Designing REDD + schemes when forest users are not forest landowners: Evidence from a survey-based experiment in Kenya

Ecological Economics 116 (2015) 46–57

Contents lists available at ScienceDirect

Ecological Economics journal homepage: www.elsevier.com/locate/ecolecon

Analysis

Designing REDD + schemes when forest users are not forest landowners: Evidence from a survey-based experiment in Kenya Marcella Veronesi a,b,⁎, Tim Reutemann c, Astrid Zabel d, Stefanie Engel e a

Department of Economics, University of Verona, Vicolo Campofiore 2, 37129 Verona, Italy Institute for Environmental Decisions, ETH Zurich, Switzerland ETH Zurich, Institute for Environmental Decisions, Universitätstrasse 22, 8096 Zurich, Switzerland d Bern University of Applied Sciences, School of Agricultural, Forest, and Food Sciences, Laenggasse 85, 3052 Zollikofen, Switzerland e University of Osnabrueck, Institute for Environmental Systems Research, Barbarastrasse 12, 49076 Osnabrück, Germany b c

a r t i c l e

i n f o

Article history: Received 29 July 2014 Received in revised form 18 March 2015 Accepted 9 April 2015 Available online xxxx JEL Classification: I38 J22 O13 Q18 Q23 Q28 Q56

a b s t r a c t This study contributes to the debate on Reducing Emissions from Deforestation and Forest Degradation (REDD+) and the relationship between land tenure and forest conservation. We investigate policies that create alternative livelihood options for people around REDD+ forests who are forest users but not forest landowners. We compare the performance of a conventional integrated conservation and development policy (ICDP) with an alternative hybrid policy that combines features of ICDP and payments for environmental services. Through a survey-based experiment in Kenya, we compare the effectiveness of different REDD+ payment schemes given rising opportunity costs of forest use. This study shows that hybrid approaches that provide alternative income opportunities to local people, target the local drivers of deforestation, are conditional on environmental outcomes, and account for changing opportunity costs could work as effective policy options. © 2015 Elsevier B.V. All rights reserved.

Keywords: REDD Payments for ecosystem services Deforestation Land tenure Integrated conservation and development policy Africa

1. Introduction Reducing Emissions from Deforestation and Forest Degradation (REDD+) has been proposed as policy measure to address deforestation and degradation, and safeguard or increase forest carbon (Angelsen, 2008; Palmer and Engel, 2009; Pistorius, 2012). The policy sets a framework for an exchange of benefits, monetary or other, for guarantees to maintain wooded areas that otherwise would be deforested or degraded. In many cases, the forests at stake are not exploited commercially, but owners have tolerated some degree of subsistence usage through local people who are not forest landowners. Once enrolled in a REDD+ program, such customary use, for example for charcoaling, could become a ⁎ Corresponding author at: Department of Economics, University of Verona, Vicolo Campofiore 2, 37129 Verona, Italy. E-mail addresses: [email protected] (M. Veronesi), [email protected] (T. Reutemann), [email protected] (A. Zabel), [email protected] (S. Engel).

http://dx.doi.org/10.1016/j.ecolecon.2015.04.009 0921-8009/© 2015 Elsevier B.V. All rights reserved.

risk to the newly-valuable trees. This situation calls for accompanying measures that prevent locals from using the forest in any way that is detrimental to REDD+ goals. Yet there are equity concerns that crude fences-and-fines policies to protect REDD+ forests jeopardize local peoples' livelihoods (Ghazoul et al., 2010; IUCN, 2010), implying a need for accompanying policies (Chhatre and Agrawal, 2009; Palmer and Silber, 2010; Groom and Palmer, 2012). In addition, as up to 800 million people worldwide are estimated to be dependent on such forests for their livelihoods (Chomitz et al., 2006; World Resources Institute, 2005), it has been argued that poverty reduction should be incorporated as a ‘co-benefit’ of REDD+ policy (Brown et al., 2009). This study contributes to the literature on the relationship between land tenure and forest carbon management (Duchelle et al., 2014; Holland et al., 2014; Pfaff et al., 2014; Resosudarmo et al., 2014; Sunderlin et al., 2014). In particular, we consider REDD+ schemes in a context where resource user communities are not forest land owners but have joint customary rights over forests. We contribute to the debate

M. Veronesi et al. / Ecological Economics 116 (2015) 46–57

on optimal REDD+ policy design by comparing the performance of a conventional integrated conservation and development policy (ICDP) with a hybrid policy that combines features of ICDP and payments for environmental services (PES), and by testing the effectiveness of different payment designs.1 The specific hybrid policy analyzed is an eco-charcoaling policy where the price paid to forest users for sustainably harvested raw material is to some degree conditional on reduced forest degradation. We move beyond conventional ICDP approaches still predominantly applied by implementing organizations. ICDPs aim to reduce pressure on forests by providing alternative income opportunities (Hughes and Flintan, 2001). For example, agricultural policies are implemented with the idea that improved agricultural production opportunities reduce forest product extraction by local people. However, a substantial body of literature has demonstrated that the effectiveness of ICDP approaches is limited (Hughes and Flintan, 2001). Some studies argue that PES that are made conditional on an improved environmental outcome are environmentally more effective and also more cost-effective than ICDP (Ferraro, 2001; Ferraro and Kiss, 2002; Engel et al., 2008). Like ICDP, PES have the potential to address poverty and environmental concerns at the same time (Pagiola et al., 2005). The PES approach has risen tremendously in popularity over the past decades (Farley and Costanza, 2010; Kosoy and Corbera, 2010; Pascual and Corbera, 2011; Matzdorf et al., 2013; Schomers and Matzdorf, 2013). Yet a number of recent studies show that PES are not always environmentally effective or cost-efficient either (Pattanayak et al., 2010; Vatn, 2010; Muradian et al., 2013). In addition, in the context where forests are subject to customary use rights that are not formalized legally, implementation of PES faces three major difficulties that hamper its implementation. First, paying people for non-use of a resource that is not formally theirs to start with has weak legal basis. Second, it has been shown that PES design with weak property rights is highly complex, can be counterproductive, and may involve trade-offs between environmental and poverty alleviation objectives (Corbera et al., 2007; Engel and Palmer, 2008, 2009; Engel et al., 2013). Third, customary rights are often held by groups of individuals such as local communities. This induces issues of a commons dilemma (Zabel et al., 2014). Given the weaknesses of both PES and ICDP approaches for the case of REDD+ design for customary forest users, in this paper we implement a survey-based experiment in Kenya to answer the following research questions: (i) how does a hybrid policy that combines ICDP and PES features compare to a conventional ICDP policy? (ii) How do different payment designs compare to each other? and (iii) what is the most effective policy under volatile opportunity costs? Is a policy that indexes payments to opportunity costs more effective than a standard policy with fixed payments? To the best of our knowledge, this is the first study that compares the effectiveness of different REDD+ payment schemes in the field, and provides some insights on the effectiveness of different policies when forest users are different from forest landowners. It is also to our knowledge the first study that assesses the environmental effectiveness of a hybrid approach. We show that the hybrid approach is environmentally effective, and worthwhile exploring in actual REDD+ policy. The paper is organized as follows. Sections 2, 3, and 4 describe the case study, the experimental design, and the data, respectively; Section 5 presents the empirical model, and Section 6 the results. Section 7 concludes, highlighting policy implications, and directions for future research. 2. Case Study Description This study focuses on the Kasigau Corridor REDD+ Project in Kenya, which is the first REDD+ project ever to issue carbon credits under an 1

In related ongoing work by some of the authors of this paper, we compare different REDD+ policy designs in Brazil, but in a setting with secure individual property rights (Reutemann et al., 2014).

47

internationally accepted carbon standard (Peters-Stanley et al., 2011). In the Kasigau Corridor, the forest users are not the forest landowners. The forest land is split into several community ranches, which are owned by shareholder companies. The shares have historically been distributed among the population living on the more fertile hills that surround the forest, but since then the population increased heavily and many shareholders also migrated to Nairobi or Mombasa. In general, the shareholders do not live close to the forest. They receive a share of the revenue from the sale of carbon credits, which is high above their opportunity cost, as the area is rather infertile and the forest is of low commercial value for the owners. The amount of land per shareholder varies greatly. The ranch with most shareholders has 2500 shareholders, while the one with least (which is about one tenth the size) has only one. Under the REDD+ agreement, the land was leased to a conservation company, which is responsible for the entire carbon accreditation and commercialization, as well as protection of the forest. Apart from various indirect measures, the company also introduced rangers who directly control the forest for illegal charcoaling and tree cutting. The focus of this study was laid on the forest users rather than the forest landowners, because the former face substantial opportunity costs of forest conservation. Despite being illegal in Kenya, charcoaling is a widespread practice and the base of many livelihoods, as well as a major cooking fuel in the entire country.2 Although domestic demand in Kenya has been reduced through the introduction of efficient charcoal stoves, we still consider demand as inelastic, as charcoal is exported through Mombasa harbor and therefore linked to international demand for energy carriers. Therefore policies with the aim to reduce unsustainable charcoaling primarily need to address the supply side. The investigated hybrid policy is an indirect payment through the financial support of eco-charcoal factories, which pay local land users for the supply of sustainably harvested raw material, i.e., scrap wood from fast growing shrubs, while at the same time supplying a sustainable substitute for the nonrenewable charcoal for the end users. In the Kasigau Corridor, charcoal production is the key driver for forest degradation. It also paves the way for deforestation, as the land becomes easier to clear for agriculture once a charcoaler removed all hardwood trees. A pilot eco-charcoal factory has already been set up and is currently producing small amounts of eco-charcoal.3 In the pilot project, hired workers cut shrubs for daily wage. The project owner made deliberate efforts to hire excharcoalers. This setup has several disadvantages when aiming to scale up: (i) the access to shrubs is limited to land owned or leased by the factory operators and public lands; and (ii) it could be perceived as unfair since only charcoalers are employed, and even lead to perverse incentives such as starting charcoaling to get a job. For scaling up to a level of production that can substitute a significant amount of charcoal, we assume that access to shrubs on private land is required, and therefore, we analyze a scheme where anybody can sell shrubs at the factory gate. 3. Experimental Design A complexity that has been discussed in PES design is that prices in developing countries are often volatile and households myopic. Also, in the light of growing world food demand, promotion of biofuels, and resulting increases in agricultural commodity prices, the opportunity costs of forest conservation may well increase over time and induce landowners to breach REDD+ contracts (Butler et al., 2009). Designing REDD+ schemes under these conditions can be challenging. Very few studies exist on how REDD+ scheme design may help to address this issue, and they mostly focus on the allocation of liability between buyers and sellers of REDD credits (Dutschke and Wong, 2003; Dutschke and 2 Hosonuma et al. (2012) emphasize that the most important driver of forest degradation in Africa is charcoaling and fuel wood collection, accounting for 48% of forest degradation. 3 The production of eco-charcoal requires equipment costing several thousands USD and is therefore only feasible when done at least at the small factory scale.

48

M. Veronesi et al. / Ecological Economics 116 (2015) 46–57

Angelsen, 2008). Engel et al. (2014) use real options theory to model land users' decisions to convert forest to agriculture in the face of uncertainty in agricultural commodity prices. They find that payments linked to the international carbon market price are larger than those linked to an agricultural commodity index. We implemented a survey based-experiment of time allocation to elicit the behavior of the forest users under different hypothetical policies. We chose an experiment on time allocation rather than for example, a willingness to pay survey, as the local economy is strongly cash constrained. The population is therefore more familiar with decisions relating to work time. Our proxy for forest degradation is the time spent making conventional charcoal at the household level. We collected data on time allocation for the status quo, and for six different realistic policy scenarios, two related to the introduction of an agricultural policy, and four related to the introduction of an eco-charcoal policy. Each policy is implemented one at a time, and not simultaneously. A detailed description of each policy follows. 3.1. Agricultural Policy A key feature of ICDP approaches is to divert labor away from the destructing activity (Hughes and Flintan, 2001). Increasing agricultural productivity is one of the main factors that could yield to both an increase in incomes and a decrease in deforestation (Sunderlin et al., 2005). In our study, the goal of the agricultural policy is to divert labor away from charcoaling to the cultivation of a new crop. We consider the cultivation of ‘jojoba’ (Simmondsia chinensis) for which seeds were previously not available. We chose jojoba because this crop is currently under consideration by the conservation company at the case study site in Kenya. They appreciate it as a high-revenue, low labor-intense crop. The choice for jojoba was exogenously given through the conservation company at the study site. However, among the ICDP approaches, the implementation of high revenue plants that require little care like jojoba is not uncommon (Hughes and Flintan, 2001). We investigate two agricultural policy scenarios: (i) Agricultural policy with low price. The value of one acre of jojoba was set at the lower estimate of 150,000 KES/year.4 (ii) Agricultural policy with high price. The value of one acre of jojoba was set to the higher estimate of 250,000 KES/year. 3.2. Eco-charcoal Policies Sunderlin et al. (2005) argue that increased opportunity cost of labor through off-farm jobs could lead to a reduction in deforestation by an increase in incomes. Eco-charcoal policies are an alternative to agricultural policies where households are given the opportunity to collect dry scrap wood from outside the forest or from the forest ground and sell it to an eco-charcoal factory. We investigate three different modalities of setting the price for eco-charcoal raw material: (1) Fixed price policy for eco-charcoal raw material. The price of raw material was set at 200 KES/unit, independent of the charcoal price.5 (2) Indexed price policy for eco-charcoal raw material. The price per kilogram of eco-charcoal raw material was set at 12.5% of the kilogram charcoal price.6 (3) Indexed conditional price. This policy also indexes the price for scrap wood to the charcoal price but additionally introduces a 4 1 KES is 0.0117 USD. The jojoba prices are based on the results of a field trial in the area and current jojoba oil prices. 5 One unit of eco-charcoal raw material was chosen as 100 kg, which is the average amount that one adult collected in one day during a trial. 200 KES is approximately equal to a daily labor salary, and thus, roughly represents the current approach of hiring workers. 6 12.5% was chosen based on a calculation of the conversion efficiency of the raw material with data from the pilot eco-charcoal factory. The factory's objective is to operate revenue neutral by selling eco-charcoal.

conditionality clause. Now the price for scrap wood consists of two parts, the first is a base payment that is indexed to the charcoal price and the second part is a premium that is also indexed to the charcoal price but additionally is a function of the household's labor in charcoaling and the labor that other villagers spend in charcoaling. The household can only decide on the amount of labor that it itself spends in charcoaling. Yet, it has expectations toward the other villagers' behavior. We investigate two types of conditionality: (3.1) Indexed price policy for eco-charcoal raw material with weak conditionality. The minimum price per kilogram of eco-charcoal raw material was set at 10% of the kilogram price of charcoal. Additionally, a premium was introduced. The premium increases by 0.2% of the charcoal price per week up to a maximum of 15% of the charcoal price. If somebody from the village is caught producing illegal bush charcoal in the forest, the premium decreases by 1% of the charcoal price per week. (3.2) Indexed price policy for eco-charcoal raw material with strong conditionality. This is the same as the previous policy, however, if somebody from the village is caught charcoaling or poaching the premium decreases by 2% of the charcoal price per kilogram per week. When these last two policies were presented, subjects were asked to answer three control questions to test if the policies were fully understood. While all three scenarios can be seen as a type of payment for environmental services by some broader definitions, only the third one satisfies conditionality in a stricter sense (e.g., Engel et al., 2008). In a broader sense, conditionality may still be given in the sense that the payments from the final buyer of verified carbon credits are conditional on the state of the forest. Thus, the level of subsidies for eco-charcoaling that an implementing agency can pay ultimately still depends on the state of the forest. Yet conventional policies generally do not make this link explicit. We randomly assigned two policy scenarios (i.e., two questions) to each household: (i) one scenario on the jojoba agricultural policy, and (ii) one scenario on an eco-charcoal policy. In addition, we randomly assigned five charcoal prices (250, 500, 750, 1000, and 1500 KES) for each policy.7 We randomized the order of the two scenarios to account for order effects, and we disclosed in advance the choice tasks.8 Fig. A1 of the Appendix reports an example of policy scenarios. These policies, while hypothetical in our experiment, are realistic. A pilot ecocharcoal factory has already been implemented, and field trials with jojoba are running since four years. The experiment was conducted together with a socio-economic survey addressed to the household head. The experiment was carried out in person by 14 locally recruited interviewers, and one locally recruited, experienced survey manager. The interviewers went through an intensive five-day training course and were given the opportunity to give inputs to the survey design. The survey was pre-tested with 50 subjects and adjusted accordingly. The data were collected during two months in spring 2011. Each interviewer worked in the villages proximate to their own home village, making it easier to find access to and gain trust from the subjects. Among other socio-economic data, the interviewers also had to collect the phone numbers of the subjects. We carefully informed the participants that this was an academic study, and that all data collected were anonymous and confidential. To monitor survey quality, the survey manager then checked via phone for half of 7 The current average price of charcoal is about 455 KES per bag. One bag is equivalent to 35 kg. 8 Day et al. (2012) find robust evidence of order effects in repeat-response stated preference studies, however, they also find that this effect is significantly mitigated by task training and information provision on the tasks.

M. Veronesi et al. / Ecological Economics 116 (2015) 46–57

the questionnaires if the interviews where done properly. The survey manager also did occasional spot checks on the interviewers in the field. The experiment was conducted with the help of visual aids following Delavande et al. (2011). In particular, the interviewers carried a bag with a piece of eco-charcoal and some sticks of the raw material to show to the subjects when presenting the scenarios with the eco-charcoal policy. Furthermore, a cardboard circle was presented on which the interviewees could see four drawings, each representing one of the forest-relevant household activities, namely (1) ‘farm work’ (e.g. cropping, field protection, livestock tending), (2) ‘firewood collection’, (3) ‘charcoaling’ and (4) ‘other activities’ (including charcoal trading) (see Panel A, Fig. A2 of the Appendix). We added ‘collection of scrap wood’ as a fifth activity in the scenarios with the eco-charcoal policy (see Panel B, Fig. A2 of the Appendix). To elicit how the individuals allocate their time during a normal week, they were asked to distribute beans on the activity drawings on the circle. Fourteen beans needed to be distributed per individual with each bean representing half a day. We opted for the use of half days as unit during the pre-study field visit since charcoaling is a time-intensive activity. It requires at least half a day to travel into the forest, cut the trees, and prepare the kiln. The interviewers talked the subjects through the week by asking separately for every half day, i.e. ‘How do you spend your Monday mornings? Your Monday afternoons? etc.’. We used different colored beans, one color per household member. The household head was asked to choose the time allocation for absent household members. Children under the age of 12 were excluded. The bean game was well accepted by the subjects. 4. Data Description The original design includes 85 villages covering all areas of the Kasigau Corridor REDD+ Project, and 1095 households randomly selected proportionally to the total village population in each village. We drop 123 observations because two interviewers where found working in an inappropriate manner at a late stage of the survey. This causes a slight underrepresentation of one region (Mwatate). Then, we carry out internal and external consistency checks. We drop five observations because the total time allocation is zero, and 96 observations because the total time did not add up to a full week (i.e., 14 beans). We include several quality control questions to elicit directly or indirectly whether there are charcoalers in the household. For example, a respondent stating directly that the household does not engage in charcoaling should not report that the household used the forest for charcoaling during the last dry season or that the household had income gains from charcoaling. We find only ten respondents declaring not charcoaling but with positive income from charcoaling. We then drop 31 observations because they did not declare time spent charcoaling while we identified them as charcoalers through the control questions. Table A1 of the Appendix compares the final ‘clean’ sample with the original sample to test for sample selection bias. We find no significant differences in the main covariates that will be used later on in the analysis, that is household size, household composition, forest and market distance as well as the time allocation at the status quo. In addition, as we will show in Section 6, our findings do not change if we keep or drop these observations. After cleaning the data, the final sample used in the analysis consists of 81 villages and 840 households. Since each respondent answered three time allocation questions (status quo plus two policy scenarios) we have three observations per household, that is a panel dataset of 2520 observations. The survey collects detailed information on household characteristics (e.g., size and composition), forest use (e.g., fuel wood collection, livestock tending, charcoaling, and hunting), and household time use in the aforementioned activities: farm work, collection of firewood, charcoaling, collection of scrap wood, and ‘other activities.’ Panel A of Table 1 presents the descriptive statistics at the status quo for the whole sample and separately for each treated group. Formal t-tests show the quality of the policy randomization: the null hypothesis that the policy treatment groups are equal for the main variables is not

49

rejected for nearly all the variables presented with the exception of household size for the group corresponding to the indexed price policy with strong conditionality, and for the share of children aged 12–17 for the group corresponding to the indexed price policy with weak conditionality (differences significant at the 5% statistical level). The average household consists of five members, in particular, about 28% are male adults, 25% female adults, 2% elderly, 24% children aged 0–11, and about 11% children aged 12–17.9 The majority of households own their land (90%) and are small farm-households (about 5 acres on average). An average household spends most of the time per week on farm work and doing ‘other activities.’ Panel B of Table 1 presents the average time allocation after the policy treatment. Farm work is still the main activity. However, a simple comparison of time allocation at the status quo and after the treatment (Panel A vs. Panel B) seems to convey that the time spent charcoaling decreased, and in particular, that eco-charcoal policies appear to be more effective than agricultural policies in reducing time spent charcoaling. 5. Empirical Model The effects of different policy scenarios and raising opportunity costs on time allocation are assessed by estimating a system of K equations. We can write the k-th equation for household h as follows
 ð1Þ ykhq−1 ¼ α0k þ Dkq α1k þ pkq  Dkq α2k þ Xkhq α3k þ εkhq where the dependent variable is the weekly number of half days spent in each activity k (k = 1, …, K) by household h at the status quo (q = 1) or under a policy scenario (q = 2, …, Q, with q = 2 corresponding to the second policy scenario in our case study, and Q to the last one); vector D includes dummy variables equal to 1 for each policy scenario described in the experimental design section, 0 otherwise10; p is a vector of relative charcoal prices with respect to the status quo, that is the charcoal price randomly assigned to each household divided by the charcoal price at the status quo; p × D represents the interaction term between relative charcoal prices and policy dummy variables. In addition, X represents a vector of individual and household characteristics such as household composition and size, hectares of land owned, number of livestock heads, distance to the market and to the forest. We also control for the order in which the eco-charcoal policy question was presented by including a dummy variable equal to 1 if an eco-charcoal policy is presented as first, 0 if second. Since each respondent answered multiple time allocation questions, it is likely that the responses are correlated. In addition, because the respondent has to choose how to allocate his/her time among K activities, our final empirical model is a system of K equations estimated simultaneously. In particular, we employ a random effect seemingly unrelated regression model (Biorn, 2004) where the error term is comprised of two components, both of which are normally distributed: εkhq = νkh + ηkhq ~ N(0,V), where the term ν is an individual-specific error component that remains unchanged within a household over questions, and is independent across households; and η is an independent and identically distributed (i.i.d.) error across and within households. We then derive the marginal effect of each policy as the difference between the expected time allocation at the status quo (ykh1) and at each policy scenario q with q ≠ 1 (ykhq−1 ), that is h i E ykhq−1 − E½ykh1  ¼ α1k þ α2k pkq :

ð2Þ

Note that Eq. (2) represents the impact of a policy when the randomly assigned charcoal prices are equal to the status quo charcoal price, i.e., when pkq is equal to one. In the other cases, when the relative price 9 There is a remaining 10% of household members that our dataset did not allow us to identify. However, formal t-tests show that the percentage of people belonging to this residual category is not different by treatment group. 10 Note that D is a vector because each respondent answered multiple policy questions.

50

M. Veronesi et al. / Ecological Economics 116 (2015) 46–57

Table 1 Descriptive statistics. Treatment policies Agricultural policy Full sample

Low price

Eco-charcoal policy High price

Fixed price

Indexed price No conditionality

Weak conditionality

Strong conditionality

Mean

Std. dev.

Mean

Std. dev.

Mean

Std. dev.

Mean

Std. dev.

Mean

Std. dev.

Mean

Std. dev.

Mean

Std. dev.

Panel A — status quo Charcoaling (weekly half days) Farm work (weekly half days) Fire wood collection (weekly half days) Other activities (weekly half days)

3.912 10.818 3.815 23.588

6.289 9.076 4.180 17.745

3.807 10.995 3.844 24.563

6.408 9.188 3.927 18.447

4.009 10.653 3.789 22.680

6.182 8.978 4.408 17.038

4.146 11.169 3.543 24.484

6.268 10.168 3.678 18.405

4.049 11.420 4.332 24.112

6.205 10.024 5.404 18.057

3.409 10.790 4.199 24.656

6.562 7.902 4.183 18.544

3.974 9.970 3.304 21.404

6.176 7.890 3.211 16.006

Control variables Household size Share of male adults Share of female adults Share of elderly Share of kids 0–11 Share of kids 12–17 Land owned (Ha) Number of livestock Village forest distance (km) Village market distance (km)

4.968 0.281 0.255 0.024 0.243 0.114 5.165 10.362 5.300 21.655

2.710 0.245 0.187 0.092 0.231 0.160 4.608 12.617 5.293 17.272

5.069 0.282 0.253 0.021 0.244 0.119 5.145 10.373 5.147 20.595

2.814 0.241 0.185 0.090 0.230 0.168 4.889 13.532 4.975 16.868

4.874 0.280 0.257 0.026 0.243 0.109 5.174 10.352 5.423 22.641

2.608 0.249 0.190 0.093 0.232 0.153 4.335 11.716 5.575 17.602

5.014 0.302 0.257 0.026 0.226 0.101 5.115 10.260 5.119 20.447

2.996 0.257 0.200 0.109 0.232 0.152 4.616 14.776 5.005 17.518

5.156 0.276 0.265 0.022 0.238 0.121 5.177 10.034 5.300 20.969

2.732 0.237 0.189 0.080 0.227 0.155 4.323 10.938 5.371 16.592

5.134 0.258 0.249 0.015 0.265 0.140 5.180 10.505 5.179 20.770

2.591 0.219 0.165 0.060 0.226 0.183 5.205 11.940 4.953 16.114

4.622 0.284 0.251 0.030 0.248 0.098 5.170 10.635 5.532 24.131

2.471 0.260 0.191 0.103 0.236 0.151 4.355 12.386 5.761 18.363

Panel B — after policy treatment Charcoaling (weekly half days) Farm work (weekly half days) Fire wood collection (weekly half days) Other activities (weekly half days) Scrap wood collection (weekly half days) Sample size

2.515 10.286 3.324 19.489 6.415 840

5.497 9.054 4.005 16.911 9.995

3.807 12.960 3.847 22.548 – 405

6.955 10.298 4.622 18.590 –

3.152 13.411 3.936 20.536 – 435

5.836 9.656 4.989 17.018 –

1.863 7.498 2.671 18.534 12.584 219

4.095 7.540 2.481 16.655 10.348

1.937 7.010 3.083 17.678 14.000 205

4.722 6.789 3.519 15.194 11.757

1.081 7.435 2.855 18.344 13.263 186

3.422 6.238 2.575 16.948 12.172

1.335 7.548 2.465 15.570 11.674 230

4.348 6.947 2.817 14.052 9.112

of charcoal pkq is different from one, we can use Eq. (2) to investigate the effect of raising opportunity costs, i.e., the impact of the randomly assigned charcoal prices (250, 500, 750, 1000, 1500) on time allocation. An advantage of this estimation procedure is that the randomly assigned prices and policy scenarios are exogenous, that is they are independent of omitted variables included in the error term. This implies that (i) endogeneity problems in estimating α1 and α2 are avoided, (ii) any systematic tendency within households to misstate their time allocation affects only the constant term, and (iii) estimates of parameters are unaffected by whether observed individual and household characteristics are included as additional covariates (we report on both versions below). However, the inclusion of control variables should help absorb any residual variation, and reduce the standard errors. Particular functional forms are chosen to remain within the spirit of previous work in this area (Fafchamps and Quisumbing, 1999; Shively and Fisher, 2004; Fisher et al., 2005; Ito and Kurosaky, 2009). Since we do not observe the shadow cost of labor we control for factors that affect labor supply such as household composition and size (e.g., shares of male and female adults aged 18 to 65, the share of elderly aged 66 and above, and the share of children in the age groups 0–11 and 12–17); hectares of land owned; number of livestock heads; and distance to the market and to the forest. 6. Results Tables 2 and 3 report, respectively, the coefficient point estimates and marginal effects of policy impact at the sample mean of the simultaneous equations random effect model [1], that is on how households change the amount of time spent charcoaling under different policy scenarios with respect to the status quo. Results are robust to different specifications.11 In particular, Model 1 presents the simplest specification 11 Table A2 of the Appendix presents the marginal effects of each policy by using the original ‘non-cleaned’ sample. The results are equivalent to those obtained by using the ‘clean’ final dataset.

where time allocation depends only on the policy treatment variables. Model 2 controls for village heterogeneity by including village fixed effects, and Model 3 presents estimates of Eq. (1) where we included also status quo control variables. The coefficients and marginal effects of policy treatment variables are strongly statistically significant with p-values close to zero in all three models. In particular, Table 2 shows that compared to Model 1 the inclusion of village fixed effects in Model 2 increases the coefficient point estimates slightly and it does not reduce the statistical significance of the coefficients. In line with our expectations, Model 3 shows that including control variables for the status quo while it adds explanatory power to the regression and helps reduce the standard errors, it has a minimal effect on the coefficient point estimates. Also as expected, the distance of the village from the market, household size and the proportion of men in the household have a significant and positive effect on the time spent charcoaling while the distance of the village from the forest has a negative effect. In addition, we tested for the order of presentation of the policy scenarios and find that the order does not matter at the 1% statistical level. Table 3 shows that both agricultural policies and eco-charcoal policies significantly decrease the amount of time allocated to charcoaling. However, the implementation of an eco-charcoal policy has the largest effect, while an agricultural policy with a low jojoba price has the smallest effect. The impact of a fixed price eco-charcoal policy is about five times the impact of a low price agricultural policy, and nearly twice the impact of an agricultural policy with a high jojoba price. Comparing the effects at different charcoal prices, we find that a policy that indexes eco-charcoal payments to charcoalers' opportunity costs is the most effective policy in providing forest conservation. When an eco-charcoal policy with an indexed price is introduced, we observe a significant decrease in time spent charcoaling even at high charcoal prices. For example, if the price of charcoal is set at 1500 and an indexed price policy with strong conditionality is implemented, then a household spends about one day less per weak charcoaling. On the contrary, if we consider the introduction of a cash-crop agricultural

M. Veronesi et al. / Ecological Economics 116 (2015) 46–57

51

Table 2 Policy impact on charcoal time allocation — coefficient point estimates. Variables

Low price agricultural policy Relative charcoal price × low price agricultural policy High price agricultural policy Relative charcoal price × high price agricultural policy Fixed price eco-charcoal policy Relative charcoal price × Fixed price eco-charcoal policy Indexed price eco-charcoal policy No conditionality Relative charcoal price × indexed price eco-charcoal policy with no conditionality Indexed price eco-charcoal policy — weak conditionality Relative charcoal price × indexed price eco-charcoal policy with weak conditionality Indexed price eco-charcoal policy — strong conditionality Relative charcoal price × indexed price eco-charcoal policy with strong conditionality Covariates Household size Share of male adults Share of female adults Share of elderly Share of kids 0–11 Share of kids 12–17 Land owned (Ha) Number of livestock Village forest distance (km) Village market distance (km) Order effects Village fixed effects R-squared Number of observations Number of households

Model 1

Model 2

Model 3

Coeff

Std. err.

Coeff.

Std. err.

Coeff.

Std. err.

−1.160⁎⁎⁎ 0.694⁎⁎⁎ −1.970⁎⁎⁎ 0.723⁎⁎⁎ −3.455⁎⁎⁎ 0.898⁎⁎⁎

0.390 0.166 0.367 0.154 0.501 0.260

−1.268⁎⁎⁎ 0.638⁎⁎⁎ −2.268⁎⁎⁎ 0.760⁎⁎⁎ −3.587⁎⁎⁎ 0.880⁎⁎⁎

0.388 0.163 0.367 0.151 0.495 0.255

−1.278⁎⁎⁎ 0.638⁎⁎⁎ −2.246⁎⁎⁎ 0.753⁎⁎⁎ −3.605⁎⁎⁎ 0.884⁎⁎⁎

0.387 0.163 0.365 0.151 0.493 0.254

−2.987⁎⁎⁎ 0.608⁎⁎⁎ −3.104⁎⁎⁎ 0.445⁎ −2.642⁎⁎⁎

0.507 0.224 0.590 0.254 0.520 0.204

−3.198⁎⁎⁎ 0.604⁎⁎⁎ −3.217⁎⁎⁎ 0.360 −2.712⁎⁎⁎

0.502 0.220 0.584 0.250 0.514 0.201

−3.224⁎⁎⁎ 0.622⁎⁎⁎ −3.196⁎⁎⁎ 0.345 −2.719⁎⁎⁎

0.500 0.219 0.582 0.249 0.512 0.200

0.152

Yes No 0.056 2850 840

0.075

Yes Yes 0.288 2850 840

0.084 0.547⁎⁎⁎ 3.088⁎⁎ 0.882 −2.035 −0.572 0.358 −0.016 0.012 −1.050⁎⁎⁎ 0.337⁎⁎ Yes Yes 0.344 2850 840

0.066 0.948 1.103 1.721 0.956 1.108 0.033 0.012 0.343 0.170

Note: The dependent variable is the household weekly number of half days spent charcoaling. Coefficients are estimated by random effect seemingly unrelated regression model for panel data. ⁎ Significant at the 10% level. ⁎⁎ Significant at the 5% level. ⁎⁎⁎ Significant at the 1% level.

policy such as jojoba, then we find that with respect to the status quo the time spent charcoaling at high charcoal prices increases, specifically, by about half a day if an agricultural policy with a high price for jojoba is introduced. This may be due to the modest amount of labor that jojoba cultivation requires. It fails to absorb labor and divert it away from charcoaling especially when charcoal prices are high. In addition, although the estimated policy coefficients are negative, at high charcoal prices a fixed payment policy does not significantly affect the time spent charcoaling. In addition, we find differences between a policy without conditionality and one with conditionality or different types of conditionality (strong/weak) at high charcoal prices. At the high charcoal price scenario (1500) there is a declining trend of charcoaling time as the policy moves from a policy with no conditionality (impact equal to −0.809) to a policy with weak and strong conditionality (impact equal to − 1.550 and − 2.068, respectively). However, the only statistically significant difference (at the 5% statistical level) is between the policy without conditionality and the one with strong conditionality: the effect under the latter is almost double the effect under the policy with no conditionality. The lack of significant effects for the other cases could be linked to the fact that in our context the payment is conditional on collective rather than individual behavior. As a consequence, the effect of conditionality is likely to depend on monitoring intensity and on the household's expectation of the effective ecocharcoal price it will receive, which depends on others' behavior as well. If a household believes that others will charcoal illegally it may consider a lower expected eco-charcoal price in the conditional scenario and thus increase own charcoaling. Because of missing information on collective action, we leave it for future research to test these hypotheses. Apart from the policies' impacts on charcoaling, the data also reveal very interesting results related to the impact on time spent on farm

work, fire wood and ‘other activities,’ as shown in Table 4.12 We find, as expected, that the agricultural policies increase the time spent for farm work (by about one day more per week), however, they do not affect the time spent collecting firewood. On the contrary, the eco-charcoal policies have a strong significant (at the 1% statistical level) negative effect on time spent collecting firewood, farming, and in particular, time spent in ‘other activities’ because more time is allocated to collect eco-charcoal raw material. 7. Discussion and Conclusions REDD+ is a major topic in the debate on policies to mitigate climate change. Developing mechanisms to ensure forest conservation and create alternative livelihood options when the forest users are not the forest landowners is an important challenge in REDD + scheme design. Conventional ICDP approaches still dominate REDD+ planning despite their weaknesses. At the same time, PES as a potential alternative policy approach faces severe complexities under weak property rights and making PES to customary forest users suffers from weak legal basis. This study contributes to the debate into two ways: first, it investigates an innovative hybrid policy that creates alternative livelihood options for people around REDD + forests that are forest users but not forest landowners; and second, it provides some insights on the effectiveness of different policies under rising opportunity costs of forest use. By linking conditionality to an alternative production option, the hybrid policy circumvents the lack of formal property rights to the forest, and

12 Table 4 presents estimated marginal effects of Model 2, which controls for village heterogeneity by including village fixed effects and no covariates, as we showed that adding covariates has a minimal effect on the coefficient point estimates. Estimated parameters of Model 1 and Model 3 yield to the same findings, and they are available upon request.

52

M. Veronesi et al. / Ecological Economics 116 (2015) 46–57

Table 3 Policy impact on charcoal time — marginal effects. Model 1

Low price agricultural policy Charcoal price 250 Charcoal price 500 Charcoal price 750 Charcoal price 1000 Charcoal price 1500 High price agricultural policy Charcoal price 250 Charcoal price 500 Charcoal price 750 Charcoal price 1000 Charcoal price 1500 Fixed price eco-charcoal policy Charcoal price 250 Charcoal price 500 Charcoal price 750 Charcoal price 1000 Charcoal price 1500 Indexed price eco-charcoal policy No conditionality Charcoal price 250 Charcoal price 500 Charcoal price 750 Charcoal price 1000 Charcoal price 1500 Indexed price eco-charcoal policy Weak conditionality Charcoal price 250 Charcoal price 500 Charcoal price 750 Charcoal price 1000 Charcoal price 1500 Indexed price eco-charcoal policy Strong conditionality Charcoal price 250 Charcoal price 500 Charcoal price 750 Charcoal price 1000 Charcoal price 1500 Order effects Village fixed effects Covariates Number of observations Number of households

Model 2

Model 3

Marg. eff.

Std. err.

Marg. eff.

Std. err.

Marg. eff.

Std. err.

−0.466⁎ −0.708⁎⁎ −0.308 0.244 0.527⁎⁎ 1.346⁎⁎⁎ −1.248⁎⁎⁎ −1.503⁎⁎⁎ −1.077⁎⁎⁎ −0.551⁎⁎

0.273 0.309 0.254 0.229 0.244 0.363 0.261 0.294 0.244 0.221 0.235 0.377 0.330 0.377 0.308 0.300 0.362 0.592

−0.630⁎⁎ −0.853⁎⁎⁎ −0.485⁎ 0.023 0.283 1.037⁎⁎⁎ −1.508⁎⁎⁎ −1.777⁎⁎⁎ −1.329⁎⁎⁎ −0.775⁎⁎⁎ −0.415⁎ 0.698⁎ −2.707⁎⁎⁎ −3.007⁎⁎⁎ −2.497⁎⁎⁎ −2.081⁎⁎⁎ −1.519⁎⁎⁎

−0.640⁎⁎ −0.863⁎⁎⁎ −0.495⁎ 0.013 0.273 1.026⁎⁎⁎ −1.493⁎⁎⁎ −1.759⁎⁎⁎ −1.314⁎⁎⁎ −0.766⁎⁎⁎ −0.409⁎ 0.694⁎ −2.722⁎⁎⁎ −3.023⁎⁎⁎ −2.511⁎⁎⁎ −2.092⁎⁎⁎ −1.528⁎⁎⁎

−0.487

0.275 0.309 0.257 0.232 0.247 0.362 0.264 0.296 0.247 0.224 0.238 0.375 0.329 0.375 0.308 0.300 0.360 0.583

−0.491

0.274 0.308 0.256 0.231 0.245 0.360 0.263 0.294 0.246 0.224 0.237 0.373 0.328 0.373 0.307 0.299 0.358 0.581

−0.809

0.355 0.399 0.337 0.309 0.335 0.502

−2.595⁎⁎⁎ −2.798⁎⁎⁎ −2.489⁎⁎⁎ −1.981⁎⁎⁎ −1.741⁎⁎⁎ −1.035⁎⁎

0.353 0.396 0.336 0.309 0.333 0.496

−2.602⁎⁎⁎ −2.812⁎⁎⁎ −2.493⁎⁎⁎ −1.970⁎⁎⁎ −1.723⁎⁎⁎ −0.996⁎⁎

0.352 0.394 0.334 0.307 0.332 0.494

−2.660⁎⁎⁎ −2.824⁎⁎⁎ −2.538⁎⁎⁎ −2.133⁎⁎⁎ −1.975⁎⁎⁎ −1.550⁎⁎⁎

0.401 0.464 0.363 0.324 0.351 0.504

−2.856⁎⁎⁎ −2.990⁎⁎⁎ −2.758⁎⁎⁎ −2.430⁎⁎⁎ −2.302⁎⁎⁎ −1.958⁎⁎⁎

0.399 0.461 0.362 0.324 0.350 0.499

−2.850⁎⁎⁎ −2.978⁎⁎⁎ −2.756⁎⁎⁎ −2.442⁎⁎⁎ −2.319⁎⁎⁎ −1.989⁎⁎⁎

0.397 0.459 0.361 0.323 0.349 0.497

−2.490⁎⁎⁎ −2.545⁎⁎⁎ −2.448⁎⁎⁎ −2.330⁎⁎⁎ −2.263⁎⁎⁎ −2.068⁎⁎⁎

0.369 0.418 0.337 0.289 0.299 0.444

−2.637⁎⁎⁎ −2.664⁎⁎⁎ −2.617⁎⁎⁎ −2.558⁎⁎⁎ −2.525⁎⁎⁎ −2.429⁎⁎⁎

0.367 0.415 0.336 0.291 0.300 0.442

−2.635⁎⁎⁎ −2.665⁎⁎⁎ −2.612⁎⁎⁎ −2.547⁎⁎⁎ −2.510⁎⁎⁎ −2.402⁎⁎⁎

0.365 0.414 0.335 0.290 0.299 0.440

−0.208 0.850⁎⁎

−2.557⁎⁎⁎ −2.863⁎⁎⁎ −2.343⁎⁎⁎ −1.917⁎⁎⁎ −1.344⁎⁎⁎ −0.290 −2.379⁎⁎⁎ −2.584⁎⁎⁎ −2.273⁎⁎⁎ −1.762⁎⁎⁎ −1.520⁎⁎⁎

Yes No No 2850 840

Yes Yes No 2850 840

Yes Yes Yes 2850 840

Note: The dependent variable is the household weekly number of half days spent charcoaling. Marginal effects are calculated by applying Eq. (2), and using the coefficients of Table 2. Standard errors are derived by applying the delta method. ⁎ Significant at the 10% level. ⁎⁎ Significant at the 5% level. ⁎⁎⁎ Significant at the 1% level.

by indexing the price paid to opportunity costs it addresses the issue of rising and/or volatile opportunity costs. We take advantage of the setting at the world's first REDD+ project certified under verified carbon standard: the Kasigau Corridor REDD+ Project in Kenya. We implement a survey-based experiment of time allocation in different activities with a sample of customary forest-using households living adjacent to the Kasigau Corridor REDD + Project. We move beyond the analysis of conventional style ICDP or PES policies and develop innovative policy alternatives combining ICDP-like indirect approaches with aspects of conditionality and indexed payments put forth in the broader literature on payments for ecosystem services. Charcoal prices were randomly assigned to allow investigating the impact of volatile opportunity costs. The results show that hybrid approaches combining the strengths of ICDPs and PES are more promising than predominant agricultural policies that promote high revenue plants, which require little care. In particular, a payment mechanism that links the price of eco-charcoal raw material to the charcoal price significantly lowers the amount of labor allocated to charcoaling relative to the status quo even at high charcoal prices. The

unambiguous positive effect of indexing in our study, compared to more ambiguous theoretical results in Engel et al. (2014), is likely due to the fact that charcoal prices were known to respondents with certainty in our study and payments were correlated to actual charcoal prices rather than an index used as a proxy for opportunity costs. Introducing conditionality further increases the environmental effectiveness of the hybrid policy in our study, yet only when conditionality is strong and opportunity costs are high. The weak effect of conditionality at low charcoal prices could be due to the complication that joint customary use of forests by communities of households induces a commons dilemma. In general, our study provides empirical evidence that hybrid policies that target the local drivers of deforestation, provide alternative income opportunities to local people, are conditional on environmental outcomes, and account for changing opportunity costs could work as effective policy options. As Fisher et al. (2011) emphasize ‘Such targeted interventions could be considered Smart-REDD’ (p. 163). In addition, Angelsen and Rudel (2013) argue that ‘applications of REDD+ should be contextually sensitive, tailoring the types of incentives and compensation available to the forest conditions in a country’ (p. 92). Our results

M. Veronesi et al. / Ecological Economics 116 (2015) 46–57

53

Table 4 Policy impact on time spent for farm work, firewood collection, and ‘other activities’ — marginal effects. Farm work

Low price agricultural policy Charcoal price 250 Charcoal price 500 Charcoal price 750 Charcoal price 1000 Charcoal price 1500 High price agricultural policy Charcoal price 250 Charcoal price 500 Charcoal price 750 Charcoal price 1000 Charcoal price 1500 Fixed price eco-charcoal policy Charcoal price 250 Charcoal price 500 Charcoal price 750 Charcoal price 1000 Charcoal price 1500 Indexed price eco-charcoal policy No conditionality Charcoal price 250 Charcoal price 500 Charcoal price 750 Charcoal price 1000 Charcoal price 1500 Indexed price eco-charcoal policy Weak conditionality Charcoal price 250 Charcoal price 500 Charcoal price 750 Charcoal price 1000 Charcoal price 1500 Indexed price eco-charcoal policy Strong conditionality Charcoal price 250 Charcoal price 500 Charcoal price 750 Charcoal price 1000 Charcoal price 1500

Firewood collection

Other activities

Marg. eff.

Std. err.

Marg. eff.

Std. err.

Marg. eff.

Std. err.

2.345⁎⁎⁎ 2.487⁎⁎⁎ 2.253⁎⁎⁎ 1.930⁎⁎⁎ 1.764⁎⁎⁎ 1.285⁎⁎⁎ 3.050⁎⁎⁎ 3.166⁎⁎⁎ 2.972⁎⁎⁎ 2.733⁎⁎⁎ 2.577⁎⁎⁎ 2.096⁎⁎⁎ −3.080⁎⁎⁎ −2.837⁎⁎⁎ −3.250⁎⁎⁎ −3.588⁎⁎⁎ −4.044⁎⁎⁎ −4.882⁎⁎⁎

0.352 0.396 0.328 0.297 0.316 0.463 0.337 0.378 0.316 0.287 0.304 0.479 0.421 0.480 0.394 0.384 0.460 0.746

0.190 0.256 0.147 −0.004 −0.081 −0.305 0.203 0.230 0.185 0.129 0.093 −0.020 −0.957⁎⁎⁎ −0.931⁎⁎⁎ −0.975⁎⁎⁎ −1.011⁎⁎⁎ −1.059⁎⁎⁎ −1.147⁎⁎⁎

0.171 0.192 0.159 0.144 0.153 0.225 0.164 0.184 0.153 0.139 0.148 0.233 0.205 0.233 0.191 0.186 0.224 0.363

−1.701⁎⁎⁎ −1.604⁎⁎⁎ −1.765⁎⁎⁎ −1.987⁎⁎⁎ −2.101⁎⁎⁎ −2.430⁎⁎⁎ −1.966⁎⁎⁎ −1.876⁎⁎⁎ −2.026⁎⁎⁎ −2.213⁎⁎⁎ −2.334⁎⁎⁎ −2.708⁎⁎⁎ −5.339⁎⁎⁎ −4.934⁎⁎⁎ −5.623⁎⁎⁎ −6.186⁎⁎⁎ −6.946⁎⁎⁎ −8.341⁎⁎⁎

0.364 0.410 0.340 0.308 0.327 0.480 0.349 0.391 0.327 0.297 0.314 0.496 0.436 0.497 0.408 0.397 0.477 0.773

−3.709⁎⁎⁎ −3.405⁎⁎⁎ −3.868⁎⁎⁎ −4.629⁎⁎⁎ −4.989⁎⁎⁎ −6.047⁎⁎⁎

0.452 0.507 0.429 0.394 0.426 0.635

−0.974⁎⁎⁎ −0.899⁎⁎⁎ −1.012⁎⁎⁎ −1.197⁎⁎⁎ −1.284⁎⁎⁎ −1.541⁎⁎⁎

0.220 0.246 0.208 0.192 0.207 0.308

−6.772⁎⁎⁎ −6.903⁎⁎⁎ −6.705⁎⁎⁎ −6.380⁎⁎⁎ −6.227⁎⁎⁎ −5.775⁎⁎⁎

0.468 0.524 0.444 0.408 0.440 0.656

−2.902⁎⁎⁎ −2.648⁎⁎⁎ −3.091⁎⁎⁎ −3.716⁎⁎⁎ −3.961⁎⁎⁎ −4.618⁎⁎⁎

0.510 0.589 0.463 0.414 0.448 0.638

−1.014⁎⁎⁎ −0.957⁎⁎⁎ −1.056⁎⁎⁎ −1.196⁎⁎⁎ −1.251⁎⁎⁎ −1.398⁎⁎⁎

0.248 0.286 0.225 0.201 0.218 0.310

−5.352⁎⁎⁎ −5.053⁎⁎⁎ −5.573⁎⁎⁎ −6.307⁎⁎⁎ −6.594⁎⁎⁎ −7.365⁎⁎⁎

0.528 0.608 0.479 0.429 0.464 0.660

−2.235⁎⁎⁎ −2.176⁎⁎⁎ −2.279⁎⁎⁎ −2.407⁎⁎⁎ −2.478⁎⁎⁎ −2.687⁎⁎⁎

0.469 0.531 0.430 0.372 0.384 0.565

−0.981⁎⁎⁎ −0.990⁎⁎⁎ −0.974⁎⁎⁎ −0.956⁎⁎⁎ −0.945⁎⁎⁎ −0.915⁎⁎⁎

0.228 0.258 0.209 0.180 0.186 0.274

−5.075⁎⁎⁎ −4.803⁎⁎⁎ −5.280⁎⁎⁎ −5.868⁎⁎⁎ −6.196⁎⁎⁎ −7.160⁎⁎⁎

0.484 0.548 0.444 0.384 0.397 0.584

Note: The dependent variable is the household weekly number of half days spent on each activity. Marginal effects are calculated by applying Eq. (2), and estimating Model 2, which includes village fixed effects and no covariates. The number of households is 840 and the number of observations is 2850. Standard errors are derived by applying the delta method. ⁎⁎⁎ Significant at the 1% level.

are context specific and future research is required to assess their generalizability, however, in general this study indicates that designing REDD+ incentives to customary forest users could benefit by adapting ideas discussed in the literature on PES for contexts of formal ownership rights and combining them with ICDP-like approaches. Though our results are based on hypothetical self-reported time allocation, the use of hypothetical policies has been applied in many areas of economics to test preferences for multiple policy scenarios, an objective that is commonly very difficult to achieve with revealed preference data.13 In addition, an advantage of the estimation procedure applied in this study is that the randomly assigned policies and charcoal prices are independent of household and individual characteristics. This implies that our estimates are not suffering from endogeneity bias and unobserved heterogeneity. Furthermore, any systematic tendency within households to misstate the time allocation is treated as a householdspecific error component that is unrelated to prices and policies, thus yielding unbiased results.14 13 Examples of studies using hypothetical policies in the development, health, and environmental economics literature include Cropper et al. (2004), Bosworth et al. (2008), Grosjean and Kontoleon (2009), Ibanez and Carlsson (2010), Krishna et al. (2013), and Gerking et al. (2014) among others. 14 The household-specific component of the error will capture systematic tendencies within households; however, if all respondents in the survey misstate their time allocation then the results will still be biased.

Our analysis relates to the literature comparing integrated conservation and development policies to more conditional approaches like PES (e.g., Ferraro and Kiss, 2002; Ferraro and Simpson, 2002; Muller and Albers, 2004; Groom and Palmer, 2012). Our results are in line with a substantial body of literature that has demonstrated that the effectiveness of ICDP approaches alone is limited (Kremen et al., 1994; UNDP, 2000; Hughes and Flintan, 2001). The analysis of policy scenarios not linked to opportunity costs also relates to a broader literature on the impacts of alternative income opportunities on damaging activities (Bluffstone, 1995; Shively, 2001; Shively and Pagiola, 2004). In addition, Benítez et al. (2006) and Dutschke and Angelsen (2008) propose indexing conservation payments to agricultural commodity prices, yet theoretically, the impact of indexing on environmental and cost effectiveness is ambiguous (Engel et al., 2014). Our hybrid approach is an innovative solution to decrease deforestation in a context where land users are not landowners. We emphasize that one could envision the hybrid approach as applied to other settings. For example, Cranford and Mourato (2014) propose, for a different setting, credit-based payments for ecosystem services, that is a hybrid microcredit instrument where microcredit repayment rates are made conditional on the adoption of agro-forestry. They show that this hybrid policy is a promising approach for incentivizing the provision of environmental services. Finally, before bringing eco-charcoal to such a scale, additional studies are needed (i) to quantify the ecological carrying capacity of

54

M. Veronesi et al. / Ecological Economics 116 (2015) 46–57

the eco-charcoal raw material; (ii) to test whether the effectiveness of the eco-charcoal policies prevails also over longer time periods; (iii) to test whether there are general equilibrium effects that we could not capture in our study such as the possibility that increasing returns to agriculture through agricultural policy could lead to more land being deforested, or improved income and labor opportunities drawing more people into the area; and (iv) to test the presence of leakage, i.e., the possibility that the damaging activity (charcoaling) will be relocated elsewhere to satisfy energy demands. We speculate that while the mere reduction of charcoaling as a consequence of an agricultural policy may induce leakage, eco-charcoaling is likely to be superior in that it provides a sustainable alternative energy source, however, this question is left for future research. Last but not least, additional studies are needed to test whether the effectiveness of indexed payment policies that are conditional on forest degradation depends on monitoring intensity and households' expectation of the effective eco-charcoal price they will receive. The latter in turn depends on other households' behavior regarding deforestation. Households from communities with better collective action potential might expect

a higher price and reduce charcoaling more. Because of missing information on collective action in our dataset, we leave it for future research to test these hypotheses. Acknowledgments The authors would like to thank for their comments the participants at seminars at the London School of Economics and Political Science, University of Basel, University of Chieti-Pescara, and at the conference ‘Payments for ecosystem services and their institutional dimensions’ in Berlin, at the 19th Conference of the European Association in Environmental and Resource Economics in Prague, and at the AEL Conference in Bonn. We would like to also thank Saraly Andrade De Sa and Adrian Müller for their comments on a preliminary draft; Jerusha Magenyi for excellent field assistance and survey management; Mike Korchinsky and the Wild Life Works Carbon Ltd. team for their support and open access to all relevant information on the project. We acknowledge the financial support of ETH Global (formerly North-South Centre). All errors and omissions are our responsibility.

Appendix A

Fig. A1. Example of policy scenarios and experimental instructions.

M. Veronesi et al. / Ecological Economics 116 (2015) 46–57

55

Table A2 Policy impact on charcoal time – marginal effects – original sample. Model 1

Fig. A2. Visual aids used in the bean game on time allocation.

Table A1 Descriptive statistics — final sample vs. original sample. (1)

(2)

(3)

Final sample (N = 850)

Original sample (N = 1095)

t-Statistic (1) ≠ (2)

Mean

Mean

Std. dev.

Dependent variables (half days per week) Charcoaling 3.912 6.289 Farm work 10.818 9.076 Fire wood collection 3.815 4.180 Other activities 23.588 17.745

4.017 10.854 3.998 23.041

6.709 9.489 4.347 18.035

−0.355 −0.085 −0.941 0.670

Control variables Household size Share of male adults Share of female adults Share of elderly Share of kids 0–11 Share of kids 12–17 Land owned (Ha) Number of livestock Village forest distance (km) Village market distance (km)

5.101 0.285 0.251 0.023 0.238 0.113 5.248 11.712 3.886 14.274

3.329 0.251 0.187 0.092 0.226 0.158 5.021 39.924 6.953 14.862

−0.971 −1.946 0.471 0.238 0.476 0.137 −0.368 −1.053 0.021 −0.114

4.968 0.281 0.255 0.024 0.243 0.114 5.165 10.362 3.893 14.197

Std. dev.

2.710 0.245 0.187 0.092 0.231 0.160 4.854 12.617 7.317 14.736

Low price agricultural policy Charcoal price 250 Charcoal price 500 Charcoal price 750 Charcoal price 1000 Charcoal price 1500 High price agricultural policy Charcoal price 250 Charcoal price 500 Charcoal price 750 Charcoal price 1000 Charcoal price 1500 Fixed price eco-charcoal policy Charcoal price 250 Charcoal price 500 Charcoal price 750 Charcoal price 1000 Charcoal price 1500 Indexed price eco-charcoal policy No conditionality Charcoal price 250 Charcoal price 500 Charcoal price 750 Charcoal price 1000 Charcoal price 1500 Indexed Price Eco-charcoal Policy Weak conditionality Charcoal price 250 Charcoal price 500 Charcoal price 750 Charcoal price 1000 Charcoal price 1500

Model 2

Model 3

Marg. eff.

Std. err.

Marg. eff.

Std. err.

Marg. eff.

Std. err.

−0.483⁎

0.254

−0.739⁎⁎⁎

0.256

−0.772⁎⁎⁎

0.255

−0.698⁎⁎

0.288

−0.937⁎⁎⁎

0.289

−0.979⁎⁎⁎

0.288

−0.339

0.235

−0.607⁎⁎

0.238

−0.634⁎⁎⁎

0.237

0.126

0.210

−0.179

0.214

−0.186

0.214

0.384⁎⁎

0.224

0.058

0.227

0.063

0.226

1.093⁎⁎⁎

0.332

0.709⁎⁎

0.332

0.745⁎⁎

0.331

−1.170⁎⁎⁎

0.243

−1.485⁎⁎⁎

0.246

−1.473⁎⁎⁎

0.245

−1.441⁎⁎⁎

0.275

−1.760⁎⁎⁎

0.278

−1.747⁎⁎⁎

0.276

−0.993⁎⁎⁎

0.226

−1.305⁎⁎⁎

0.230

−1.294⁎⁎⁎

0.229

−0.431⁎⁎

0.205

−0.733⁎⁎⁎

0.209

−0.725⁎⁎⁎

0.209

−0.106

0.220

−0.403⁎

0.223

−0.397⁎

0.222

0.814⁎⁎

0.333

0.533

0.332

0.534⁎

0.331

−2.302⁎⁎⁎

0.307

−2.580⁎⁎⁎

0.307

−2.576⁎⁎⁎

0.305

−2.530⁎⁎⁎

0.348

−2.807⁎⁎⁎

0.346

−2.803⁎⁎⁎

0.345

−2.148⁎⁎⁎

0.287

−2.427⁎⁎⁎

0.287

−2.424⁎⁎⁎

0.286

−1.833⁎⁎⁎

0.272

−2.114⁎⁎⁎

0.272

−2.111⁎⁎⁎

0.271

−1.345⁎⁎⁎

0.319

−1.629⁎⁎⁎

0.318

−1.626⁎⁎⁎

0.317 0.504

−0.526

0.512

−0.815

0.506

−0.814⁎

−2.541⁎⁎⁎

0.331

−2.783⁎⁎⁎

0.330

−2.743⁎⁎⁎

0.329

−2.732⁎⁎⁎

0.371

−2.969⁎⁎⁎

0.369

−2.914⁎⁎⁎

0.368

−2.455⁎⁎⁎

0.316

−2.699⁎⁎⁎

0.316

−2.666⁎⁎⁎

0.314

−2.062⁎⁎⁎

0.282

−2.315⁎⁎⁎

0.283

−2.312⁎⁎⁎

0.281

−1.734⁎⁎⁎

0.306

−1.995⁎⁎⁎

0.305

−2.018⁎⁎⁎

0.304

−1.062⁎⁎

0.456

−1.340⁎⁎

0.451

−1.415⁎⁎

0.449

−2.320⁎⁎⁎

0.372

−2.685⁎⁎⁎

0.374

−2.704⁎⁎⁎

0.372

−2.418⁎⁎⁎

0.431

−2.789⁎⁎⁎

0.432

−2.807⁎⁎⁎

0.431

−2.241⁎⁎⁎

0.333

−2.599⁎⁎⁎

0.335

−2.620⁎⁎⁎

0.334

−2.031⁎⁎⁎

0.300

−2.374⁎⁎⁎

0.302

−2.398⁎⁎⁎

0.301

−1.945⁎⁎⁎

0.320

−2.281⁎⁎⁎

0.321

−2.307⁎⁎⁎

0.320

−1.641⁎⁎⁎

0.497

−1.955⁎⁎⁎

0.494

−1.985⁎⁎⁎

0.492

56

M. Veronesi et al. / Ecological Economics 116 (2015) 46–57

Table A2 (continued) Model 1

Indexed price eco-charcoal policy Strong conditionality Charcoal price 250 Charcoal price 500 Charcoal price 750 Charcoal price 1000 Charcoal price 1500 Village fixed effects Covariates Number of observations Number of households

Model 2

Model 3

Marg. eff.

Std. err.

Marg. eff.

Std. err.

Marg. eff.

Std. err.

−2.551⁎⁎⁎

0.339

−2.812⁎⁎⁎

0.338

−2.789⁎⁎⁎

0.337

−2.582⁎⁎⁎

0.388

−2.811⁎⁎⁎

0.385

−2.784⁎⁎⁎

0.384

−2.535⁎⁎⁎

0.317

−2.813⁎⁎⁎

0.317

−2.791⁎⁎⁎

0.316

−2.465⁎⁎⁎

0.270

−2.815⁎⁎⁎

0.272

−2.803⁎⁎⁎

0.271

−2.431⁎⁎⁎

0.283

−2.816⁎⁎⁎

0.284

−2.808⁎⁎⁎

0.283

−2.326⁎⁎⁎

0.430

−2.820⁎⁎⁎

0.428

−2.825⁎⁎⁎

0.426

No

Yes

Yes

No 3276

No 3276

Yes 3276

1095

1095

1095

Note: The dependent variable is the household weekly number of half days spent charcoaling. Marginal effects are calculated by applying Eq. (2). Standard errors are derived by applying the delta method. ⁎ Significant at the 10% level. ⁎⁎ Significant at the 5% level. ⁎⁎⁎ Significant at the 1% level.

References Angelsen, A., 2008. Moving Ahead with REDD: Issues, Options and Implications. CIFOR, Bogor, Indonesia. Angelsen, A., Rudel, T.K., 2013. Designing and implementing effective REDD+ policies: a forest transition approach. Rev. Environ. Econ. Policy 7 (1), 91–113. Benítez, P., Kuosmanen, T., Olschewski, R., van Kooten, G., 2006. Conservation payments under risk: a stochastic dominance approach. Am. J. Agric. Econ. 88 (1), 1–15. Biorn, E., 2004. Regression system for unbalanced panel data: a stepwise maximum likelihood procedure. J. Econ. 122, 281–291. Bluffstone, R.A., 1995. The effect of labor market performance on deforestation in developing countries under open access: an example form rural Nepal. J. Environ. Econ. Manag. 29, 42–63. Bosworth, R.C., Cameron, T.A., DeShazo, J.R., 2008. Demand for environmental policies to improve health: evaluating community-level policy scenarios. J. Environ. Econ. Manag. 57 (3), 393-308. Brown, D., Seymour, F., Peskett, L., 2009. How do we achieve REDD co-benefits and avoid doing harm? In: Angelsen, A., et al. (Eds.), Realising REDD+: National Strategy and Policy Options. Centre for International Forestry Research (CIFOR), Bogor, Indonesia, pp. 107–118 Butler, R., Koh, L., Ghazoul, J., 2009. REDD in the red: palm oil could undermine carbon payment schemes. Conserv. Lett. 2 (2), 67–73. Chhatre, A., Agrawal, A., 2009. Trade-offs and synergies between carbon storage and livelihood benefits from forest commons. PNAS 106 (42), 17667–17670. Chomitz, K.M., Buys, P., De Luca, G., Thomas, T.S., Wertz-Kanounnikoff, S., 2006. At loggerheads? Agricultural expansion, poverty reduction and environment in the tropical forests. World Bank Policy Research Report. Development Research Group, World Bank, Washington D.C. Corbera, E., Kosoy, N., Tuna, M.M., 2007. Equity implications of marketing ecosystem services in protected areas and rural communities: case studies from Meso-America. Glob. Environ. Chang. 17, 365–380. Cranford, M., Mourato, S., 2014. Credit-based payments for ecosystem services: evidence from a choice experiment in Ecuador. World Dev. 64, 503–520. Cropper, M., Hailec, M., Lampietti, J., Poulos, C., Whittington, D., 2004. The demand for a malaria vaccine: evidence from Ethiopia. J. Dev. Econ. 75, 303–318. Day, B., Bateman, I.J., Carson, R.T., Dupont, D., Louviere, J.J., Morimoto, S., Scarpa, R., Wang, P., 2012. Ordering effects and choice set awareness in repeat-response stated preference studies. J. Environ. Econ. Manag. 63, 73–91. Delavande, A., Giné, X., McKenzie, D., 2011. Measuring subjective expectations in developing countries: a critical review and new evidence. J. Dev. Econ. 94 (2), 151–163. Duchelle, A.E., Cromberg, M., Gebara, M.F., Guerra, R., Melo, T., Larson, A., et al., 2014. Linking forest tenure reform, environmental compliance and incentives: lessons from REDD+ initiatives in the Brazilian Amazon. World Dev. 55, 53–67. Dutschke, M., Angelsen, A., 2008. How do we ensure permanence and assign liability? In: Angelsen, A. (Ed.), Moving Ahead With REDD: Issues, Options and Implications. Center for International Forestry Research. CIFOR, Bogor, Indonesia

Dutschke, M., Wong, J., 2003. Can permanence be insured? Consideration of some technical and practical issues of insuring carbon credits from afforestation and reforestation. HWWA Discussion Paper No. 235. Engel, S., Palmer, C., 2008. Payments for environmental services as an alternative to logging under weak property rights: the case of Indonesia. Ecol. Econ. 65, 799–809. Engel, S., Palmer, C., 2009. Designing payments for environmental services with weak property rights and external interests. In: Lipper, L., Sakuyama, T., Stringer, R., Zilberman, D. (Eds.), Payments for Environmental Services in Agricultural Landscapes: Economic Policies and Poverty Reduction in Developing Countries. Food and Agriculture Organization of the United Nations (FAO) and Springer Press. Engel, S., Pagiola, S., Wunder, S., 2008. Designing payments for environmental services in theory and practice: an overview of the issues. Ecol. Econ. 65 (4), 663–674. Engel, S., Palmer, C., Pfaff, A., 2013. On the endogeneity of resource co-management: theory and evidence from Indonesia. Land Econ. 89 (2), 308–329. Engel, S., Palmer, C., Taschini, L., Urech, S., 2014. Conservation payments under uncertainty when nonuse benefits have market value. Land Econ. 91 (1), 36–56. Fafchamps, M., Quisumbing, A.R., 1999. Human capital, productivity, and labor allocation in rural Pakistan. J. Hum. Resour. 34 (2), 369–406. Farley, J., Costanza, R., 2010. Payments for ecosystem services: from local to global. Ecol. Econ. 69 (11), 2060–2068. Ferraro, P.J., 2001. Global habitat protection: limitations of development interventions and a role for conservation performance payments. Conserv. Biol. 15 (4), 990–1000. Ferraro, P.J., Kiss, A., 2002. Direct payments to conserve biodiversity. Science 298, 1718–1719. Ferraro, P.J., Simpson, R.D., 2002. The cost-effectiveness of conservation performance payments. Land Econ. 78 (3), 339–353. Fisher, M., Shively, G.E., Buccola, S., 2005. Activity choice, labor allocation, and forest use in Malawi. Land Econ. 81 (4), 503–517. Fisher, B., Lewis, S.L., Burgess, N.D., Malimbwi, R.E., Munishi, P.K., Swtnam, R.D., et al., 2011. Implementation and opportunity costs of reducing deforestation and forest degradation in Tanzania. Nat. Clim. Chang. 1, 161–164. Gerking, S., Dickie, M., Veronesi, M., 2014. Valuation of human health: an integrated model of willingness to pay for mortality and morbidity risk reductions. J. Environ. Econ. Manag. 68 (1), 20–45. Ghazoul, J., Butler, R.A., Mateo-Vega, J., Koh, L.P., 2010. REDD: a reckoning of environment and development implications. Trends Ecol. Evol. 25 (7), 396–402. Groom, B., Palmer, C., 2012. REDD+ and rural livelihoods. Biol. Conserv. 154, 42–52. Grosjean, P., Kontoleon, A., 2009. How sustainable are sustainable development programs? The case of the sloping land conversion program in China. World Dev. 37, 268–285. Holland, M.B., Koning, F.D., Morales, M., Naughton-Treves, L., Robinson, B., Suàrez, L., 2014. Complex tenure and deforestation: implications for conservation incentives in the Ecuadorian Amazon. World Dev. 55, 21–36. Hosonuma, N., Herold, M., De Sy, V., De Fries, R.S., Brockhaus, M., Verchot, L., Angelsen, A., Romijn, E., 2012. An assessment of deforestation and forest degradation drivers in developing countries. Environ. Res. Lett. 7 (2), 1–12. Hughes, R., Flintan, F., 2001. Integrating Conservation and Development Experience: A Review and Bibliography of the ICDP Literature. International Institute for Environment and Development, London. Ibanez, M., Carlsson, F., 2010. A survey-based choice experiment on coca cultivation. J. Dev. Econ. 93, 249–263. Ito, T., Kurosaky, T., 2009. Weather risk, wages in kind, and the off-farm labor supply of agricultural households in a developing country. Am. J. Agric. Econ. 91 (3), 697–710. IUCN, 2010. Indigenous Peoples and REDD-plus: Challenges and Opportunities for the Engagement of Indigenous Peoples and Local Communities in REDD-plus. International Union for Conservation of Nature, Washington DC. Kosoy, N., Corbera, E., 2010. Payments for ecosystem services as commodity fetishism. Ecol. Econ. 69, 1228–1236. Kremen, C., Merenlender, A.M., Murphy, D.D., 1994. Ecological monitoring: a vital need for integrated conservation and development programs in the tropics. Conserv. Biol. 8 (6), 388–397. Krishna, V.V., Drucker, A.G., Pascual, U., Raghu, P.T., King, E.D.I.O., 2013. Estimating compensation payments for on-farm conservation of agricultural biodiversity in developing countries. Ecol. Econ. 87, 110–123. Matzdorf, B., Sattler, C., Engel, S., 2013. Institutional frameworks and governance structures of PES schemes. For. Policy Econ. 37, 57–64. Muller, J., Albers, H., 2004. Enforcement, payments, and development projects near protected areas: how the market setting determines what works where. Resour. Energy Econ. 26 (2), 185–204. Muradian, et al., 2013. Payments for ecosystem services and the fatal attraction of win– win solutions. Conserv. Lett. 6 (4), 274–279. Pagiola, S., Arcenas, A., Platais, G., 2005. Can payments for environmental services help reduce poverty? An exploration of the issues and the evidence to date from Latin America. World Dev. 33 (2), 237–253. Palmer, C., Engel, S., 2009. Introduction: reducing CO2 emissions through avoided deforestation? In: Palmer, C., Engel, S. (Eds.), Avoided deforestation: prospects for mitigating climate change. Routledge Explorations in Environmental Economics, Routledge, Oxford Palmer, C., Silber, T., 2010. Trade-offs between carbon sequestration and poverty alleviation in community-based land use systems: preliminary evidence from Mozambique. Res. Pap. Spat. Environ. Anal. 130. Pascual, U., Corbera, E., 2011. Payments for environmental services: perspectives and innovative experiences for nature conservation and rural development. Rev. Esp. Estud. Agrosoc. Pesq. 228 (01/11/2029). Pattanayak, S.K., Wunder, S., Ferraro, P.J., 2010. Show me the money: do payments supply environmental services in developing countries? Rev. Environ. Econ. Policy 4 (2), 254–274.

M. Veronesi et al. / Ecological Economics 116 (2015) 46–57 Peters-Stanley, M., Hamilton, K., Marcello, T., Sjardin, M., 2011. Back to the future. State of the Voluntary Carbon Markets 2011. Ecosystem Marketplace and Bloomberg New Energy Finance Publishers. Pfaff, A., Robalino, J., Lima, E., Sandoval, C., Herrera, L.D., 2014. Governance, location and avoided deforestation from protected areas: greater restrictions can have lower impact, due to differences in location. World Dev. 55, 7–20. Pistorius, T., 2012. From RED to REDD+: the evolution of a forest-based mitigation approach for developing countries. Curr. Opin. Environ. Sustain. 4, 638–645. Resosudarmo, I.A.P., Atmadja, S., Ekaputri, A.D., Intarini, D.Y., Indriatmoko, Y., Astri, P., 2014. Does tenure security lead to REDD+ project effectiveness? Reflections from five emerging sites in Indonesia. World Dev. 55, 68–83. Reutemann, T., Engel, S., Pareja, E., 2014. Designing payments for avoided deforestation under a capital constraint — a framed lab-in the-field experiment. Paper Presented at the 5th World Congress of Environmental and Resource Economists, Istanbul, 28 June–02 July, 2014. Schomers, S., Matzdorf, B., 2013. Payments for ecosystem services: a review and comparison of developing and industrialized countries. Ecosyst. Serv. 6, 16–30. Shively, G.E., 2001. Agricultural change, rural labor markets, and forests clearing: an illustrative case from the Philippines. Land Econ. 77 (2), 268–284. Shively, G., Fisher, M., 2004. Smallholder labor and deforestation: a system approach. Am. J. Agric. Econ. 86 (5), 1361–1366.

57

Shively, G., Pagiola, S., 2004. Agricultural intensification, local labour markets, and deforestation in the Philippines. Environ. Dev. Econ. 9 (2), 241–266. Sunderlin, W.D., Angelsen, A., Belcher, B., Burgers, P., Nasi, R., Santoso, L., Wunder, S., 2005. Livelihoods, forests, and conservation in developing countries: an overview. World Dev. 33 (9), 1383–1402. Sunderlin, W.D., Larson, A.M., Duchelle, A.E., Resosudarmo, I.A.B., Huynh, T.B., Awono, A., et al., 2014. How are REDD+ proponents addressing tenure problems? Evidence from Brazil, Cameroon, Tanzania, Indonesia, and Vietnam. World Dev. 55, 68–83. UNDP, 2000. Proceedings of Integrated Conservation and Development Projects Lessons Learned Workshop. June 12–13, 2000, accessed on March 03, 2015. http://www.vn. undp.org/content/dam/vietnam/docs/Publications/4283_icdp0600e.pdf. Vatn, A., 2010. An institutional analysis of payments for environmental services. Ecol. Econ. 69 (6), 1245–1252. World Resources Institute, 2005. World Resources 2005: The Wealth of the Poor Managing Ecosystems to Fight Poverty, in Collaboration with United Nations Development Programme, United Nations Environment Programme, and World Bank. WRI, Washington D.C. Zabel, A., Bostedt, G., Engel, S., 2014. Performance payments for groups: the case of carnivore conservation in Northern Sweden. Environ. Resour. Econ. 59, 613–631.