analysis originally intended to gather data from three REDD+ pilot sites in Southern Palawan, Southern Leyte, and Quezon province during November 2013 to January 2014. Typhoon Haiyan, however, struck the provinces of Leyte and Samar on 8 November 2013, with the northern municipalities of Palawan also affected. Fieldwork in Southern Leyte became impractical as a result and a focus was placed instead on Southern Palawan and Quezon. During data collection it became apparent that conducting focus group discussions was potentially dangerous in the Philippines. Given the range of actors involved in corruption, from politicians and officials to neighbours and family members, interviewees felt more comfortable speaking with project researchers individually and privately. Nine one-on-one interviews were conducted with key stakeholders in government, civil society, and academia.
In Indonesia, individual key informants from national-level institutions involved in implementing REDD+ were interviewed to discuss issues identified in the literature review. In contrast to the situation in the Philippines, focus groups were possible in Indonesia, and two large focus group discussions were held to discuss state losses in the forest sector due to forest-related crimes and confirm preliminary findings on corruption risks and relevant reforms. Focus group participants again came from national-level institutions involved in implementing REDD+.
In Vietnam, 15 semi-structured, in-depth qualitative interviews were conducted with key REDD+ stakeholders at the national and sub-national level. The interview data was supplemented by a literature review and an online survey soliciting REDD+ stakeholder views on potential corruption risks and suggestions for how to respond to these risks in national REDD+ implementation via explicit anti-corruption strategies. The survey consisted of 23 questions distributed via the Vietnam REDD+ Network, comprised of over 200 individuals from 56 organisations. However, the number of responses to this survey was too small (just 30 responses were received) to be representative and the results only supplement insights from the qualitative interviews.
Will REDD+ safeguards mitigate corruption? 2135 6. Qualitative case analysis: corruption risks and REDD+ readiness in three Southeast Asian contexts

Indonesia
At the September 2009 G20 meeting in Pittsburgh, then Indonesian President Yudhoyono committed to reduce carbon emissions by 2020 through domestic and foreign financing. Emissions from the forest sector constitute a large proportion of Indonesia's carbon emissions and REDD+ quickly became a cornerstone of the country's forest sector. The Indonesian government set up regulatory frameworks and established an agency to manage REDD+ activities (although responsibilities for REDD+ recently moved to the newly named Ministry of Environment and Forestry). A measurement, reporting and verification (MRV) system was in place by the time of fieldwork, while a Constitutional Court decision had set out procedures to clarify the status of Indonesia's forest estate: both developments considered to provide solid foundations for REDD+ implementation.
The rate and causes of deforestation in Indonesia have been the subjects of longstanding discussions (see Sunderlin & Resosudarmo, 1996;Hansen et al., 2013;Margono et al., 2014). Discrepancies exist between Ministry of Environment and Forestry statistics and those contained in independent studies. Margono et al. (2014) find, for example, a deforestation rate for the period 2011-2012 of about 0.85 million hectares, but this has been contested by the ministry which argues that differences in methodology and forest definitions explain their lower deforestation figures for the same period. There is broad agreement that the proximate drivers of deforestation are logging, forest fires, the development of timber plantations and estate crops, and the expansion of mining activities (Indrarto et al., 2012;Hosonuma et al., 2012;Abood, Lee, Burivalova, Garcia-Ulloa, & Koh, 2014), though the relative importance of each driver tends to vary across studies.
Several surveys indicate corruption as a continuing challenge in Indonesia (CPI, 2013;Republik Indonesia, 2012). A UNDP survey on forest, land, and REDD+ governance in the country shows governance to be generally weak, with lower scores at provincial than at national level (Situmorang et al., 2013). The then Ministry of Forestry identified the following areas of the forest sector as being at risk from corruption: licensing, misreporting on data of forest product harvests and trade, manipulation of taxes or other charges, and manipulation of auctions of confiscated timber (Kementerian Kehutanan, 2014). For REDD+, the following corruption risks were in the literature and emerged from our interviews: collusion in initial REDD+ policy-making, collusion in the coordination of processes among multiple implementation actors, fraud in relation to the benefit-sharing mechanism and in relation to the REDD+ accounting system (Dermawan et al., 2011). Outside the REDD+ 'value-chain', corruption risks also arise from weak or uneven law enforcement of rules and regulations in the forest sector.
A new policy for which REDD+ provided impetus, the so-called 'one map' initiative, has aimed at aligning cartographic resources available to public agencies through one national land-use map. Published online and regularly updated, the map has reportedly reduced the scope for competing interpretations of forestland use and improved the potential for multi-agency coordination in REDD+ implementation. Yet, the 'one map' policy has not been matched with countrywide improvements in forest data generation, and such data is still often missing or inaccurate, complicating reporting procedures, tax accounting, and the calculations necessary for determining losses to the state in court-cases of forest corruption.
In terms of mitigating corruption risks in licensing and benefit-sharing calculations, the then Ministry of Forestry established an online licensing system providing figures on numbers of permits issued. Interviewees state that it is still cumbersome, however, to get an overview of whether a particular firm has fulfilled different requirements from different public agencies, for example reporting on tax revenues from timber. To help improve law enforcement approaches to forest crimes, the respective ministers of finance, forestry, and environment, along with the attorney general, the national police chief, and the head of the national Transaction Reporting Analysis Centre, all signed a memorandum of understanding in 2012 to pursue a 'multi-door approach' to combat forest sector criminality, including corruption. Yet the domestic anti-corruption body (KPK) has pointed to the involvement of high-ranking law enforcement officers and judges in more recent corruption cases as reasons to suspect forest-related corruption cases may not get the hearing they deserve (Kompas, 2014).
There is evidence to suggest that nepotism, collusion and other forms of corruption continue to shape patterns of forest use, particularly around the time of provincial elections (Burgess et al., 2011). Corruption is a frequently voiced concern of domestic commentators and various government initiatives point to the tenacity of pro-corruption social norms. At the same time, visible steps are improving monitoring and accountability regimes for REDD+ specifically, and for the forest sector in general. This points to a possible domestic political-bureaucratic struggle between, broadly, those in favour of strengthening principal-agent type control mechanisms, and those that wish to maintain pro-corruption norms because they benefit from illicit wealth extraction from the country's forest estate.

The Philippines
The Philippines is rapidly losing its remaining forest cover, as indicated by the country's fourth national report submitted to the Convention on Biological Diversity (2009). Media reporting of massive illegal logging is commonplace and disturbances from forest fires extensive (Republic of the Philippines, 2009). Formulation of a national strategy on REDD+ began in 2009 when a network of NGOs established a partnership with the Department of Environment and Natural Resources (DENR). This national strategy (dubbed PNRPS) became part of the country's National Climate Action Plan (NCCAP) in 2011 as well as the Philippine Development Plan for 2011-2016. A National Multi-stakeholder REDD+ Council is envisaged as the primary coordinating and implementing body, with membership drawn from the DENR, but also other public agencies, indigenous' peoples groups, CSOs, and academia. This multi-stakeholder body was, at the time of fieldwork, not yet formally constituted, however.
Beyond global corruption indices on which the Philippines scores poorly, national analyses of corruption have stated it is 'systemic, political and grand in nature' with 'powerful elites who mould and direct policies that favour specific private interests' ( Transparency and Accountability Network, 2011, p. 15). The Philippine Development Plan 2011-2016 notes corruption and a lack of accountability are rooted in the 'country's historically evolved political processes and traditions', characterised by the dominance of elite interests, a weak bureaucracy, and corruption in elections through patronage and money politics (National Economic Development Authority Philippines, 2011, p. 213). In the country's forest sector, two overarching corruption issues identified by Mayo-Anda (2011) are: (i) bribery and manipulation of land classification documents; and (ii) extortion and bribery by officers of the Department of Environment and Natural Resources (DENR), by the military, and by police as part of transaction costs for the harvesting, transportation, and sale of forest products. Van der Ploeg, van Weerd, Masipiueña, and Persoon (2011) provide an example of the latter from the Northern Sierra Madre National Park in Isabela province: tracing the production chain of an illegal logging operation (financed by prominent businessmen and run by village-level officials) they found that forest guards were bribed to allow illegal timber to pass highway checkpoints.
Earlier literature, including a corruption risk assessment commissioned by the UN-REDD Programme (UN-REDD and Ateneo School of Government, 2013), and our interviews suggest REDD+ projects are likely to encounter a range of corruption risks during implementation. Improper political influence on REDD+ via the misuse of official resources (financial, human, assets) was viewed as the risk most likely to occur, while fraudulent reporting of the results of reforestation efforts was considered the highest impact risk, given the potential for undermining national climate change mitigation goals. The potential for diversion of REDD+ financing was also classified as a high financial management risk.
Developing and ensuring the implementation of legal, social, and environmental safeguards for REDD+ has been a key strategy under the PNRPS and draft REDD+ safeguard guidelines have been prepared with assistance from the UN-REDD Programme. Corruption is included as a 'governance risk' in these safeguards, with six subsets of risk identified: abuse of authority, bribery, patronage, nepotism, misrepresentation, and the abuse of funds (Safeguards Technical Working Group, 2013).
Our case interviews show that, in the Victoria-Anepahan Range REDD+ Pilot in the municipalities of Narra and Quezon, Southern Palawan, land use conflicts, environmental crimes and graft reportedly abound. Community members allege bribery helps in evading environmental permits, while DENR and local government checkpoints are colloquially named 'cash points' given they are focal sites for bribe extractions (similar to the practices described by Van der Ploeg et al., 2011). Interviewees noted similar issues in the General Nakar REDD+ Pilot in Quezon province. Here, local-level elite capture was alleged to take several forms, notably favouritism towards certain families in the processing of environmental permits. Acts of bribery were alleged by some community members to constitute 'standard operating procedure', although not all interviewees agreed with such perceptions.
Our findings suggest a mix of principal-agent and normative corruption problems in the Philippines' REDD+ pilots. Members of public agencies tasked with implementing and monitoring REDD+ safeguards are reported to engage in various forms of corruption, suggesting both weak principalagent accountability relationships and the continued prevalence of pro-corruption social norms. Although detailed corruption risk assessments for REDD+ and specific anti-corruption measures have been identified at national level, these did not appear to translate into meaningful accountability outcomes at the pilot sites we studied.

Vietnam
Approximately USD 72 million has been committed to support REDD+ readiness activities in Vietnam since 2009 (Huynh, 2016) and the country has taken several steps towards establishing a national REDD+ programme since piloting began in the Lam Dong province. Quy and Can (1994) found Vietnam's forest cover to have declined from 43 to 27 per cent in the period 1943-1990, and the country was one of the first UN-REDD partner countries to move towards Phase 2 of its national REDD+ programme (which involves the implementation of a national REDD+ strategy with a view to receive results-based payments in a third phase). A measurement, reporting, and verification (MRV) framework has been endorsed by the National REDD+ Office, while a Forest Protection and Development Plan sets a target to increase forest cover by 45 per cent by 2020. Institutional arrangements put in place include the establishment of a National REDD+ Steering Committee, which is the ultimate decision-maker, chaired by the Ministry of Agriculture and Rural Development (MARD), but also including other ministries. A Vietnam REDD+ Office (VRO), based in the Administration of Forestry, coordinates implementation activities.
A 2012 national survey conducted by the World Bank with the Government Inspectorate of Vietnam (the country's anti-corruption agency), found that 72 per cent of citizens view corruption as a serious problem. Petty corruption and unofficial payments are particularly common, causing public dissatisfaction (World Bank and Government Inspectorate of Vietnam, 2012). A UNDP-supported survey (PAPI) has found that corruption is rated among the top five most serious socio-economic issues for Vietnamese citizens, with one in four viewing it as the most serious issue. 6 Several studies find corruption to be an endemic feature of the Vietnamese forest sector (Hoang & Do, 2011;McElwee, 2004;Sikor & To, 2011). Three main explanations for the high incidence of corruption in the country's forest sector are identified by AusAID (2000): (i) inconsistencies in legislation and policies that create gaps in which bribery can thrive; (ii) the slow operation of the 'one-stop-shop' office for forest licensing; and (iii) unclear division of duties and allocation of tasks among state officials.
Given that Vietnam has historically experienced weak anti-corruption enforcement, the UN-REDD Programme conducted a 'participatory governance assessment' for REDD+ in the country (UN-REDD, 2014). Corruption risk areas highlighted by this study were the benefit distribution system, forest law enforcement, and forest sector contracting. A further corruption risk assessment has been completed by the local chapter of Transparency International in Vietnam (Cao & Le, 2013), identifying seven areas of REDD+ at severe risk of corruption: (i) design and development of the national REDD+ strategy (PRAP); (ii) land use planning and land use allocation for REDD+; (iii) allocation of funds for internal capacity building to ministries and agencies; (iv) allocation of finance to ministries and agencies for REDD+ implementation; (v) allocation of finance to ministries and agencies; (vi) registration of projects; and (vii) procurement of goods and services.
Contrary to Cao and Le's (2013) assertion that corruption risks linked to the development of the PRAP are severe, the majority of our case interviewees did not agree that individuals and groups would seek to influence PRAP design in order to benefit private interests or to entrench political and economic power. One interviewee from an international organisation noted that the manner in which the PRAP process was conducted reduced possible corruption risks. The PRAP process in Lam Dong province, for example, identified interventions via a series of workshops using participatory stakeholder analysis as well as spatial analysis of forest resources. At the same time, other interviewees suggested that stakeholder analyses of this type were still strongly influenced by national government actors, with limited involvement of local citizens at pilot sites. In terms of corruption risks in forestland allocation, provincial and district agencies in Lam Dong province have asserted that private firms investing in small hydropower unit construction and rubber plantations have strongly influenced provincial politicians in order to gain access to forestland. Da Chay and Da Sar communes were particularly concerned with a lack of transparency and participation in project decision-making processes. Just over a third of our case interviewees (mostly from international organisations) noted that corruption risks relating to rightful landowners not receiving a fair proportion of benefits due to inadequate consultation were high. This was interpreted by some interviewees as relating to the absence of clear goals and effective consultation mechanisms. Others (mostly non-state actors) viewed inadequate consultation in turn as the result of power and resource concentrations.
Our findings reveal attempts in Vietnam to address specific corruption risks for REDD+ identified in earlier studies. Some aspects of participatory approaches introduced by international organisations in collaboration with both governmental and domestic non-governmental actors were considered to reduce certain corruption risks. The types of anti-corruption activities being implemented (social safeguards, grievance, and fund management mechanisms) reveal attempts both to strengthen principal-agent type interventions and to promote participation as a norm within policymaking processes. Yet, examples of lobbying at the interface between the private sector and politicians at provincial level, as well as limited involvement of local citizens in certain stakeholder processes, point to competing pro-corruption social norms as well as the relative superficiality of some REDD+ consultations. 7

Conclusion
It is difficult for external actors to offer the right incentives for developing countries to limit deforestation, particularly where economic drivers (such as commerce in agricultural commodities and mining) pushing deforestation are strong and in some cases growing (Leblois, Damette, & Wolfersberger, 2017), and incentives to engage in various forms of corruption in REDD+, as identified in earlier literature, are strong. Our findings confirm that the safeguards put in place by actors in official development cooperation to mitigate identified corruption risks, at least as these safeguards have recently been implemented in the Philippines, Indonesia, and Vietnam, are still largely based on principal-agent type accountability mechanisms. At the same time, our case analysis shows that procorruption social norms are part of the landscape at REDD+ implementation sites in all three countries. It appears that REDD+ safeguards have not integrated lessons from earlier research on anti-corruption intervention effectiveness (for example Marquette & Peiffer, 2015;Persson, Rothstein, & Teorell, 2012), such as the need to explicitly unpack assumptions about the existence of 'principled principals' and to integrate an adaptive management approach so that the effectiveness of various safeguards can be regularly revisited. This may be part of broader explanations for why REDD+ has been more difficult to implement than expected (Angelsen et al., 2017;Lund, Sungusia, Mabele, & Scheba, 2017). We conclude that REDD+ safeguards need overhauling to mitigate the likely impact of procorruption social norms on the principal-agent type accountability mechanisms put in place to date.
Will REDD+ safeguards mitigate corruption? 2139 Our advice for colleagues working on improving REDD+ safeguards is to engage with the recent literature and recommendations on corruption risk management in aid interventions (for example Johnsøn, 2015;OECD, 2016). Further research is also required. In future, scholars might usefully extend the number of REDD+ implementing countries and projects scrutinised, while longitudinal studies may offer opportunities for testing the effects of different versions of REDD+ safeguards on corruption prevalence over time. Theoretical advances may be possible through further study of the interactions between principal-agent type accountability mechanisms and pro-corruption social norms at particular REDD+ sites. Notes 1. UN Sustainable Development Goal 13 is to 'take urgent action to combat climate change and its impacts'. On 22 April 2016, 175 UN Member States signed the Paris Agreement under the United Nations Framework Convention on Climate Change, aiming to reduce the pace of climate change. 2. A regional and country overview of activities is provided by the UN-REDD Programme at: http://www.unredd.net/index.php? option=com_unregions&view=overview&Itemid=495. 3. NICFI pledged USD 1 billion of Norwegian government money to REDD+ programming in both Indonesia and Brazil in 2010. 4. Criticisms of REDD+ relate to broader scepticism towards performance-based aid allocation systems, of which REDD+ represents one type (see McGillivray and Pham 2017). 5. In Indonesia, for example, a multi-stakeholder REDD+ Taskforce was created, later leading to a National REDD+ Agency, the first cabinet-level REDD+ body to exist in any country. The agency was discontinued in 2015 with Indonesian Presidential Decree No. 16. Responsibility for REDD+ in Indonesia rests at the time of writing with a sub-directorate of the Ministry of Environment and Forestry (see http://www.redd-monitor.org/2015/01/30/indonesias-decision-to-put-the-reddagency-in-the-ministry-of-environment-and-forestry-is-not-in-accordance-with-norways-us1-billion-redd-deal). 6. PAPI is the largest time-series national governance and public administration survey in Vietnam based exclusively on citizen experiences. The index was first piloted in three provinces in 2009, and was extended to 30 provinces in 2010. By 2011 it covered all 63 provinces (see http://papi.org.vn/eng/). 7. Implementation challenges in REDD+ consultation processes, such as the Free, Prior, and Informed Consent (FPIC) mechanism, reportedly abound. Fontana and Grugel (2016) report from Bolivia that FPIC there has sometimes embedded existing social, cultural, and economic tensions.

Disclosure statement
No potential conflict of interest was reported by the authors.