Modest forest and welfare impacts from current REDD+ initiatives

doi:10.21203/rs.3.rs-2429873/v1

Download PDF

Analysis

Modest forest and welfare impacts from current REDD+ initiatives

https://doi.org/10.21203/rs.3.rs-2429873/v1

This work is licensed under a CC BY 4.0 License

Version 1

posted

You are reading this latest preprint version

Reduced emissions from deforestation and forest degradation (REDD+) promise to deliver performance-based, cost-effective climate change mitigation. 15 years after REDD+’s conception, we analyse the rigorous counterfactual-based ev-idence for environmental and welfare effects from national and subnational initiatives, along a REDD+ Theory of Change. Using machine-learning tools for literature review, we compare 32 quantitative studies including 26 primary forest-related and 12 socioeconomic effect sizes. Average environmental impacts were positively significant yet moderately sized, comparable to impacts from other conservation tools, and mostly impermanent over time. Socioeconomic impacts were welfare-neutral to slightly positive, especially at outcome stage (e.g. rising incomes). Moderator analysis shows that REDD+ environmental additionality was likely restricted by project proponents’ ‘high-and-far’ spatial targeting of low-threat areas (adverse selection bias). Disappointingly scarce funding flows from carbon markets and ill-enforced condi-tionality probably also limited impacts. Hence, important policy and implementation lessons emerge for boosting effec-tiveness in the current global transition towards larger-scale, jurisdictional REDD+ action.

Earth and environmental sciences/Environmental social sciences/Climate-change mitigation

Earth and environmental sciences/Environmental social sciences/Environmental economics

In 2003, at the United Nations Framework Convention on Climate Change (UNFCCC) 9th Conference of the Parties (COP9), researchers from Brazil and the USA launched the notion of “compensated reduction”: tropical countries should be ex-post rewarded for reducing their national forest loss below a pre-agreed baseline.¹ The European Union’s (EU) Joint Research Centre also linked national forest-cover baselines to possible compensations between countries.² Simultaneously, a high-level review of the economics of climate change concluded that for US$5–10 billon, two thirds of global deforestation could be ‘bought out’, thus curbing one major source of global greenhouse gas emissions at low costs.³ Since much land clearing in the forested hinterlands of the Global South only provides marginal economic returns, conservation opportunity costs there often remain modest, so allegedly they could be compensated or ‘bought out’ rather cheaply. The conceptual scope was later broadened towards an all-inclusive term of political consensus: “reduced emissions from deforestation, forest degradation and the role of conservation, sustainable management of forests and enhancement of forest carbon stocks in developing countries” (short: REDD+).⁴

REDD + would basically work as an international multilevel system of conditional, performance-based payments for environmental services (PES).^5,6,7 In this global architecture, carbon markets would mobilize funding, while recipient national governments would incentivize on-the-ground landholders and forest-dwelling indigenous populations, invest in economic alternatives, capacity building, and improve protected areas – thus delivering the enabling conditions for achieving emission reductions on the ground.⁸ Hence, REDD + as a model of intervention is usually associated with global performance-based forest-carbon funding, but implementation is de facto an umbrella term for a broad mix of ground-level initiatives, designed in contextually customized ways.

A swath of local-level REDD + projects has been implemented since the COP13 Bali Action Plan in 2007.^9,10 Across governance scales, “nested approaches” were to resolve issues of attributing carbon credits between projects, subnational programmes, and the national level, including to avoid double-counting.¹¹ REDD + came out strengthened from the UNFCCC Paris Agreement (UNFCCC 2015). 377 REDD + projects cover 53 million hectares in 56 countries, based on CIFOR’s International Database on REDD + Projects (ID-RECCO).¹² These projects were to avoid some 1% of annual forest-based emissions, but quite recently only 5% of their carbon credits had been sold¹⁰.

Across the Global South, Brazil (48), Colombia (33), and Peru (25) are project-leading; the density of REDD + implementation (project area/national forest area) is highest in Kenya, Nepal, Central America, and the Andes region (Fig. 1); conversely, REDD + implementation is negatively biased against Central Africa and South Asia. Our map also clearly depicts the nesting challenges of REDD + credits between overlapping national, subnational, and project scales of action (e.g., in Brazil). Finally, mapping our 32 quantitative study cases (red-coloured; for sample selection, cf. Methods—case sources are listed in the Supplementary Information, SI) also points to some light research imbalances: compared to REDD + implementation, only few rigorous impact evaluations have been done in Asia, Africa, Colombia, and Mexico.

Rather than carbon markets assuming the hoped-for lead role in financing large-scale jurisdictional-level implementation, REDD + has remained ‘project-ified’, with bilateral or conservation donors financing only incipient actions¹³. Bilateral donors and the UN-REDD programme implement so-called REDD+ “readiness” preparatory processes (increasing forest monitoring capacities, analysing deforestation drivers, etc.); 50 + countries have created national REDD + programmes.¹⁴

The originally envisaged model of REDD + large-scale national implementation has in practice only in a few countries advanced towards large-scale conditional payments. Those comprise notably Norway’s International Climate and Forest Initiative (NICFI), launched as early as 2007,⁵ followed by Germany’s REDD Early Movers programme.¹⁵ More recently, two multilateral organizations started piloting large-scale, results-based payments: the Green Climate Fund (GCF) and the Forest Carbon Partnership Facility (FCPF), as well as efforts to mobilize public-private blended funding under the LEAF coalition.

A sometimes-harsh debate has accompanied REDD + projects. Strong criticisms of REDD + processes and impacts have focused inter alia on problems related to social inclusion, indigenous rights, and other welfare outcomes.^16,17,18,19 Conversely, more optimistic outlooks stressed the experimental nature of project-scale REDD + initiatives (with some encouraging outcomes), the time-consuming complexity of governance transitions, and the embryonic stage of the genuine national-level REDD + that largely remains untested.^14,20,21

At this stage, how much do we know about the on-the-ground successes or failures of REDD+? Impact evaluations should answer this question; they are becoming standard tools in many sectors.²² Following emphatic calls for solidly formalized empirical impact assessments also in environmental and biodiversity conservation,²³ the field of environmental impact evaluation has recently expanded, as evidenced by various reviews and meta-studies.^24,25,26

For REDD + funded initiatives, so far less stylized evidence is available. The literature-synthesizing work has been mostly qualitative, ranging from reviews of the REDD + literature,⁴ its governance challenges²⁷ and perspectives for future REDD + implementation¹⁴. More formal evaluations include lessons from early carbon project,²⁸ and first REDD + environmental and welfare effects.²⁹

Against this backdrop, this REDD + meta-study systematically takes stock of the currently expanding evidence, which required careful delimitations (cf. Methods). In our Web-of-Knowledge based literature search using text-mining algorithms, we target rigorous quantitative evaluations of the environmental and welfare-related impacts of REDD + interventions. This includes (corporate or NGO) projects, public programmes (e.g., national PES schemes with forest-carbon components), and a few bilateral jurisdictional agreements (e.g. Norway’s forest agreements with Guyana and Indonesia). Our focus is on avoided deforestation and degradation, rather than re-, afforestation, or restoration. Included evaluation studies sport impact estimates that can be scaled and ranked, i.e., effect sizes (and their precision) are comparable including across categories defined by relevant contextual and design variables. Finally, as called for in this journal,³⁰ we compare REDD + impact sizes to those of other forest conservation instruments. To our knowledge, no such analysis pre-exists in the REDD + literature.

1. Environmental impacts

We used the correlated hierarchical effects model with random effects for our impact calculations. In total, 32 quantitative studies (listed in SI) with 26 forest-related and 12 socioeconomic primary effect sizes fulfilled our data selection criteria (cf. Methods). Figure 2 shows a forest plot of our meta-regression results for comparable forest impact sizes from REDD + treatments (results fully reproduced in SI). In our sample, we only have one (insignificant) estimate directly for forest carbon—the primary goal and final impact of REDD + according to its Theory of Change. Most estimates are instead for forest-cover proxy outcomes (including absolute and relative forest loss) leading to these impacts, which can be more easily compared. Our mean overall estimated REDD + effect of 0.07 (95% Confidence Interval: 0.03–0.11) can be considered “small”. The Q-test indicates heterogeneity, meaning one can find true effects outside of this confidence interval. Overall, the positive significant estimate confirms modest forest conservation gains from REDD+.

Two intervention subgroups can be distinguished in Fig. 2: self-declared REDD + projects (commercial, NGO-led, or national—cf. upper panel) versus carbon-inclusive multipurpose conservation PES (public, mostly national programmes schemes—middle panel). We found no significant effect difference between the two (p = 0.57). The precision of estimates is lower among especially some of the smaller-sized projects. Even in public PES-for-carbon schemes, the same programmes evaluated in different studies reached quite divergent estimates – including studies carried out by the same authors (i.e. Arriagada et al. 2011 vs. 2012 both on Costa Rica’s PSA; Mohebalian and Aguilar 2016 vs. 2018, both on Ecuador’s Socio Bosque), seemingly reflecting both variations in output variables and in matching methods (cf. Methods). Of special interest would also be the performance of larger-scale, jurisdictional-level REDD+, given the ongoing implementation shift towards those. These results are moderately encouraging, with conservation impacts in Guyana (Roopsind et al. 2019), Indonesia (Groom et al. 2022) (both NICFI) and Amazonas, Brazil (Cisneros 2022) all being significantly positive, although the latter two very small-sized (all case references in SI).

Looking at the secondary impacts, i.e., indicators not directly comparable to forest-cover proxies (SI, Fig.S3), we observe here also some larger, more significant impacts, such as boosting tree-species richness, avoiding wildfire incidence or slowing forest encroachment. Yet, this extended picture remains variable, too. Notably, impacts on forest degradation, the second “D” in REDD+, are just like for deforestation small or statistically insignificant.

Finally, little is globally known so far about the permanence of REDD+, i.e. to what extent prospective conservation impacts last after the intervention has ended. Estimates across the three permanence studies (excluded from our REDD + effect sizes in the upper panels) originating from Uganda (Jayachandran et al. 2018), Colombia (Pagiola et al. 2016) and Brazil (Carrilho et al. 2022) differ somewhat, but all three coefficients are insignificant (Fig. 3, lower panel). The dominant pattern is thus that post-REDD + deforestation returns by default approximately to its pre-intervention speed, but without eliminating the temporary conservation and climate mitigation gains achieved during REDD + implementation.

2. Socioeconomic impacts

In our REDD + Theory of Change (SI, Fig.S2), the most important side-objective of REDD + is it to improve local people’s wellbeing. Figure 3 thus shows comparable socioeconomic effects from rigorous REDD + impact evaluations. Like for environmental impacts, these are divided between outcomes (changes in income, consumption, or asset holdings) and proper impacts: the self-stated subjective wellbeing, and changes herein, on behalf of REDD + participants and other residents. Also here, our outcome variables are very close impact proxies – and can often be verified more objectively.

Empirically, the two types of indicators perform differently (p = 0.03): at the outcome stage REDD + has a significant positive, welfare-improving effect of 0.09 (95% CI: 0.03–0.15), while at the impact stage the mean effect of -0.01 is statistically indifferent from zero. Hence, the good news is that REDD + on average tends to make benefit recipients slightly better off materially. The bad news is that this may not always boost self-perceived welfare. The few studies where outcomes and impacts are measured simultaneously (Arriagada et al. 2015; Montoya-Zumaeta et al. 2021, 2022) confirm this trend: the material benefits provided may come to fall short of community expectations, especially when these are ex-post assessed, when benefit flows have ceased. Thus, self-stated subjective wellbeing may also become a strategic vote of protest by local participants over REDD + benefit sharing.

Looking at other socioeconomic outcomes (SI, Fig.S4), we observe that several impacts for subgroups can come out negative, such as the subjective wellbeing of female REDD + participants (Larson et al. 2018). This is a reminder that modest average gains in material welfare from REDD + does not necessarily warrant equity or a do-no-harm principle: distributional and non-material effects may still create (objective or self-perceived) losers.

3. Contextual and design factors

We conducted a moderator analysis for potential hints about the implicit role different REDD + context and design factors in co-determining the above-assessed effect sizes. Hence, we plot conditional impact sizes against selected variables (Fig. 4, A-H). We used all available observations to account for within-study subgroup differences in the design variables. Figure 4 shows the estimated coefficients along with its 95% confidence intervals. Our number of effect-size observations is low for solidly exploring correlations, but we can set hypotheses for future research.

First, this includes baseline deforestation pressure (Fig. 4A,B,E,F), which for other conservation tools correlated positively with impact size:^26,31 intuitively, the lower ex-ante forest-loss threats are, the harder it would be to counterfactually demonstrate impacts. For REDD+, we confirm a positive relationship between threat and impact, statistically insignificant at national (4A) but significant (p = 0.057) at the zoomed-in scale of subnational deforestation pressure (4B) (cf. Methods for classification). Socioeconomically, low-threat REDD + might go along with higher welfare gains (e.g. through lower opportunity costs), as also indicated by the coefficient sign here, yet this correlation is insignificant (4E,F).

Turning to design factors, the PES literature indicates that providing differentiated, beneficiary-customized, rather than uniform benefits can boost environmental effectiveness.^32,33 Again, we find the expected coefficient sign, but the correlation is insignificant (4C). Yet, socioeconomic welfare improvements were significantly higher in programs with differentiated, rather than uniform benefits (p = 0.045). This indicates that social customization may be important in REDD + benefit-sharing strategies. As for the second REDD + design variable, spatially targeting lands with high density of/ high threat towards environmental benefits is again in the PES literature featured as key for additionality outcomes.^32,33 This is forcefully confirmed for our REDD + sample: spatial targeting for forest-carbon density and/or expected deforestation correlates strongly (p = 0.001) with environmental impacts (4D), while it does not affect welfare outcomes (4F).

4. Comparing with other conservation instruments

Finally, as emphasized in this journal, we know generally too little about the comparative performance across conservation instruments. ³⁰ In Fig. 5, we compare the normalized Hedges’ G effect sizes recorded for the two types of REDD + with those of pre-existing conservation instruments, such as other incentives, disincentives, and enabling actions–-drawing on previous studies.^26,31 As in Fig. 2, we accounted for dependent effect sizes from the same study by assuming a correlation of 0.8 and used robust variance estimation.

As for performance, the two REDD + subgroups (defined as in Fig. 2) compare fairly well to the other three instrument categories, in terms of mean effect sizes to protect forests (2nd and 3rd ranks, among five). However, there is also a large variability of REDD + outcomes underlying the rather small intervention samples. Consequently, no statistically significant differences between REDD + and any of the alternative conservation instruments could be found. Comparatively speaking, REDD + exhibited a middling, yet also changeable conservation performance.

Since 2007, the world has incipiently gathered experiences with REDD+, a tool designed to conserve and enhance forest carbon in non-Annex I countries (i.e. largely developing/ emerging economies, plus China), in exchange for economic compensations from the industrialized Global North. REDD + is an objectively desirable end (the goal of reducing forest-based emissions) but has equally become a controversial means of using ‘market-based’ international offsets to help accelerate climate change mitigation.

A broad range of REDD + pilot projects has thus emerged. Jointly, they planned to avoid 84 million tCO₂ annually (over 33 years of mean lifespan) of emissions, or “around 1% of annual emissions from deforestation, forest degradation, harvesting and peat fires in the tropics”.¹⁰ While potentially significant, leakage and credit performance apart only 5% of the correspondingly needed carbon credits have so far been sold on the voluntary market. In terms of de facto avoiding existing deforestation at scale, REDD + projects have thus been but a drop in the sea.¹⁰

Effectively, REDD + projects have thus been starved out by a grossly insufficient global willingness to pay for mitigating climate change. Uncertain funding prospects have also made many projects quasi-placeholders waiting whether funding flows would materialize, meanwhile keeping on-the-ground treatment intensities low.^9,14 In particularly, implementers have been hesitant to introduce PES-type of continuous compensations to landholders, since implementers currently cannot promise continuity.^34,35 Obviously, this has deteriorated the framework conditions under which REDD + projects were hoped to deliver efficient results.

Yet, this does not per se question the potential usefulness of REDD + projects in providing valuable pilot lessons for the hoped-for upscaling to jurisdictional REDD+. Above, we have taken stock of the experiences so far. We carefully delimited exactly which initiatives deserved to be labelled as REDD+ -- either by proponents, or by analysts. We have also screened which impact evaluations were sufficiently rigorous to ensure internal validity and deliver trustworthy results, based on realistic counterfactuals enabling credible causal attribution. With many new empirical studies emerging recently, our larger and geographically more balanced sample than in previous meta-studies should also increase confidence in our results.

We can thus shed some new light on the effectiveness and welfare implications of REDD + initiatives. As for forest-cover and carbon effects – the ultimate raison d’être – REDD + initiatives have had small-sized effects, similar to what other conservation instruments have (not) achieved.^26,31 This holds for both specialized REDD + projects/ programmes and cases where REDD + has been integrated into national PES programmes. When interventions stop, prior pressures tend to resume, yet typically without fully undoing REDD + gains (partial impermanence). Overall, given disappointing carbon-market flows, and the harsh critique against REDD+, environmental effectiveness does not look great, but probably still better than expected by many observers.

Nevertheless, for lack of uniform cost data we have above not compared cost efficiency parameters. The few available case studies with REDD + costs data point to highly variable, in some cases elevated transaction costs, but also declining with scale.^36,37,38 A move towards larger-scale jurisdictional REDD + programs may thus also push towards more mitigation ‘bang for the buck’.

On the socioeconomic side of local benefits provided – REDD+’s primary side-objective – our results on average portray small positive contributions to local livelihood outcomes (e.g. incomes, assets), yet insignificant impacts (e.g. subjective wellbeing). New incentive-based projects also tend to locally build easy-to-disappoint expectations regarding future benefit flows.³⁹ Customized, rather than equal benefit transfers seem to improve socioeconomic outcomes. Single cases apart, no evidence points to REDD + making local people systematically worse off: while not everybody locally may gain, a narrow outcome range from welfare neutrality to light livelihood improvements is most common.

But in raising ambitions, why have the environmental impact results of this highly innovative tool overall not been much better? Above we have pointed to several design factors, such as adverse selection bias and inadequate spatial targeting. Insufficient on-the-ground enforcement of contractual conditionality is, however, another common deficiency in REDD + implementation, as noted in several case studies: often implementers will prefer to tolerate land-use violations, safeguarding instead the social capital built with local communities (Brazil:⁴⁰ Cisneros et al. 2020; Peru: Montoya-Zumaeta et al. 2021 & Giudice et al. 2019). Globally, by tolerating exaggerated baselines of future deforestation, the bar for REDD + credits was set far too low; unsurprisingly, the majority of credits are non-performing,⁴¹ accelerating in turn a public fatigue with environmental offsets.

Another critical qualitative issue surrounds complexity. REDD + is to forest carbon what ICDP projects have been to biodiversity: an umbrella term under which a ratatouille of composite, heterogeneous interventions has gathered. Many REDD + initiatives are ‘ICDP-like’, in terms of using the same integrated, multifaceted approach: trying a bit of everything to satisfy multiple stakeholders and minimize risks of total failure. Unfortunately, the ICDP approach does not have a good record in terms of producing sizeable impacts.²⁶

After 2007, many pre-existing ICDP projects looking for fresh funding were remodelled as REDD + initiatives, producing an opportunistic self-selection bias. In Indonesia, for instance, many REDD + projects were implemented by biodiversity-focused organizations; the targeted forest areas were more biodiversity-rich than carbon-dense, and only about one quarter of the project areas was truly threatened by deforestation.⁴² In turn, many genuinely new private sector initiatives adversely targeted de facto low-threat areas: avoiding deforestation would here become a low-hanging fruit.⁴¹ Hence, many REDD + projects may have served more as a proof of concept than as a real test of whether the approach is holding water.

Arguably, it is no shame for pilot projects to underperform or fail, if useful lessons are learned for future initiatives. Did early REDD + interventions thus maximize this learning and upscaling potential? Hardly so, mainly since too many projects were carried out in ‘high-and-far’, low-pressure settings, thus not taking the bull by the horns. Particularly the ICDP-type model was also too complex in design and transaction cost-heavy in implementation to replicate at scale. Furthermore, REDD + implementers almost never facilitated impact evaluation through (semi-)experimental rollout of multiple design options of action. Hence, we rather stand back with many highly customized ICDP-like ‘boutique projects’, including multiple components of action; yet we know very little about how well these components worked, and why.

What about full national REDD + programmes, as an alleged future of upscaled REDD+? For now, only the NICFI programmes in Guyana (Roopsind et al 2019) and Indonesia (Groom et al 2022) have been evaluated, finding for both small yet significant forest-protecting impacts. The Guyana case is not without controversy though: deforestation especially from gold mining actually increased during NICFI support, but less so than was predicted by a synthetic matching model for a no-REDD + counterfactual, based on other comparable high-forest low-deforesting (HFLD) countries with significant mining sectors.

For future research, doing further analyses of larger-scale REDD + programs, be it NICFI or more recently GCF and FCPF, looks promising, but impact evaluation needs to be integrated early into programme design. These impact evaluation analyses should in turn not just provide average effect estimates, but equally be challenged to investigate heterogeneous impacts, telling causally plausible stories about where, how, and why REDD + might work.

Delineating the concept of REDD+

As we saw above, REDD + is typically seen as a prototype type of action (i.e., a means) that generically remains described exclusively by its outcomes of reduced emissions (i.e., an end). This is fundamentally different from other interventions; for instance, “protected areas” or “forest certification” that describe means not ends. Observers can thus conceptually come to confuse a model for action (the alleged market-based offsetting strategy of REDD+) with an expected final goal (of having forests store more carbon).³⁵

In this subsection, we thus explicitly walk through the typical stages and assumptions underlying a REDD + intervention, using a Theory of Change (ToC) approach, designed for causally linking the stages of inputs, treatments, outputs, outcomes and impacts.⁴³ Figure S2 (SI) outlines these stages going from left to right, with key critical assumptions flagged in bubble shapes.

As for inputs, REDD + is directly triggered by, and thus essentially dependent on the presence of external finance flows, be it from global markets for carbon credits (as originally envisaged), or from donors: bilateral development or environmental donors (such as Norway’s NICFI), multilaterals with a climate mandate (e.g. the Global Environment Facility or the Green Climate Fund), and private-sector non-market flows for direct emission offsets, based on some notion of corporate social responsibility.

However, serious claims for REDD + achievements can eventually only be made if knowledge about pre-existing carbon stocks, land-use trends, key drivers and stakeholders triggering forest loss (and protection) jointly can be merged into a credible baseline: what would have happened under the laissez-faire baseline assumption of ‘no REDD + intervention’? Notably, a proper assessment of levels/ changes in threat is quintessential: if threats from deforestation drivers are rising, treatment may have to be intensified. Conversely, if a projected threat was not to materialize at all, then neither the dynamic counterfactual nor the project will exhibit any deforestation.

REDD + treatments are highly heterogeneous in their composition. We thus distinguish between the subcategories of incentives, disincentives, and enabling measures.⁴⁴ First, invariably some incentives are present in REDD + as a general local benefit-sharing mechanism, or compensation for the opportunity costs of newly introduced/ enforced restriction in forest use or conversion to alternative land uses. Incentives can either be conditioned upon compliance with certain land-use rules (e.g. PES-type of contracts), or unconditional investments into alternative, environmentally more benign livelihoods, social sectors (health, education), etc. Often, REDD + interventions also entail disincentives, through newly introduced restrictions or a more thorough monitoring and sanctioning of incompliance with already existing ones. Typically, REDD + has thus included both carrots and sticks. Third, enabling measures as a residual category include tools such as the free prior informed consent (FPIC) of local people’s participation in REDD+, a clarification of land tenure and access rules, etc.

Many real-world REDD + projects and programmes, such as the Bolsa Floresta Programme in Brazil’s Amazonas State (Cisneros et al. 2019), or the Sustainable Settlements in the Amazon (PAS) project in the Transamazon region of Pará State (Simonet et al. 2019; Carrilho et al. 2022) have been using the full spectrum of conditional and unconditional incentives, disincentives, and enabling measures to reach their goals. They used these pilot interventions to experiment with different components, but also with an underlying belief that holistic, locally customized approaches carried higher probabilities of success, especially in market-remote, cash-strapped frontier regions. Unsurprisingly, many REDD + projects are in their holistic range of actions ‘ICDP-like’, with a predominant focus on non-conditional livelihood enhancements.^9,34 For the same reason, REDD + projects have also had much to learn from ICDPs.⁴⁵

Public PES programmes with a partial focus on forest carbon goals constitute a second type of intervention. Often, carbon financing has helped to boost the funding of these national-level, or at least larger-scale programmes. Costa Rica’s PSA, Peru’s National Forest Conservation Programme, and Ecuador’s Socio Bosque all constitute such examples, although the latter two combined PES with ICDP components (Giudice et al. 2019; Jones and Lewis 2015). Hence, with forest carbon enhancement for climate change mitigation being flagged as an explicit goal, these PES-based interventions need to be included as another pathway of implementing REDD+.

Outputs are to be understood as the immediate, often short-term results of the ‘treatment’: the treated recipients understand the goals and modalities of the intervention, the rules of the game (incl. land and resource tenure) are clear, and (dis)incentives are well-applied. Delivered outputs imply that stakeholder motivations have been successfully aligned with the goals of the intervention. For this to occur, treatments need to have been well-designed and carefully implemented. From the PES literature, we know that spatial targeting in the selection of participants and their to-be-treated land areas has been important, vis-à-vis two complementary dimensions: a) the site-specific environmental service density (here: forest carbon stocks per hectare), and b) the on-site projected threat (here: of deforestation/ degradation) of that stock to be endangered over time. Also, customization of the benefits (e.g. multiple payment levels) can help making the intervention more cost-effective and equitable.^32,33

The outcome level is where the REDD + rubber hits the road: do critical stakeholders make the required behavioural on-the-ground changes? That is, do they reduce forest clearing, charcoal making, or timber harvesting in the REDD + required manner (environmental outcomes)? Similarly, do income, consumption, and perceptions of welfare and security increase among those targeted stakeholders (socioeconomic outcomes)? These are measurable indicators that can also be included in impact evaluations.

The final transition towards impacts – the overarching primary carbon-related goal of reduced forest-based emissions, as well as ethically and politically important side-objectives related to biodiversity, human wellbeing and equity, indigenous land rights etc. – still entails some subtleties.

First, a reality check is to what extent intervention-targeted stakeholders and deforestation drivers have been adequately aligned. For instance, many REDD + projects are focused on addressing smallholders to reduce their deforestation, but a local surge in land grabbing from more powerful external agents might render these efforts less fructiferous in terms of mitigating deforestation.

Second, income and consumption outcomes trigger development feedback loops on the final impacts. Rebound effects refer to the fact that treatment-induced changes in household incomes may also change consumption patterns (e.g., higher incomes stimulating meat and dairy consumption) that per se change ecological footprints. Magnet effects refer to the potential of these income changes to attract outside migrants, e.g. through successful employment creation in a REDD + project. Pull migration could have a bearing on land use, as migrants open up new land plots for subsistence agriculture. Both effects are well-established in the PES literature.³¹

Third, the goal of mitigating climate change is both universal and perpetual. Classical concerns vis-à-vis REDD + projects are thus to what extent these time- and place-bound interventions contribute to the hoped-for universal and perpetual impacts. As for permanence, the impact of a time-limited treatment on carbon stocks may also only be transitory – though as such still important for mitigating climate change in the short run. Conversely, to the extent the treatment triggers desirable structural changes at the output and outcomes level, permanence might be increased.

Likewise, a REDD + treatment may not only reduce on-site deforestation, but also push some of these pressures outside of the intervention area – a phenomenon known as leakage. This spillover effect will typically diminish, though not fully erase the (universal) REDD + mitigation impact.^46,47 The larger the scale of the REDD + intervention, the less leakage we should expect – a key argument for favouring national programmes over REDD + projects. The size of leakage in conservation incentive programmes is seldom quantified.³¹ For high-value products sold on international markets, such as harvesting precious timbers, leakage may be exceptionally high.⁴⁸ In general, the higher the price elasticity and the geographical mobility on output and input markets (incl. access to land), the larger leakage we should expect.³¹

Delimitation

As briefly mentioned above, our aim is to take stock of the currently available evidence from rigorous quantitative impact evaluations for REDD + interventions. This means that we need to apply a number of apriori filters of inclusion (cf. Table S1), related both to the underlying REDD + intervention (Factors 1–4), and subsequently to the case study evaluating its impacts (Factors A-F).

As “REDD + interventions” (1), we understand here, firstly, actions that implementers self-denominated using the RED(D) + label, and secondly, other actions that fully or partially featured forest-based climate mitigation/ carbon outcomes in an explicit way. As mentioned above, this would include also national-level PES programmes that pretend to further forest-carbon objectives; in turn, some large national-level watershed-focus PES programmes (e.g. in China, Mexico, and Vietnam) would be excluded. As for actions (2), many forest carbon programmes include both conservation/ regeneration of standing forests and afforestation/reforestation (A/R) activities; those focused entirely on A/R would not seem to functionally fit the REDD + definition, and we thus exclude them. In terms of scale (3), we choose to be inclusive of both subnational REDD+ (incl. projects) and emerging national programmes, keeping in mind they likely have different characteristics – cf. also (1). Finally, as a temporal cut-off point for the start of implementation (4), we use the year 2007, coinciding with the Bali UNFCCC COP13: everything that happened in the forest carbon sphere before 2007 (Joint Implementation, Clean Development Mechanism, etc.) is of comparative interest,²⁸ but is inevitably bound to differ from REDD+.

A second layer of filters refers to the analytical level. First, we choose to be inclusive with respect to the screened literature (A), incorporating not only peer-reviewed but also grey-literature studies – allowing us to include in a quickly moving field some more recent contributions (assessed by us as high-quality) that are still at the working paper stage. As for analysed impacts (B), we looked at both forest carbon (main goal) and welfare effects (primary side-objective). As “bottom-line”, we understand effects to be observed at the right-hand side of the REDD + ToC, i.e. both outcomes and impacts (see above). Impact evaluations are often stated in terms of outcomes (C), such as forest-cover (deforestation areas, rates) and land-use proxies (e.g. fire incidences), which are more precisely observable than forest carbon in the short-to-medium run. The more process-oriented, intermediary outputs (middle part of ToC) are not of our interest here (D): they are often less clearly (sometimes, ambiguously) linked to REDD + bottom-line outcomes, and are often more qualitative than quantitative. However, we included also subjectively stated wellbeing (“do you now feel better/worse off/ unchanged than prior to the REDD + project start?”), as a popular socioeconomic bottom line of evaluation (E). We realize though that it is an indicator with its own potential response biases, which is best triangulated with more objectively measurable socioeconomic outcomes.

The final filtering criterion is arguably the trickiest: the quality of impact evaluation (F). To rigorously attribute impacts to interventions, counterfactuals need to be constructed: what would have happened without the REDD + intervention? We only included impact studies using counterfactuals, i.e. experimental and quasi-experimental methods. This includes the alleged ‘gold standard’ of randomized controlled trials (RCT), or Before-After-Control-Intervention (BACI) design. Various econometric techniques attempt to ex-post model a counterfactual, including using matching to identify adequate control observations, or selecting non-treated units to synthesize a control unit. Yet, different recall techniques can also be used to gather baseline data ex post in the field. To make impact estimates quantitatively comparable, we also needed standard deviation estimates. Many case-study authors did not publish these; we had to contact several for this supplement.

Study identification strategy

We started by screening our pool of REDD + studies from prior reviews.^26,29,49,50 An initial set of 15 eligible studies with quantitative estimates of REDD + and carbon-focused PES projects using counterfactual impact evaluation methods was identified. A Boolean search string based on title and abstract of this initial sample was semi-automatically generated following the method described by Grames et al. (2019) (cf. SI, Fig.S1).

We extracted study characteristics such as location, intervention details, sample characteristics along with Hedge’s G effect sizes. Our final sample comprises a total of 30 REDD + interventions, analysed in 32 studies, with 52 effect sizes being included (35 forest-related, 17 socioeconomic outcomes). This includes disaggregated effects being used in the moderation analysis. For the main analyses, we aggregated effects resulting in 23 and 12 estimates for environmental and socioeconomic indicators, respectively. For a meta-study, this remains a fairly small sample, restricting also our analytical options: although the number of rigorous impact studies has expanded rapidly in recent years (around half of the articles included have been published since 2017), but more is still needed to reach a critical mass for detailed statistical analysis. Our studies are just about equally divided between specialized REDD + projects/programmes and PES schemes; yet the latter concentrate on fewer cases. In the former category, some studies are multi-case comparisons, e.g. a pool of Amazon Fund-financed and VCS-certified private REDD + projects⁴¹ and cases from CIFOR’s Global Comparative Study on REDD+ (GCS-REDD) (Bos et al. 2017, Duchelle et al. 2017, Larson et al. 2018; Sunderlin et al. 2017).

How well does our final sample represent the REDD + universe? For recall, it is shaped by the filters we have applied (cf. Tables 1), overlaying geographically an initial implementation bias (where have REDD + investors gone?) with a research bias (where have scientists preferred to work, and found access to data?), and publication bias (is it more likely that positive results are published than negative or null results?). Our small sample mirrors an ‘absolute’ implementation bias towards Latin America (Brazil, Andes, Mesoamerica); it covers less well some ‘high-density’ REDD + countries (Kenya, Colombia, Guatemala). We did find evidence for a moderate publication bias based on Egger’s regression test on funnel plot asymmetry (cf. SI): environmentally positive, significant results have a slightly higher likelihood of getting published. On aggregate, the external validity of our sample is deficient, yet still vastly exceeds that of earlier meta-studies of forest conservation incentives, having been based on much smaller and geographically more biased samples.^24,25

Meta-analysis

The meta-analysis was carried out using the standardized mean difference as the outcome measure. We use the metafor along with clubSandwich packages in R. A multi-level random-effects model was fitted to the data, including random effects at the study and country level. For the main estimates we assumed a correlation of 0.8 within studies and countries, and report robust variance estimates based on the correlated hierarchical effects procedure.⁵² We conducted subgroup analyses, testing for differences between self-declared REDD + and PES-cum-carbon programmes. Similarly, we tested for differences between the outcome and impact levels of the socioeconomic variables. For the moderation analysis we also included binary moderators indicating a) deforestation pressure (1 for high pressure; 0 otherwise); b) spatial targeting (1 if study explicitly mentions ecosystem service density and/or deforestation threat as determining factors for the location and/or intensity of the intervention; 0 otherwise); and c) benefit differentiation (1 if study explicitly mentions differently sized benefit levels within the same scheme; 0 otherwise).

Our binary division between high and low deforestation threat was based on the position vis-à-vis the mean annual deforestation rate over the period 2001-21 across all countries (0.28% y-1) from Global Forest Watch (GFW). We compare this threshold with the average case-level deforestation rate considering the last five years previous to the year start of the evaluated REDD + intervention.

Robustness checks

Lacking significant differences between the two forest-size outcomes (forest cover, deforestation rate), we included them both in the same primary-effect analysis. In addition, we found no evidence that impact estimates would vary systematically with programme duration. We tested to what extent our results were driven by a few influential studies by a) consecutively excluding studies with high weights, namely Groom et al. (2022) and Guizar et al. (2022); b) excluding studies using the synthetic control method, and c) excluding studies with a Cook’s Distance larger than two standard deviations. The coefficient sizes slightly changed, but our conclusions remain robust.

Several studies employed matching techniques, and to calculate effect sizes one requires the correlation between pairs of observations (Borenstein 2009:29, Formula 4.27).⁵³ Due to missing data, we assumed a correlation coefficient of 0.5 for our main specification but tested also more extreme values (0.3 and 0.7) as a robustness check. Indeed, we found that REDD + effect estimates are sensitive to the assumed parameter in the calculation method, but not enough to alter our findings based on Figs. 2 and 3.

Acknowledgments

We are grateful for financial support from Germany’s Federal Ministry for Food and Agriculture, (BMEL) through EFI’s NewGo project, Norad through CIFOR’s Global-Comparative Study on REDD+, and the German Research Foundation under its Excellence Strategy [EXC-2070-390732324-PhenoRob]. We thank multiple authors of primary studies for sharing with us unpublished quantitative material from their respective impact evaluations, and commentators on previous versions of this document: Amy Duchelle, Steven Dörr, Leif-John Fosse, Yitagesu Tekle, Thales West, and Georg Winkel.

Supplemental Materials

hyperlink-TBD

Santilli, M.P., P. Moutinho, S. Schwartzman, D.C. Nepstad, L. Curran, C. Nobre. (2005). Tropical deforestation and the Kyoto Protocol: an editorial essay. Climatic Change 71: 267-276.
Mollicone, D., F. Achard, S.Federici, H.D. Eva, G. Grassi, A. Belward, F. Raes, G. Seufert. H.J. Stibig, G. Matteucci, E.D. Schulze. (2007). An incentive mechanism for reducing emissions from conversion of intact and non-intact forests. Climatic Change 83: 477-493.
Stern, N. (2006). Stern Review: the economics of climate change. Cambridge, UK, Cambridge University Press.
Turnhout, E., Gupta, A., Weatherley-Singh, J., Vijge, M.J., de Koning, J., Visseren-Hamakers, I.J., Herold, M., Lederer, M., 2017. Envisioning REDD+ in a post-Paris era: between evolving expectations and current practice. Wiley Interdiscip. Rev. Clim. Chang. 8, 1–13. https://doi.org/10.1002/wcc.425
Angelsen, A. (2017). REDD+ as result‐based aid: General lessons and bilateral agreements of Norway. Review of Development Economics, 21(2), 237-264.
Pagiola, S. (2011). Using PES to implement REDD. Washington DC: World Bank, Latin America and Caribbean Sustainable Development Department (pp.22).
Wunder, S. (2009). Can payments for environmental services reduce deforestation and forest degradation? In A. Angelsen and M. Brockhaus (Eds.), Realising REDD+: National strategy and policy options (pp. 213-221). Bogor: CIFOR.
Wertz-Kanounnikoff, S., and Angelsen, A. (2009). Global and national REDD+ architecture. In A. Angelsen (Ed.), Realising REDD (pp. 13-24). Bogor: CIFOR.
Sills, E.O., Atmadja, S.S., de Sassi, C., Duchelle, A.E., Kweka, D.L., Resosudarmo, I.A.P., and Sunderlin, W. D. (Eds.). (2014). REDD+ on the ground: A case book of subnational initiatives across the globe. Bogor: CIFOR.
Simonet G., Agrawal A., Bénédet F., Cromberg M., de Perthuis C., Haggard D., Jansen N., Karsenty A., Liang W., Newton P., Sales A-M, Schaap B., Seyller C., Vaillant G., (2018a) ID-RECCO, International Database on REDD+ projects and programs, linking Economic, Carbon and Communities data. version 3.0. http://www.reddprojectsdatabase.org
Pedroni, L., M. Dutschke, C. Streck, M. Porrúa. (2009). Creating incentives for avoiding further deforestation: the nested approach. Climate Policy 9(2): 207-220.
Atmadja, S. S., Duchelle, A. E., De Sy, V., Selviana, V., Komalasari, M., Sills, E. O., & Angelsen, A. (2022). How do REDD+ projects contribute to the goals of the Paris Agreement? Environmental Research Letters, 17(4), 044038. doi:10.1088/1748-9326/ac5669
Nepstad, D., S. Irawan, T. Bezerra, W. Boyd, C. Stickler, J. Shimada, O. Carvalho Jr., K. MacIntyre, A. Dohong, A. Alencar, A. Azevedo, D. Tepper, S. Lowery. 2013. More Food, More Forest, Few Emissions, Better Livelihoods: Linking REDD+, Sustainable Supply Chains and Domestic Policy in Brazil, Indonesia and Colombia. Carbon Management 4 (6): 639–58.
Duchelle, A.E., F. Seymour, M. Brockhaus, A. Angelsen, A.M. Larson, M. Moeliono, G.Y. Wong, T.T. Pham, C. Martius. 2019. Forest-based climate mitigation: lessons from REDD+ implementation. World Resources Institute Issues Paper.
Pistorius, T. and Kiff, L. (2015). The Politics of German Finance for REDD. Center for Global Development Working Paper No. 390. http://dx.doi.org/10.2139/ssrn.2622776
Fletcher, R., Dressler, W, Buescher, B., and Anderson, Z.R. (2016). Questioning REDD+ and the future of market-based conservation. Conservation Biology 30:673–675.
Corbera, E. (2012). Problematizing REDD+ as an experiment in payments for ecosystem services. Current Opinion in Environmental Sustainability, 4(6), 612-619. https://doi.org/10.1016/j.cosust.2012.09.010
Griffiths, T. (2007). Seeing “RED”? “Avoided deforestation” and the rights of Indigenous Peoples and local communities., Forest Peoples Programme, Moreton-in-Marsh, UK.
Chomba, S., Kariuki, J., Lund, J.F., Sinclair, F., (2016). Roots of inequity: How the implementation of REDD+ reinforces past injustices. Land Use Policy 50, 202–213. https://doi.org/10.1016/j.landusepol.2015.09.021
Angelsen, A., Brockhaus, M., Duchelle, A.E., Larson, A., Martius, C., Sunderlin, W.D., Verchot, L., Wunder, S. (2017). Learning from REDD+: a response to Fletcher et al. Conservation Biology, 31(3), 718-720.
Pham T.T., M. Moeliono, A. Angelsen, M. Brockhaus, P. Gallo, T.L. Hoang, T.L.C. Dao, C. Ochoa and K. Bocanegra. 2018. Strategic Alignment: Integrating REDD+ in NDCs and National Climate Policies. In Transforming REDD+: Lessons and New Directions, ed. by A. Angelsen, C. Martius, V. De Sy, A.E. Duchelle, A.M. Larson, and T.T. Pham. Bogor, Indonesia: CIFOR.
Gertler, P.J., S. Martinez, P. Premand, L.B. Rawlings, and C.M.J. Vermeersch. (2016). Impact Evaluation in Practice, 2^nd edition. Washington, DC: Inter-American Development Bank and World Bank. doi:10.1596/978-1-4648-0779-4.
Ferraro, P. and S. Pattanayak (2006). "Money for nothing? A call for empirical evaluation of biodiversity conservation investments." PLoS Biology 4(4): e105. doi:110.1371/journal.pbio.0040105.
Samii, C., M. Lisiecki, P. Kulkarni, L. Paler and L. Chavis (2014). Effects of payment for environmental services and decentralized forest management on deforestation and poverty in low and middle income countries: a systematic review. Campbell Systematic Reviews. Oslo, The Campbell Collaboration: 91.
Snilsveit, B., Stevenson, J., Langer, L., da Silva, N., Rabat, Z., Nduku, P., . . . Ferraro, P. J. (2019). Incentives for climate mitigation in the land use sector—the effects of payment for environmental services (PES) on environmental and socio-economic outcomes in low‐and middle‐income countries: A mixed‐methods systematic review.
Börner, J., D. Schulz, S. Wunder and A. Pfaff (2020). The Effectiveness of Forest Conservation Policies and Programs. Annual Review of Resource Economics 12(1): 45-64.
Ravikumar, A., Larson, A. M., Duchelle, A. E., Myers, R., & Gonzales Tovar, J. (2015). Multilevel governance challenges in transitioning towards a national approach for REDD+: evidence from 23 subnational REDD+ initiatives. International Journal of the Commons, 9(2), 909–931. DOI: http://doi.org/10.18352/ijc.593
Caplow, S., Jagger, P., Lawlor, K., & Sills, E.O. (2011). Evaluating land use and livelihood impacts of early forest carbon projects: Lessons for learning about REDD+. Environmental Science & Policy, 14(2), 152-167. doi:https://doi.org/10.1016/j.envsci.2010.10.003
Duchelle A.E., G. Simonet, W.D. Sunderlin, and S.Wunder. (2018). “What is REDD+ achieving on the ground?” Current Issues in Sustainability 32: 134–140.
Editorial: Weighing up policy tools. Nature Sustainability 4, 561-561, doi:10.1038/s41893-021-00754-y (2021).
Wunder, S., Börner, J., Ezzine-de-Blas, D., Feder, S., & Pagiola, S. (2020). Payments for Environmental Services: Past Performance and Pending Potentials. Annual Review of Resource Economics, 12(1), 209-234. doi:10.1146/annurev-resource-100518-094206
Ezzine-de-Blas, D., Wunder, S., Ruiz-Pérez, M., & Moreno-Sanchez, R. d. P. (2016). Global Patterns in the Implementation of Payments for Environmental Services. PLoS ONE, 11(3), e0149847. doi:10.1371/journal.pone.0149847
Wunder, S., Brouwer, R., Engel, S., Ezzine-de-Blas, D., Muradian, R., Pascual, U., & Pinto, R. (2018). From principles to practice in paying for nature’s services. Nature Sustainability, 1(3), 145-150. doi:10.1038/s41893-018-0036-x
Sunderlin W.D., Ekaputri AD, Sills EO, Duchelle AE, Kweka D, Diprose R, Doggart N, Ball S, Lima R, Enright A, Torres J, Hartanto H and Toniolo A. (2014). The challenge of establishing REDD+ on the ground: Insights from 23 subnational initiatives in six countries. Occasional Paper 104. Bogor, Indonesia: CIFOR.
Wunder, S., A. E. Duchelle, C. d. Sassi, E. O. Sills, G. Simonet and W. D. Sunderlin (2020). REDD+ in theory and practice: how lessons from local projects can inform jurisdictional approaches." Frontiers in Forests and Global Change 3(11). https://doi.org/10.3389/ffgc.2020.00011
Nantongo, M., & Vatn, A. (2019). Estimating transaction costs of REDD+. Ecological Economics, 156, 1-11. https://doi.org/10.1016/j.ecolecon.2018.08.014
Thompson, O. R. R., Paavola, J., Healey, J. R., Jones, J. P., Baker, T. R., & Torres, J. (2013). Reducing Emissions from Deforestation and Forest Degradation (REDD+): Transaction Costs of six Peruvian Projects. Ecology and Society, 18(1).
Rakatama, A., Pandit, R., Ma, C., & Iftekhar, S. (2017). The costs and benefits of REDD+: A review of the literature. Forest Policy and Economics, 75, 103-111. https://doi.org/10.1016/j.forpol.2016.08.006
Montoya-Zumaeta, J.G., Rojas, E., & Wunder, S. (2019). Adding rewards to regulation: The impacts of watershed conservation on land cover and household wellbeing in Moyobamba, Peru. PLoS ONE, 14(11), e0225367. doi:10.1371/journal.pone.0225367
Rosa da Conceição, H., Börner, J., & Wunder, S. (2018). REDD+ as a Public Policy Dilemma: Understanding Conflict and Cooperation in the Design of Conservation Incentives. Forests, 9 (11), 725. Retrieved from https://doi.org/10.3390/f9110725
West, T.A.P.; Börner, J.; Sills, Erin O.; Kontoleon, Andreas (2020): Overstated carbon emission reductions from voluntary REDD+ projects in the Brazilian Amazon. In Proceedings of the National Academy of Sciences of the United States of America 117 (39), pp. 24188–24194. DOI: 10.1073/pnas.2004334117.
Murray, J. P., Grenyer, R., Wunder, S., Raes, N., & Jones, J. P. (2015). Spatial patterns of carbon, biodiversity, deforestation threat, and REDD+ projects in Indonesia. Conservation Biology, 29(5), 1434-1445.
Weiss, C. H. (1997). How Can Theory-Based Evaluation Make Greater Headway? Evaluation Review, 21(4), 501-524. doi:10.1177/0193841X9702100405
Börner, J. and S. Vosti (2013). Managing tropical forest ecosystem services: an overview of options. Governing the provision of ecosystem services. R. Muradian and L. Rival. Dordrecht, Springer: 21-46.
Blom, B., Sunderland, T., & Murdiyarso, D. (2010). Getting REDD to work locally: lessons learned from integrated conservation and development projects. Environmental Science & Policy, 13(2), 164-172. doi:https://doi.org/10.1016/j.envsci.2010.01.002
Pfaff, A., & Robalino, J. (2017). Spillovers from Conservation Programs. Annual Review of Resource Economics, 9(1), 299-315. doi:10.1146/annurev-resource-100516-053543
Meyfroidt, P., J. Börner, R. Garrett, T. Gardner, J. Godar, K. Kis-Katos, B. Soares-Filho, and S. Wunder (2020). Focus on Leakage and Spillovers: Informing Land-Use Governance in a Tele-Coupled World. Environmental Research Letters (15):9 https://iopscience.iop.org/article/10.1088/1748-9326/ab7397.
Sohngen, B. and S. Brown (2004). "Measuring leakage from carbon projects in open economies: a stop timber harvesting project in Bolivia as a case study." Canadian Journal of Forestry Resarch 34(4): 829-839.
Burivalova, Z., Allnutt, T. F., Rademacher, D., Schlemm, A., Wilcove, D. S., & Butler, R. A. (2019). What works in tropical forest conservation, and what does not: Effectiveness of four strategies in terms of environmental, social, and economic outcomes. Conservation Science and Practice, 1(6), e28. https://doi.org/10.1111/csp2.28
Montoya-Zumaeta, J. G. (2020). Reconciling conservation and development? Impacts of incentive-based forest conservation in Peru. (PhD Thesis), Australian National University (ANU), Canberra, Australia.
Grames, E. M., Stillman, A. N., Tingley, M. W., & Elphick, C. S. (2019). An automated approach to identifying search terms for systematic reviews using keyword co‐occurrence networks. Methods in Ecology and Evolution, 10(10), 1645–1654. https://doi.org/10.1111/2041-210x.13268
Pustejovsky, James (2020). clubSandwich: Cluster-Robust (Sandwich) Variance Estimators with Small-Sample Corrections. R package version 0.5.2. https://CRAN.R-project.org/package=clubSandwich
Borenstein, M. (2009). Effect sizes for continuous data. In H. Cooper, L. V. Hedges, & J. C. Valentine (Eds.), The handbook of research synthesis and meta-analysis (2nd ed., pp. 221–235). New York: Russell Sage Foundation.

There is NO Competing Interest.

Wunderetal.SIREDDimpactsforNatSust2022.docx
All SI material

Download PDF

Version 1

posted

You are reading this latest preprint version

Modest forest and welfare impacts from current REDD+ initiatives

Status:

Version 1

Abstract

Figures

Introduction

RESULTS

DISCUSSION

METHODS

Declarations

REFERENCES AND NOTES

Additional Declarations

Supplementary Files

Status:

Version 1