Promoting sustainable land use choices in Indonesia: Experimental evidence on the role of changing mindsets and structural barriers

Romero, Miriam; Wollni, Meike; Rudolf, Katrin; Asnawi, Rosyani; Irawan, Bambang Working Paper Promoting sustainable land use choices in Indonesia: Experimental evidence on the role of changing mindsets and structural barriers EFForTS Discussion Paper Series, No. 25 Provided in Cooperation with: Collaborative Research Centre 990: Ecological and Socioeconomic Functions of Tropical Lowland Rainforest Transformation Systems (Sumatra, Indonesia), University of Goettingen Suggested Citation: Romero, Miriam; Wollni, Meike; Rudolf, Katrin; Asnawi, Rosyani; Irawan, Bambang (2019) : Promoting sustainable land use choices in Indonesia: Experimental evidence on the role of changing mindsets and structural barriers, EFForTS Discussion Paper Series, No. 25, GOEDOC, Dokumenten- und Publikationsserver der Georg-August-Universität, Göttingen, https://nbn-resolving.de/urn:nbn:de:gbv:7-webdoc-3990-7 This Version is available at: https://hdl.handle.net/10419/195913 Standard-Nutzungsbedingungen: Die Dokumente auf EconStor dürfen zu eigenen wissenschaftlichen Zwecken und zum Privatgebrauch gespeichert und kopiert werden. Sie dürfen die Dokumente nicht für öffentliche oder kommerzielle Zwecke vervielfältigen, öffentlich ausstellen, öffentlich zugänglich machen, vertreiben oder anderweitig nutzen. Sofern die Verfasser die Dokumente unter Open-Content-Lizenzen (insbesondere CC-Lizenzen) zur Verfügung gestellt haben sollten, gelten abweichend von diesen Nutzungsbedingungen die in der dort genannten Lizenz gewährten Nutzungsrechte. Terms of use: Documents in EconStor may be saved and copied for your personal and scholarly purposes. You are not to copy documents for public or commercial purposes, to exhibit the documents publicly, to make them publicly available on the internet, or to distribute or otherwise use the documents in public. If the documents have been made available under an Open Content Licence (especially Creative Commons Licences), you may exercise further usage rights as specified in the indicated licence. https://creativecommons.org/licenses/by-nc-nd/4.0/ GOEDOC - Dokumenten- und Publikationsserver der Georg-August-Universität Göttingen 2019 Promoting sustainable land use choices in Indonesia - experimental evidence on the role of changing mindsets and structural barriers Miriam Romero, Meike Wollni, Katrin Rudolf, Rosyani Asnawi, Bambang Irawan EFForTS discussion paper series Nr. 25 Romero, Miriam; Wollni, Meike; Rudolf, Katrin; Asnawi, Rosyani; Irawan, Bambang: Promoting sustainable land use choices in Indonesia :experimental evidence on the role of changing mindsets and structural barriers Göttingen : GOEDOC, Dokumenten- und Publikationsserver der Georg-August-Universität, 2019 (EFForTS discussion paper series 25) Verfügbar: PURL: http://resolver.sub.uni-goettingen.de/purl/?webdoc-3990 This work is licensed under a Creative Commons Attribution 4.0 International License Bibliographische Information der Deutschen Nationalbibliothek Die Deutsche Nationalbibliothek verzeichnet diese Publikation in der Deutschen Nationalbibliographie; detaillierte bibliographische Daten sind im Internet über <http://dnb.dnb.de> abrufbar. Erschienen in der Reihe EFForTS discussion paper series ISSN: 2197-6244 Herausgeber der Reihe SFB 990 EFForTS, Ecological and Socioeconomic Functions of Tropical Lowland Rainforest Transformation Systems (Sumatra, Indonesien) - Ökologische und sozioökonomische Funktionen tropischer Tieflandregenwald-Transformationssysteme (Sumatra, Indonesien) Georg-August-Universität Göttingen Johann-Friedrich-Blumenbach Institut für Zoologie und Anthropologie, Fakultät für Biologie und Psychologie Abstract: This study evaluates the effects of two environmental policy instruments on the adoption of native tree planting in oil palm plantations. The first instrument is an information campaign on tree planting in oil palm. The second instrument combines the information campaign with a structural intervention that provides native tree seedlings for free. We implemented a randomized controlled trial in oil palm growing villages in Jambi, Indonesia. Our study addresses the underlying mechanisms of behavioral change, by investigating how the policy instruments shape farmers’ perceptions, intentions and actual adoption decisions. The results show that information campaigns and structural interventions can motivate tree planting among smallholder oil palm farmers in Indonesia. While both treatments have a positive and significant effect, the intervention combining information with seedling provision leads to significantly higher adoption rates, indicating that overcoming structural barriers is critical. While changes in perceptions and intentions fully mediate the effect of the information campaign on adoption, they can only partially explain the effect of the combined intervention. Thus, to promote a transition towards more sustainable development pathways, facilitating easy access to critical inputs may be key to motivate adoption among large numbers of potential users. Keywords: tree-planting; oil palm; intentions; mediation; Asia Promoting sustainable land use choices in Indonesia: experimental evidence on the role of changing mindsets and structural barriers Miriam Romero, Meike Wollni, Katrin Rudolf, Rosyani Asnawi, Bambang Irawan EFForTS Discussion Paper Series No. 25 March 2019 This publication was funded by the Deutsche Forschungsgemeinschaft (DFG, German Research Foundation) – project number 192626868 – in the framework of the collaborative German-Indonesian research project CRC 990 (SFB): “EFForTS, Ecological and Socioeconomic Functions of Tropical Lowland Rainforest Transformation Systems (Sumatra, Indonesia)” https://www.uni-goettingen.de/de/about+us/413417.html SFB 990, University of Goettingen Untere Karspüle 2, D-37073 Goettingen, Germany ISSN: 2197-6244 Managing editors: at the University of Goettingen, Germany Prof. Dr. Heiko Faust, Faculty of Geoscience and Geography, Division of Human Geography (Email: h[email protected]) Dr. Jana Juhrbandt, Environmental and Resource Economics, Department for Agricultural Economics and Rural Development (Email: jjuh[email protected]) at the University of Jambi, Indonesia Prof. Dr. Zulkifli Alamsyah, Faculty of Agriculture, Dept. of Agricultural Economics (Email: [email protected].id) Promoting sustainable land use choices in Indonesia: experimental evidence on the role of changing mindsets and structural barriers Miriam Romero1, Meike Wollni1,*, Katrin Rudolf1, Rosyani Asnawi2, Bambang Irawan2 1 Department of Agricultural Economics and Rural Development, Georg-August-University of Goettingen, Platz der Göttinger Sieben 5, 37073 Goettingen, Germany 2 University of Jambi, Indonesia * Corresponding author, email: [email protected] , tel:+49-551-3924843 Abstract This study evaluates the effects of two environmental policy instruments on the adoption of native tree planting in oil palm plantations. The first instrument is an information campaign on tree planting in oil palm. The second instrument combines the information campaign with a structural intervention that provides native tree seedlings for free. We implemented a randomized controlled trial in oil palm growing villages in Jambi, Indonesia. Our study addresses the underlying mechanisms of behavioral change, by investigating how the policy instruments shape farmers’ perceptions, intentions and actual adoption decisions. The results show that information campaigns and structural interventions can motivate tree planting among smallholder oil palm farmers in Indonesia. While both treatments have a positive and significant effect, the intervention combining information with seedling provision leads to significantly higher adoption rates, indicating that overcoming structural barriers is critical. While changes in perceptions and intentions fully mediate the effect of the information campaign on adoption, they can only partially explain the effect of the combined intervention. Thus, to promote a transition towards more sustainable development pathways, facilitating easy access to critical inputs may be key to motivate adoption among large numbers of potential users. Keywords: tree-planting; oil palm; intentions; mediation; Asia 1 1 Introduction In many tropical regions land conversion from tropical rainforest and other diverse and complex land-use systems into monoculture plantations is progressing rapidly. In Southeast Asia for example, oil palm cultivation is expanding at the cost of tropical lowland rainforest and traditional land use systems, like rubber agroforest, leading to the homogenization of landscapes (Carter et al., 2007; Burgess et al., 2012; Corley and Tinker, 2016). In Indonesia, the area cultivated with oil palm increased 106 fold to about 9 million hectares between 1961 and 2016 (FAOSTAT, 2018) and current investment plans of the government foresee further expansion (Coordinating Ministry of Economic Affairs, 2011). Land conversion towards oil palm is increasingly driven by independent smallholder farmers, to whom oil palm has brought increases in welfare and food security (Euler et al., 2015). However, the conversion of land towards homogenous structures leads to the degradation of important ecosystem functions and an unprecedented loss of tropical biodiversity (Foley et al., 2005; Fitzherbert et al., 2008). Identifying more biodiversity-friendly oil palm management options is therefore considered critical to reconcile economic and ecological functions in tropical lowland regions. Recent research has shown that biodiversity enrichment can restore important ecosystem functions in monoculture systems (Klasen et al., 2016; Teuscher et al., 2016). 1 Biodiversity enrichment refers to the integration of native tree species in existing oil palm plantations and has been shown to increase abundance and diversity of birds and invertebrate communities at the plantation scale (Teuscher et al. 2016; Teuscher et al. 2015). While these positive externalities accrue to society at large, potential costs of lower oil palm yields or revenues are borne by the farmer. On the other hand, lower oil palm yields may be compensated for by benefits derived from the trees, including timber and fruits (Teuscher et al., 2015). Only few studies have examined which policy instruments effectively change behavior towards tree planting among landholders in developing countries. Given the positive externalities generated by trees, most of the experimental studies focus on the effects of Payments for Ecosystem Services on tree planting (Leimona, Joshi and van Noordwijk, 2009; Cole, Holl and Zahawi, 2010; Jack et al., 2013). These studies thus only shed light on the role of financial rewards, whereas evidence on other policy instruments is scarce. However, as experimental studies on agricultural technology adoption have shown, alternative instruments like information and input provision can be important especially during early stages of technology diffusion (Carter, Laajaj and Yang, 2013; Duflo, Kremer and Robinson, 2011). Furthermore, there is a lack of research on the underlying mechanisms driving observed changes in behavior. Previous research has provided descriptive evidence for the link between attitudes, beliefs and tree planting behavior. Meijer et al. (2015) find that positive attitudes and intentions are associated with a higher probability of actual tree planting among Malawian farmers. Zubair & Garforth (2006) and Ndayambaje et al. (2012) provide evidence that tree planting decisions in Pakistan and Rwanda are driven by expected economic gains, rather than by perceived environmental benefits. These studies are however based on cross- 1 Earlier research on biodiversity conservation in oil palm plantations also shows that the management of ground vegetation, conservation of forest fragments inside the plantation or having forest at the edge of the plantation have positive effects on species richness (Fitzherbert et al., 2008; Koh and Wilcove, 2008; Edwards et al., 2010; Azhar et al., 2015). 2 sectional data and thus cannot derive conclusions about the drivers of attitudinal change and the association between attitudinal and behavioral change. The current study aims to fill these research gaps by evaluating the impact of two policy interventions on perceptions towards tree planting, intention to plant trees and actual tree planting behavior. The first intervention consists of an information campaign that aims at filling knowledge gaps and changing mindsets of farmers. The second intervention combines the information campaign with the provision of native tree seedlings to farmers, thereby addressing structural barriers of missing seed markets. With our study we contribute to the scarce experimental literature field-testing the effects of non-monetary policy instruments on actual behavior. Our main contribution is that we address the underlying mechanisms of behavioral change by investigating how the policy instruments shape farmers’ perceptions, intentions and actual adoption decisions. To evaluate these policy instruments, we implemented a randomized controlled trial in oil palm growing villages in Jambi province, Indonesia. Jambi is a biodiversity hotspot and characterized by rapid expansion of intensively managed oil palm plantations, which is to a large extent driven by independent smallholder farmers (Gatto et al., 2017). 2 Conceptual framework Conversion to oil palm monoculture plantations is mainly driven by profitability considerations (Clough et al., 2016). Thus, to promote more biodiversity-friendly land use systems, it seems straightforward to provide economic incentives that shift relative profitability (Table 1). This can be achieved, e.g., through payments for ecosystem services or subsidies for certain land-use types. The challenge is, however, that the profitability of tropical cash crops is high, thus requiring substantial funds in order to achieve a tangible impact (Butler, Koh and Ghazoul, 2009). In addition, palm oil prices are subject to fluctuations in world markets – which makes setting adequate incentives for conservation difficult. Due to the high priority given to profitability considerations, financial incentives and compensation will most likely be an important component in a toolbox of incentive mechanisms to achieve more sustainable land use. Nonetheless, given the challenges, it is critical to consider the role of other mechanisms as well. Table 1 Policy instruments to change behavior Mechanism Instruments Examples Shifting relative profitability Provide financial rewards and compensation PES, subsidies for diverse land-use types, certification and price premiums Changing mindsets Raise awareness, provide how- to and principles knowledge, create social consensus Information campaigns, tv and radio, extension Overcoming structural barriers Facilitate access to resources Provision of seed material, technical assistance, credit for conservation activities 3 Studies investigating pro-environmental behavior have shown that it is determined by intrinsic factors (e.g. motivations, moral values, attitudes) and by the external environment in which the behavior is performed (Steg et al., 2014). Thus, changing mindsets and overcoming structural barriers can be critical components of a strategy to induce more environmentally friendly behavior (Table 1). According to socio-psychological theory, the adoption decision is a cognitive process shaped by knowledge, information exposure and contextual factors (Ajzen, 1991; Steg et al., 2014). Social-psychology theories suggest that the antecedent knowledge an individual has about the benefits, use, and cost of a technology shape perceptions and intentions and eventually drive adoption (Rogers, 1983; Sood et al., 2004; Hansson, Ferguson and Olofsson, 2012; Ndayambaje, Heijman and Mohren, 2012; Klöckner, 2013; Meijer et al., 2015; Campos et al., 2017). Thus, informing people about the consequences of their actions, providing them with alternatives, and creating social consensus are important measures that can alter mindsets. These measures can induce more sustainable land use through changing people’s perceptions and intentions, which are then translated into actions. However, in some environments access to resources that are necessary to implement biodiversity-friendly land use systems may be limited or lacking completely. In this context, individuals may be constrained by the costs and structural barriers associated with adoption (Bamberg, 2003; Steg and Vlek, 2009). Thus, measures that aim to overcome structural barriers should influence action directly by removing the existing constraints. Figure 1 illustrates these relationships. According to Steg & Vlek (2009) and Meijer et al. (2014) the characteristics of the decision-maker, the environment and the technology to be implemented create knowledge, new experiences and perceptions that in turn will shape intentions and eventually actual behavior. We hypothesize that information provision can induce a positive and significant change in perceptions, intentions, and actual behavior. We further expect the effect of information provision on actual adoption behavior to be fully mediated by changes in farmers’ perceptions and intentions. The effect of information provision on actual behavior will be limited in the presence of structural constraints. A structural intervention can help to overcome such barriers – we therefore expect it to have a stronger and direct effect on actual adoption, which is not mediated by changes in perceptions and intentions. Figure 1 Conceptual framework Note: Adapted from Meijer et al. (2014) and Steg & Vlek (2009) 10 1. 𝑎𝑖 in eq 2.3 and eq 2.4 is significant. There is a linear relationship between 𝑋 and 𝑀𝑖 2. 𝑐 in eq 2.1 is significant. There is a linear relationship between 𝑋 and 𝑌 3. 𝑏𝑖 in eq 2.2 is significant. 𝑀𝑖 helps to predict the outcome variable 𝑌 4. Finally, 𝑐′ in eq 2.2 is significantly smaller in size compared to 𝑐 in eq 2.1. We can then conclude that if 𝑎 or 𝑏 are not significant, there is no mediation, and assume that the variance of 𝑌 is attributable to the direct effect of 𝑋. If all four conditions hold, we conclude that there is full or partial mediation. This means that, the variance of 𝑌 attributable to 𝑋 is explained partly by an indirect effect mediated by 𝑀𝑖 . If 𝑐′ is no longer significant, we assume that all the effect runs through 𝑀𝑖 . 𝑀𝑖 has only a partial effect when 𝑐′ is smaller than 𝑐, but still significant (Lacobucci, 2008). Figure 3 shows the mediation analysis explored in this article. We model the direct effect and three mediating effects for each treatment. These are obtained as follows: the causal effect of T1 on adoption can be mediated through perceptions (𝑎𝑇11𝑏1), mediated by intention (𝑎𝑇12𝑏2), and mediated through perceptions and intention (𝑎𝑇11*𝑑𝑝𝑖*𝑏2). The sum of these mediating effects gives the total indirect effect of T1 on actual adoption. The direct effect of T1, without the mediators, on actual adoption is observed in 𝑐′𝑇1. The sum of the direct and indirect effects equals the total effect 𝑐𝑇1. Similarly, we obtain the mediating effects for T2. The causal effect of T2 on actual adoption can be mediated by perceptions (𝑎𝑇21𝑏1), by intention (𝑎𝑇22𝑏2), and by perceptions and intentions together (𝑎𝑇21*𝑑𝑝𝑖*𝑏2). The direct effect of T2 on actual adoption is provided by 𝑐′𝑇2. Figure 3 Schematic representation of mediation analysis Note: Adapted from Hayes (2018). 4.3 Measurement of key outcome variables We measure three outcomes in this study. First, we are interested in farmers’ perceptions of the provision of ecosystem functions by trees in oil palm. The measure was designed according to similar studies on tree planting (Meijer et al., 2015) including 17 items assessed 11 on a Likert scale. Using exploratory factor analysis, we constructed a total score to reflect perceptions. Second, the intention to plant was elicited by the subjective belief that the farmer will plant trees in his or her oil palm plantation. To obtain this information, farmers were asked to assess the probability that they will decide to plant, using elicitation methods recommended by Delavande et al. (2011). Third, actual adoption was measured as selfreported tree planting in oil palm plantations. Chandon et al. (2005) suggest capturing changes in perceptions and intentions before observing actual adoption. Therefore, we elicited perceptions and intentions in the follow-up survey in February 2016, shortly after the intervention was completed 5 . Data on actual adoption stems from the endline survey, which was implemented in October 2016. 4.3.1 Perceptions of the provision of ecosystem functions by trees in oil palm The scale that we use to assess farmers’ perceptions captures regulation, habitat, information, and provisioning functions (Groot et al. 2002) (see Table 6 in the Appendix). Regulating functions include those that maintain and regulate ecosystems through bio-geochemical and biospheric processes. Habitat functions provide refuge and reproduction of wild plants and animals allowing succession of biological and genetic diversity. Provisioning functions provide ecosystem goods for human consumption (e.g. food and raw materials). Information functions support cultural services such as spiritual enrichment, reflection, recreation and aesthetic experience. Items are measured on a 5-point-Likert scale, where 5 represents strongly agree. Looking at the mean values, we see that farmers perceive that trees in oil palm provide regulation and provisioning functions. Similarly, on average farmers agree that trees in oil palm provide habitat for bird and insect diversity, while farmers are indifferent to the role that trees have in terms of aesthetic services in oil palm. We use exploratory factor analysis to summarize all the statements in one latent factor. Since the statements were measured on a 5-point Likert scale, we use the Polychoric correlation. Due to the large number of statements we define the loading with a Varimax rotation and retain factors with an Eigen value greater than one (Yong and Pearce, 2013). Six statements did not load significantly on the factor. Internal validity was checked with help of the Kaiser- Meyer-Olkin indicator, which measures sampling adequacy (KMO=0.85), and Cronbach’s alpha (α=0.8205). Values above 0.7 are acceptable for both indicators. Generally, perceptions of ecosystem functions provided by native trees in oil palm are positive (see total factor score in Table 6 in the Appendix). Comparing the distribution across treatment groups, we can observe that on the average farmers assigned to the treatments have more positive perceptions than those in the control group. 6 4.3.2 Intention to plant trees The intention to plant trees is elicited by subjective expectations. A subjective expectation is the belief of a person regarding the probability that an event will occur in the future (Manski, 5 Perceptions and intentions can be subject to social desirability that could lead to over/under reporting (Clayton 2012; Gifford & Nilsson 2014). It has been found that social desirability is only weakly correlated with environmental attitudes and not related to pro-environmental behavior (Milfont 2009). To minimize potential bias in our results, we carefully phrased and tested the scales prior to data collection. In addition, we explained to the respondents the importance of their honest answer, as it was done in other studies (Meijer et al. 2015). 6 Mean differences for each statement between control and treatment groups can be observed in Table 6 in the Appendix. 12 1990). Conventionally, subjective probabilities have been assessed on Likert scales, with open-ended or binary questions. However, these approaches generate less information than assessing probabilities, which can be done with the help of visual aids, such as beans (Delavande et al. 2011; Manski 1990). We gave farmers 20 beans to illustrate their subjective expectations. At first, we explained to them that the amount of beans they choose represents the likelihood that a future event will happen. Training questions were used to ensure that farmers understood the concept of probabilities (Delavande, Giné and McKenzie, 2011). Then, we asked farmers to assess how likely they consider it to be that in the next 12 months they will plant native trees within their oil palm plantation. 7 The number of beans chosen by the farmer was multiplied by 0.05 to obtain probabilities. Figure 4 shows the distribution of farmers’ subjective expectations that they will plant trees in their oil palm plantations for the full sample and for the different groups. We observe that, on the average, farmers assigned to the treatments stated a higher intention to plant than farmers assigned to the control group. Note: Distribution of the subjective belief that farmer will plant trees in oil palm. Below each quadrant is the corresponding mean and standard deviation in parenthesis. Sample size: 670 7 The question asked to farmers was: “How likely do you think it is that in the next 12 months you will plant native trees within your oil palm plantation?” Figure 4 Intention to plant trees in oil palm 13 5 Econometric results 5.1.1 Effects of policy interventions Table 4 reports the intent-to-treat estimates. We observe that assignment to the information campaign only (T1) on average increases the perception factor by 0.34 points, and assignment to the information campaign plus seedling provision (T2) by 0.27 points in comparison to the control group. Although the difference between T1 and T2 is statistically significant, it is small in absolute terms. Intent-to-treat estimates further reveal that farmers’ subjective probability that they will plant trees is 20 percentage points higher in both, T1 and T2, compared to the control group. Our findings are in line with earlier, non-experimental, studies emphasizing that informational interventions succeed in increasing the knowledge of an individual and creating awareness about a specific topic (Zelenski, Dopko and Capaldi, 2015; De Martino et al., 2016). Column 3 of Table 4 shows marginal effects of the interventions on actual tree planting adoption. We observe that farmers assigned to the information campaign only (T1) are on average 7 percentage points more likely to plant trees in their oil palm plantations in comparison to the control group. Farmers assigned to the information campaign plus seedling provision (T2) are 42 percentage points more likely to plant trees in their oil palm plantations compared to the control group. While both interventions have a significant and positive effect, t-test results show that the effect of T2 is significantly larger than that of T1, suggesting that the structural intervention is crucial to induce behavioral change more widely. Table 4 Intent-to-treat effects (1) (2) (3) Perceptions Intention to plant Actual tree planting T1 0.34*** (0.028) 0.20*** (0.046) 0.07*** (0.024) T2 0.27*** (0.028) 0.20*** (0.040) 0.42*** (0.030) Control variables1 Y Y Y P-values of t-test for T1=T2 0.003 0.932 0.000 Observations 670 670 670 R2 0.362 0.117 Pseudo R2 0.236 Note: Each column is a separated weighted regression. Columns 1 and 2 show the estimated coefficient of an OLS regression. Column 3 shows marginal effects from a logit regression. Standard errors are cluster-corrected at village level, shown in parenthesis. Results for the full regressions are provided in Table 7 in the Appendix). 1Control variables include household characteristics and stratification variables. 14 5.2 Are the treatment effects mediated by changes in perceptions and intentions? As postulated in the conceptual framework, we hypothesize that the effect of the information campaign on actual adoption is mediated through changes in perceptions and intentions. Accordingly, the information campaign, which was delivered in both treatments, will have positive effects on farmers’ perceptions and intentions, which in turn will increase the likelihood of actual adoption among the treated farmers. Results of the mediation analysis are depicted in Figure 5. 8 The causal effect of the assignment to the information campaign only (T1) is mediated through perceptions and intentions: The indirect effect of T1 on adoption, which runs through perceptions and intentions, is 0.045 and statistically significant at the one percent level. Once controlling for the mediated effect, the direct effect of T1 on adoption turns insignificant (𝑐′𝑇2 = 0.135). This indicates that the effect of the information campaign on adoption is fully explained by increases in perceptions and intentions. We further find that the causal effect of the information campaign plus seedling provision (T2) is also mediated by perceptions and intentions: The indirect effect of T2 on adoption, mediated by perceptions and intentions, is 0.038 and statistically significant. However, even when controlling for the mediating effect of perceptions and intentions, the direct effect of T2 on actual adoption is still positive and significant. Thus, in the combined intervention (information campaign plus seedling provision) a significant portion of the causal effect on adoption remains unexplained. Beyond the pathway of changing mindsets, there seem to be other mechanisms through which T2 leverages adoption, most likely related to the free and easy access to seedlings, which may facilitate adoption and experimentation even among farmers who are not fully convinced by the information campaign. 8 Full model results are shown in Table 8 in the Appendix Figure 5 Results of the mediation analysis 15 Note: N=670. Estimates for each path are given in standardized form. Full estimations are provided in Table 8 in the Appendix. We control for additional baseline covariates as a robustness check. The regression coefficients are next to their respective path. Effects of T1 are shown in the upper part of the diagram, while effects of T2 are shown in the lower part. The model was performed with a Maximum Likelihood Robust (MLR) estimator in Mplus. Since the mediators are continuous variables those regressions can be interpreted as in an OLS regression. The regression coefficient for actual behavior can be interpreted as in a Logit regression. Model fit: Log-likelihood user model (H0): - 320.477, Akaike (AIC): 688.955, Bayesian (BIC): 797.130, Sample-size adjusted Bayesian: 720.928 6 Conclusion Rapid loss of biodiversity caused by oil palm expansion is likely to continue in Indonesia. This trend urges for policies that influence behavioral change towards the adoption of more biodiversity-friendly oil palm management. This article evaluates the effects of two policy instruments on attitudes, intentions and tree planting behavior among smallholder oil palm farmers in Sumatra, Indonesia. Using a randomized controlled trial, we test the effects of an information campaign only and an information campaign combined with seedling provision on attitudes , intentions, and adoption of native tree planting in smallholder oil palm plantations. Both interventions have positive and significant effects on attitudes towards tree planting and on the intention to plant trees in oil palm. Furthermore, both interventions have a positive and significant effect on actual tree planting, increasing the probability of adoption by 7 and 42 percentage points respectively. This result suggests that despite profitability considerations that favor tropical cash crops, non-financial instruments and mechanisms can be effective in steering behavior towards more biodiversity-friendly land use choices. It is encouraging that we do find these positive effects even though our intervention was relatively short and low cost (one video-screening per village, distribution of a manual, provision of a small number of seedlings) and despite the fact that farmers were initially quite skeptical about biodiversity enrichment in their oil palm plantations. What are the underlying mechanisms then through which farmers are motivated to change their behavior and plant trees in oil palm? Using a path model, we tested to what extent the effects of our interventions can be explained by changes in farmers’ mindsets, measured here in terms of their perceptions and intentions. Results show that the effect of the information campaign (T1) on actual tree planting is fully mediated by changes in mindsets. Assignment to T1 has a significantly positive effect on perceptions and intentions, which then translates into actual adoption. In contrast, changes in mindsets are only partial mediators for the combined intervention (T2). There is a significant portion of T2’s effect on actual tree planting that cannot be explained by the observed changes in perceptions and intentions triggered by the information campaign. Providing free and easy access to seed material may thus have motivated even some of the farmers that were not fully convinced of the advantages of tree planting to experiment with a few trees on their oil palm plots. We thus conclude that an information campaign is important to change mindsets, but if the goal is to spread the technology widely to induce experimentation, facilitating easy access to critical inputs may be key to motivate adoption among large numbers of potential users. 16 References Acharya, A., Blackwell, M. and Sen, M. (2016) ‘Explaining Causal Findings Without Bias : Detecting and Assessing’, American Political Science Review, 110(3), pp. 512–529. doi: 10.1017/S0003055416000216. Ajzen, I. (1991) ‘The Theory of Planned Behavior’, Organizational Behavior and Human Decision Processes, 50(2), pp. 179–211. doi: 10.1016/0749-5978(91)90020-T. Azhar, B. et al. (2015) ‘Promoting landscape heterogeneity to improve the biodiversity benefits of certified palm oil production: Evidence from Peninsular Malaysia’, Global Ecology and Conservation, 3, pp. 553–561. doi: 10.1016/j.gecco.2015.02.009. Bamberg, S. (2003) ‘How does environmental concern influence specific environmentally related behaviors? A new answer to an old question’, Journal of Environmental Psychology, 23(1), pp. 21–32. doi: 10.1016/S0272-4944(02)00078-6. Baron, R. M. and Kenny, D. a. (1986) ‘The Moderator-Mediator Variable Distinction in Social Psychological Research: Conceptual, Strategic, and Statistical Considerations’, Journal of Personality and Social Psychology, 51(6), pp. 1173–1182. doi: 10.1037/0022- 3514.51.6.1173. Bloom, H. S. (2006) ‘The core analytics of randomized experiments for social research’, MDRC Working Papers on Research Methodology, (August), pp. 1–41. doi: 10.4135/9781446212165. de Brauw, A. et al. (2015) ‘Biofortification, Crop Adoption, and Health Information: Impact Pathways in Mozambique and Uganda. IFPRI HarvestPlus Working Paper N. 21’, (November). Available at: www.ifpri.org/cdmref/p15738coll2/id/129795/filename/130006.pdf. Burgess, R. et al. (2012) ‘The Political Economy of Deforestation in the Tropics’, The Quarterly Journal of Economics, 127(4), pp. 1707–1754. doi: 10.1093/qje/qjs034.Advance. Butler, R. A., Koh, L. P. and Ghazoul, J. (2009) ‘REDD in the red: palm oil could undermine carbon payment schemes’, Conservation Letters, 2(2), pp. 67–73. doi: 10.1111/j.1755- 263X.2009.00047.x. Campos, F. et al. (2017) ‘Teaching personal initiative beats traditional training in boosting small business in West Africa’, Science, 357(6357), pp. 1287–1290. doi: 10.1126/science.aan5329. Carter, C. et al. (2007) ‘Palm oil markets and future supply’, European Journal of Lipid Science and Technology, 109(4), pp. 307–314. doi: 10.1002/ejlt.200600256. Carter, M. R.; Laajaj, R.; Yang, D. (2013): The Impact of Voucher Coupons on the Uptake of Fertilizer and Improved Seeds. Evidence from a Randomized Trial in Mozambique. In American Journal of Agricultural Economics 95 (5), pp. 1345–1351. Chandon, P., Morwitz, V. G. and Reinartz, W. J. (2005) ‘Do Intentions Really Predict Behavior? Self-Generated Validity Effects in Survey Research’, Journal of Marketing, 69(April), pp. 1–14. doi: 10.1509/jmkg.69.2.1.60755. 17 Clough, Y. et al. (2016) ‘Land-use choices follow profitability at the expense of ecological functions in Indonesian smallholder landscapes’, Nature Communications, 7. doi: 10.1038/ncomms13137. Cole, R. J. ., Holl, K. . D. . and Zahawi, R. . (2010) ‘Seed rain under tree islands planted to restore degraded lands in a tropical agricultural landscape.’, Ecological Society of America, 20(5), pp. 1255–69. Available at: https://doi.org/10.1890/09-0714.1%0A. Coordinating Ministry of Economic Affairs (2011) Masterplan: Acceleration and Expansion of Indonesia Economic Development 2011–2025. Jakarta. Available at: https://www.dezshira.com/library/treaties/master-plan-acceleration-and-expansion-of- indonesia-economic-development-2011-2025-2764.html (Accessed: 15 May 2017). Corley, R. H. V. . and Tinker, P. B. (2016) The Oil Palm. Fifth Edit. Chichester, England: Wiley Blackwell. Delavande, A., Giné, X. and Mckenzie, D. (2010) Eliciting probabilistic expectations with visual aids in developing countries: How sensitive are answers to variations in elicitation design?. The World Ban. doi: 10.1002/jae.1233. Delavande, A., Giné, X. and McKenzie, D. (2011) ‘Measuring subjective expectations in developing countries: A critical review and new evidence’, Journal of Development Economics. Elsevier B.V., 94(2), pp. 151–163. doi: 10.1016/j.jdeveco.2010.01.008. Drescher, J. et al. (2016) ‘Ecological and socio-economic functions across tropical land use systems after rainforest conversion’, Philosophical transactions of the Royal Society of London. Series B, Biological sciences, 231(20150275), pp. 1–7. doi: 10.1098/rstb.2015.0275. Duflo, E., Glennerster, R. and Kremer, M. (2008) ‘Using Randomization in Development Economics: A Toolkit’, Handbook of Development Economics, 53, p. 68. doi: 10.1016/S1573-4471(07)04061-2. Duflo, E.; Kremer, M.; Robinson, J. (2011): Nudging Farmers to Use Fertilizer. Theory and Experimental Evidence from Kenya. In American Economic Review 101 (6), pp. 2350–2390. Edwards, D. P. . et al. (2010) ‘Wildlife-friendly oil palm plantations fail to protect biodiversity effectively’, Conservation Letters, 3(4), pp. 236–242. doi: 10.1111/j.1755- 263X.2010.00107.x. Euler, M. et al. (2015) ‘Oil palm adoption, household welfare and nutrition among smallholder farmers in Indonesia’, World Development. Elsevier Ltd. doi: 2197-6244. Euler, M. et al. (2016) ‘Oil Palm Expansion among Smallholder Farmers in Sumatra, Indonesia’, Journal of Agricultural Economics, 67(3), pp. 658–676. doi: 10.1111/1477- 9552.12163. FAOSTAT (2018) Statistics division. Food and Agricultural Organization, Rome. Available at: http://faostat.fao.org (Accessed: 12 January 2018). Fitzgerald, J., Gottschalk, P. and Moffitt, R. (1997) ‘An analysis of sample attrition in panel data: The Michigan panel study of income dynamics’. Available at: www.nber.org/papers/t0220.pdf. 18 Fitzherbert, E. B. . et al. (2008) ‘How will oil palm expansion affect biodiversity?’, Trends in Ecology and Evolution, pp. 538–545. doi: 10.1016/j.tree.2008.06.012. Foley, J. A. et al. (2005) ‘Global consequences of land use’, Science, 309(5734), pp. 570– 574. doi: 10.1126/science.1111772. Gatto, M. et al. (2017) ‘Oil Palm Boom, Contract Farming, and Rural Economic Development: Village-Level Evidence from Indonesia’, World Development. Elsevier Ltd, 95, pp. 127–140. doi: 10.1016/j.worlddev.2017.02.013. Gérard, A. et al. (2017) ‘Oil palm yields in diversified plantations: initial results from a biodiversity enrichment experiment in Sumatra, Indonesia’, Agriculture, Ecosystems and Environment, 240, pp. 253–260. doi: 10.1016/j.agee.2017.02.026. Hansson, H., Ferguson, R. and Olofsson, C. (2012) ‘Psychological Constructs Underlying Farmers’ Decisions to Diversify or Specialise their Businesses - An Application of Theory of Planned Behaviour’, Journal of Agricultural Economics, 63(2), pp. 465–482. doi: 10.1111/j.1477-9552.2012.00344.x. Hayes, A. F. (2018) Introduction to Mediation, Moderation, and Conditional Process Analysis. A regression-Based Approach. 2nd Ed. New York: Guilford Publications. Imai, K. et al. (2016) ‘Unpacking the Black Box of Causality : Learning about Causal Mechanisms from Experimental and Observational Studies’, The American Political Science Review, 105(4), pp. 765–789. Available at: http://www.jstor.org/stable/23275352. Imai, K., Keele, L. and Yamamoto, T. (2010) ‘Identification, Inference and Sensitivity Analysis for Causal Mediation Effects’, Statistical Science, 25(1), pp. 51–71. doi: 10.1214/10-STS321. Jack, B. K. et al. (2013) ‘Private Information and the Allocation of Land Use Subsidies in Malawi’, American Economic Journal: Applied Economics, 5(3), pp. 113–135. doi: 10.1257/app.5.3.113. Klasen, S. et al. (2016) ‘Economic and ecological trade-offs of agricultural specialization at different spatial scales’, Ecological Economics. The Authors, 122, pp. 111–120. doi: 10.1016/j.ecolecon.2016.01.001. Klöckner, C. A. (2013) ‘A comprehensive model of the psychology of environmental behaviour-A meta-analysis’, Global Environmental Change, 23(5), pp. 1028–1038. doi: 10.1016/j.gloenvcha.2013.05.014. Koh, L. P. and Wilcove, D. S. (2008) ‘Is oil palm agriculture really destroying tropical biodiversity?’, Conservation Letters, 1(2), pp. 60–64. doi: 10.1111/j.1755- 263X.2008.00011.x. Lacobucci, D. (2008) Mediation Analysis: Quantitative Applications in the Social Sciences. 156th edn. California: SAGE. Leimona, B., Joshi, J. and van Noordwijk, M. (2009) ‘Can rewards for environmental services benefit the poor ? Lessons from Asia’, International Journal of the Commons, 3(1), pp. 82– 107. Available at: http://doi.org/10.18352/ijc.121. 19 MacKinnon, D., Cheong, J. and Pirlott, A. G. (2012) An Introduction to Statistical Mediation Analysis, Handbook of Research Methods in Psychology. Manski, C. F. (1990) ‘The Use Of Intentions Data To Predict Behaviour : A Best- Case Analysis’, Journal of the American Statistical Association, 85(412), pp. 934–940. doi: 10.1093/restud/rdu044. De Martino, S. et al. (2016) Do They Do As They Say? Stated versus revealed preferences and take up in an incentives for conservation prgram, Environmental and Natural Resources. Available at: https://openknowledge.worldbank.org/handle/10986/25177. Meijer, S. S. . et al. (2014) ‘The role of knowledge, attitudes and perceptions in the uptake of agricultural and agroforestry innovations among smallholder farmers in sub-Saharan Africa’, International Journal of Agricultural Sustainability. Taylor & Francis, 13(1), pp. 40–54. doi: 10.1080/14735903.2014.912493. Meijer, S. S. . et al. (2015) ‘Tree planting by smallholder farmers in Malawi: Using the theory of planned behaviour to examine the relationship between attitudes and behaviour’, Journal of Environmental Psychology. Elsevier Ltd, 43, pp. 1–12. doi: 10.1016/j.jenvp.2015.05.008. Morgan, K. L. and Rubin, D. B. (2012) ‘Rerandomization to improve covariate balance in experiments’, The Annals of Statistics, 40(2), pp. 1263–1282. doi: 10.1214/12-AOS1008. Ndayambaje, J. D. ., Heijman, W. J. M. and Mohren, G. M. J. (2012) ‘Household Determinants of Tree Planting on Farms in Rural Rwanda’, Small-scale Forestry, 11(4), pp. 477–508. doi: 10.1007/s11842-012-9196-0. Pamuk, H., Bulte, E. and Adekunle, A. A. (2014) ‘Do decentralized innovation systems promote agricultural technology adoption ? Experimental evidence from Africa’, Food Policy. Elsevier Ltd, 44, pp. 227–236. doi: 10.1016/j.foodpol.2013.09.015. PROSEA (2016) Plant resources of South-East Asia. Available at: http://www.proseanet.org (Accessed: 1 September 2016). Ray, B. and Bhattacharya, R. (2004) Stakeholder Attitudes and Conservation of Natural Resources: Exploring Alternative Approaches, Development and Sustainability. doi: 10.1007/978-81-322-1124-2. Rogers, E. M. (1983) Diffusion of Innovations. doi: 82-70998. Sood, K. K. et al. (2004) ‘Do Socio-psychological Factors Matter in Agroforestry Planning? Lessons from Smallholder Traditional Agroforestry Systems’, Small-scale Forest Economics, Management and Policy, 3(2), pp. 239–255. doi: 10.1007/s11842-004-0017-y. Steg, L. et al. (2014) ‘An Integrated Framework for Encouraging Pro-environmental Behaviour: The role of values, situational factors and goals’, Journal of Environmental Psychology, 38, pp. 104–115. doi: 10.1016/j.jenvp.2014.01.002. Steg, L. and Vlek, C. (2009) ‘Encouraging pro-environmental behaviour: An integrative review and research agenda’, Journal of Environmental Psychology. Elsevier Ltd, 29(3), pp. 309–317. doi: 10.1016/j.jenvp.2008.10.004. 26 Table 8 Path analysis to test perceptions and intention as mediators Unstandardized Standardized (1) (2) (3) (4) (5) Perceptions pvalues pvalues Odds ratio T1 0.341 (0.020) 0.000 0.643 (0.037) 0.000 T2 0.276 (0.021) 0.000 0.534 (0.041) 0.000 Access to seedlings 0.050 (0.019) 0.008 0.087 (0.033) 0.008 Autochtonous 0.027(0.023) 0.239 0.047 (0.040) 0.239 Share of oil palm 0.040 (0.019) 0.040 0.080 (0.039) 0.040 Age 0.001 (0.001) 0.410 0.029 (0.035) 0.410 Years of education 0.003 (0.002) 0.251 0.040 (0.035) 0.251 Household members 0.005 (0.006) 0.351 0.031 (0.034) 0.351 Homegarden -0.010 (0.028) 0.728 - 0.011(0.031) 0.728 If trees in oil palm 0.042 (0.018) 0.021 0.076 (0.033) 0.021 Perceptions-baseline 0.018 (0.008) 0.025 0.076 (0.034) 0.025 =1 if cut trees last 12 m -0.012 (0.042) 0.776 -0.009 0.032) 0.776 R-square 0.364 0.364 Intention Perceptions 0.499 (0.056) 0.000 0.349 (0.040) 0.000 T1 0.033 (0.039) 0.406 0.044 (0.053) 0.406 T2 0.057 (0.034) 0.092 0.077 (0.046) 0.092 27 R-square 0.147 0.147 Actual adoption Intention 1.307 (0.335) 0.000 0.202 (0.049) 0.000 3.694 Perceptions 0.548 (0.603) 0.363 0.059 (0.065) 0.363 1.730 T1 0.651 (0.471) 0.167 0.135 (0.095) 0.154 1.918 T2 2.761 (0.431) 0.000 0.577 (0.073) 0.000 15.815 R-square 0.360 0.372 Intercepts Intention -0.036 (0.029) 0.221 -0.101 (0.083) 0.221 Perceptions 0.280 (0.067) 0.000 1.133 (0.278) 0.000 Thresholds OP$1 3.952 (0.507) 0.000 1.726 (0.170) 0.000 Residual Variance Intention 0.107 (0.005) 0.000 0.853 (0.023) 0.000 Perceptions 0.039 (0.002) 0.000 0.636 (0.035) 0.000 Direct effect T1 0.651 (0.471) 0.167 0.135 (0.095) 0.154 Indirect effect T1 (through only perceptions) 0.183 (0.202) 0.364 0.038 (0.042) 0.362 Indirect effect T1 (through only intention) 0.043 (0.053) 0.419 0.009 (0.011) 0.418 Indirect effect T1 (through perceptions and intention) 0.218 (0.063) 0.001 0.045 (0.013) 0.000 Total indirect effect T1 0.444 (0.204) 0.030 0.092 (0.011 0.027 Total effect T1 1.095 (0.450) 0.015 0.092 (0.042) 0.027 28 Direct effect T2 2.761 (0.431) 0.000 0.577 (0.073) 0.000 Indirect effect T2 (through only perceptions) 0.151 (0.167) 0.366 0.032 (0.035) 0.364 Indirect effect T2 (through only intention) 0.074 (0.048) 0.124 0.015 (0.010) 0.120 Indirect effect T2 (through perceptions and intention) 0.180 (0.052) 0.000 0.038 (0.010) 0.000 Total indirect effect T2 0.405 (0.171) 0.018 0.085 (0.035) 0.015 Total effect T2 3.166 (0.419) 0.000 0.661 (0.063) 0.000 Direct effect perceptions 0.548 (0.603) 0.363 0.059 (0.065) 0.912 Indirect effect perceptions (through intention) 0.652 (0.183) 0.000 0.070 (0.019) 0.000 Total indirect effect perceptions 0.652 (0.183) 0.000 0.070 (0.019) 0.000 Total effect perceptions 2.761 (0.431) 0.041 0.129 (0.063) 0.039 N 670 Note: Results of a weighted structural equation model. Estimator: Maximum Likelihood Robust (ML), the model was performed in Mplus Loglikelihood user model (H0): -338.722 /Akaike (AIC): 701.445 / Bayesian (BIC): 755.621 / Sample-size adjusted Bayesian (BIC): 715.220 Marginal effects were estimated with the unstandardized estimates from Column 1. We follow the formula: (𝑢 = 1|𝑥 = 1)= 1 1+𝑒−𝐿 ; where 𝐿 = 𝑇ℎ𝑟𝑒𝑠ℎ𝑜𝑙𝑑𝑠 𝑜𝑝 + 𝛽𝑑𝑖𝑟𝑒𝑐𝑡 𝑇1→𝑎𝑐𝑡𝑢𝑎𝑙 𝑎𝑑𝑜𝑝𝑡𝑖𝑜𝑛 for T1 and 𝑇ℎ𝑟𝑒𝑠ℎ𝑜𝑙𝑑𝑠 𝑜𝑝 + 𝛽𝑑𝑖𝑟𝑒𝑐𝑡 𝑇2→𝑎𝑐𝑡𝑢𝑎𝑙 𝑑𝑜𝑝𝑡𝑖𝑜𝑛 for T2.