1. Introduction
A common argument for participation of local stakeholders within the epistemic development and nature conservation communities is that participation directly improves institutions' ability to deliver effective, efficient or equitable outcomes.Footnote 1 Dubbed ‘sense of ownership,’ the argument hypothesizes that local stakeholders feel a sense ownership of institutions they are involved in initiating, designing and/or implementing. This subsequently increases stakeholder support and thus institutional performance. The argument is found both among project proponentsFootnote 2 and in the academic literature (e.g., Lachapelle, Reference Lachapelle2008; Reed, Reference Reed2008; Marks and Davis, Reference Marks and Davis2012). Schultz et al. (Reference Schultz, Duit and Folke2011: 662) summarize the hypothesis as ‘people are more likely to support and implement decisions they have participated in making’, and Buchy and Hoverman (Reference Buchy and Hoverman2000: 19) argue that ‘people, feeling a sense of ownership, will be more compliant to bear the costs.’ A Google Scholar search reveals 162 articles in ten top development journals mentioning ‘sense of ownership’ (1990–2016).Footnote 3
Despite its popularity, the hypothesis lacks a precise definition. Details remain fuzzy on how participation induces a sense of ownership (SoO) and how SoO in turn improves institutional performance, controlling for other positive aspects of participation, such as access to local knowledge. One reason for this is inadequate theoretical tools. ‘Sense’ implies a psychological mechanism, which is rarely explored in the relevant literature. Another reason is challenges in empirically identifying SoO effects. Participation and SoO as elusive concepts make comparative studies difficult, and persistent biases in observations make more quantitative investigations difficult. Several scholars note that research assessing the effects of stakeholder participation on institutional performance is scarce (Agrawal, Reference Agrawal2007; Reed, Reference Reed2008; Schultz et al., Reference Schultz, Duit and Folke2011; Mansuri and Rao, Reference Mansuri and Rao2013b).
Viewing participation as a means, understanding SoO can help initiate, design and implement more efficient, effective and equitable institutions. Viewing participation as an end, finding empirical support for SoO can motivate more and stronger participatory approaches among policy makers.
This paper applies psychology and behavioral economics for a theoretical understanding of SoO, and proposes three possible drivers: biased information processing, associative self-anchoring and cognitive dissonance (section 3). Economic experiments can help improve our empirical understanding by testing drivers and controlling for possible confounding factors. Since SoO may be context specific, a specific form of participation is explored through framed field experiments in the relevant context of forest conservation in Tanzania (section 4). The results (section 5) are discussed and related to development and nature conservation efforts (sections 6 and 7). To explore the participation–SoO link, the following section presents a typology of participation.
2. Participation
2.1 Definitions and classifications
Over the last 40 years, participation has become a core element in the implementation of development and nature conservation institutions in developing countries (Cohen and Uphoff, Reference Cohen and Uphoff1980; Cornwall, Reference Cornwall2008; Lie, Reference Lie2015). World Bank, for instance, spent US$ 85 billion on participatory projects during the last decade alone (Mansuri and Rao, Reference Mansuri and Rao2013b).
Despite the long focus on, and the positive aspects of, participatory approaches, empirical investigations often find questionable implementations. In some cases, ‘participation’ is merely a buzzword in project documents with few real-world implications (Baland and Platteau, Reference Baland and Platteau1996; Cornwall and Brock, Reference Cornwall and Brock2005; Leal, Reference Leal2007; Reed, Reference Reed2008; Ribot et al., Reference Ribot, Lund and Treue2010; Lie, Reference Lie2015). Another reason for varying practices is varying definitions of participation, depending on context and objective (Lawrence, Reference Lawrence2006; Cornwall, Reference Cornwall2008).
This paper views participation as the process where stakeholders make choices that determine (or co-determine) new institutions. The stakeholders of interest are local individuals who are affected by the institution and the choices (Freeman, Reference Freeman1984; Reed, Reference Reed2008). We focus on localized and not societal-wide participation (cf. Agrawal and Ostrom, Reference Agrawal and Ostrom2001).
There are numerous classifications of participation. We focus on two dimensions: degree of involvement and timing. The ladder of Arnstein (Reference Arnstein1969), and later extensions by others (Reed, Reference Reed2008), ranks participation by degree of stakeholder involvement and power, where the top rung is majority or full control of decisions. Figure 1, presented in section 2.3, outlines a simplified ranking from weak to strong participation. Weak participation is consultation, where stakeholders inform decision makers but have no direct influence over decisions. A medium level of participation is consent, where stakeholders may veto decisions and thus make constrained choices, with impact. Lastly, strong participation is control, where stakeholders are in the driving seat and make choices under few, if any, constraints, as in Arnstein's top rung.
Participation as a means.
Stakeholders may be involved at different time stages. First, the initiative can arise from the stakeholders or externally. Mansuri and Rao (Reference Mansuri and Rao2013a) distinguish between organic participation, where social movements from below initiate the process, and induced participation, where the initiative is external (but stakeholders are later involved within the set agenda). Second, stakeholders may participate in designing institutions (i.e., setting the rules). Third, stakeholders may participate in implementing initiatives from stage one according to the rules set in stage two. The involvement and time dimensions with three stages/degrees each are depicted in the two leftmost columns of figure 1.
While these categories are not exhaustive and demarcations blur, such a typology can be useful to understand participation as a means. Particularly, it allows for applying theories from behavioral economics to explain possible positive effects of participation.
2.2 Why participation may matter
From a project proponent's perspective, participation of local stakeholders can be a means to improve an institution's effectiveness (achievement of the set objectives), efficiency (costs of achieving the objectives) and/or equity (fair distribution of costs and benefits). Institutional performance refers to the ability to deliver these three Es according to given objectives. This instrumental significance of participation is advocated in many settings, e.g., in organizations (Pierce et al., Reference Pierce, Kostova and Dirks2001), in medicine (Hickey and Kipping, Reference Hickey and Kipping1998) and in marketing (Krugman, Reference Krugman1966).
Similar arguments have also been made related to development and nature conservation institutions (Bardhan, Reference Bardhan2000; Chatterley et al., Reference Chatterley, Linden and Javernick-Will2013). Buchy and Hoverman (Reference Buchy and Hoverman2000, 16) argue for ‘participation as a management tool’ in Australian forestry, and Tanzania's participatory forest management (PFM) has arguably been somewhat successful in reaching its goal of sustainable forest use (Brockington, Reference Brockington2007; Lund and Treue, Reference Lund and Treue2008; Treue et al., Reference Treue2014). Mansuri and Rao (Reference Mansuri and Rao2013b: 6) in their extensive review, conclude that ‘greater community involvement seems to modestly improve resource sustainability and infrastructure quality’, but that the benefits are unequally distributed.Footnote 4
Participation effects have also been explored in laboratory experiments. In public good experiments, participants have been shown to be more likely to follow norm behavior under a sanctioning mechanism when the mechanism is chosen through a referendum than when it is imposed externally (Tyran and Feld, Reference Tyran and Feld2006; Sutter et al., Reference Sutter, Haigner and Kocher2010; Markussen et al., Reference Markussen, Putterman and Tyran2013). Dal Bó et al. (Reference Dal Bó, Foster and Putterman2010) investigate the ‘endogeneity premium effect,’ by allowing participants to vote on their preferred treatment in prisoners' dilemma games, but discard the votes in some sessions and exogenously impose the treatment. By comparing those who voted for the policies in the sessions that chose the policy and in the sessions that had the policy imposed, they control for the selection effect. They find that choosing has a positive impact, and that the effect is both due to the selection effect and a mere choice effect. A potential issue with the design is the comparison between participants who vote and subsequently choose a policy and participants who vote, but whose vote is disregarded and the policy (they voted for) is implemented exogenously. In addition to having the policy imposed, participants in the latter sessions know that their vote is discarded. Merely having one's vote discarded could affect behavior, for instance by reducing the policy's credibility and thus the ability to coordinate behavior. These studies are also in abstract form and conducted in laboratories, ignoring possible context-relevant effects (Henrich et al., Reference Henrich2001; Harrison et al., Reference Harrison, List and Towe2007; Maddux et al., Reference Maddux2010).
Table 1 summarizes key arguments advanced for participation having positive effects on institutional performance in five mechanisms. First, by involving local stakeholders in project design, local knowledge that may be hidden to external ‘experts’ is accessed, thus creating locally-adapted institutions that rely on deeper knowledge and more updated information (both in setting objectives and creating measures). Second, the process of including local stakeholders is empowering, which helps to better sustain the institution (as stakeholders are better equipped with skills, power, knowledge, etc.). Third, as participation requires some degree of openness in the decision-making process, it improves transparency and holds the project proponents more accountable, which helps create more efficient institutions. Fourth, dialogues between stakeholders and project proponents, and among stakeholders, reduce the number and intensity of conflicts, and thus reduce the costs related to solving such conflicts. Fifth, local participation induces a sense of ownership among the stakeholders, as described in section 1.
Mechanisms through which participation may improve institutional performance
2.3 A typology
Summarizing sections 2.1 and 2.2, participation – at the initiative, design and/or implementation stage – can improve institutional effectiveness, efficiency and/or equity through one or more of the five listed mechanisms (figure 1). The effect of the mechanism(s) on institutional performance may vary by stage, level of participation and context.
Sense of ownership (SoO), although often unspecified, could also apply to any of the three stages and outcomes. Stakeholders may feel a sense ownership of the idea to be initiated, of the design process, or of the implementation. These may affect the effectiveness, efficiency and/or equity differently, and may also interact with the context and the other mechanisms. Section 4 aims to test the effect of participation (choice) in the design stage, inducing a SoO among local stakeholders, which improves institutional effectiveness.
3. Sense of ownership – a behavioral perspective
SoO implies a psychological effect, as participation in itself increases institutional performance. Theoretical support for SoO may thus be found in the related literature.
A useful distinction is between real and symbolic ownership (Etzioni, Reference Etzioni1991). While the former strictly applies to objects and often relates to legal property rights, the latter mostly refers to non-objects, which can be ‘owned’ by multiple parties simultaneously. Symbolic ownership is an attribute of our minds, while real ownership also exists outside our minds. Sense of ownership, as interpreted here, induces symbolic ownership (of a project, policy, etc.) to stakeholders.
Related literature within behavioral economics and psychology tends to focus on real ownership and differences in appreciations depending on whether a good is owned or not (Thaler, Reference Thaler1980; Kahneman et al., Reference Kahneman, Knetsch and Thaler1991). The tendency to value a good more if it is owned is referred to as the endowment effect. Similar discrepancies have also been identified for more abstract goods, such as clean air, local tree density, job health risk and work effort (Horowitz and Mcconnell, Reference Horowitz and Mcconnell2002; Norton et al., Reference Norton, Mochon and Ariely2012). Reasons (drivers) for the discrepancies are disputed, and Morewedge and Giblin (Reference Morewedge and Giblin2015) summarize several that are posited. Three of these may relate to SoO: (i) biased information processing, (ii) associative self-anchoring, and (iii) cognitive dissonance.
Biased information processing: owners are in a different framing from non-owners and thus weigh attributes of the good in question differently (Carmon and Ariely, Reference Carmon and Ariely2000; Johnson et al., Reference Johnson, Häubl and Keinan2007). This difference in weighting is biased towards considering the good more favorably when owned. Related to SoO, stakeholders who participate become owners and thus focus more on the positive aspects of the given institution than non-owners do, and this favorable evaluation spurs support for the given institution.
Associative self-anchoring: choosing or owning establishes an association, and as one in general evaluates oneself positively (or wants to do so) (Benabou and Tirole, Reference Benabou and Tirole2002), one values what one chooses more favorably (Gawronski et al., Reference Gawronski, Bodenhausen and Becker2007; Weiss and Johar, Reference Weiss and Johar2013). That we tend to implicitly self-evaluate ourselves positively implies that we unconsciously think of our self-concept and self-esteem in positive terms (Bosson et al., Reference Bosson, Swann and Pennebaker2000; Greenwald and Farnham, Reference Greenwald and Farnham2000; Gawronski et al., Reference Gawronski, Bodenhausen and Becker2007; Stieger et al., Reference Stieger, Voracek and Formann2012). To elicit this, the related literature suggests the Initial Preference Task, which asks individuals to rate letters on a scale of likability or attractiveness (Stieger et al., Reference Stieger, Voracek and Formann2012). The consistent finding that individuals rate letters in their names more favorably than non-name letters indicates positive implicit self-evaluation. SoO may thus occur through (positively self-evaluated) stakeholders' association between themselves and the institution.
Choosing implies choosing away something else, creating a potential cost of regret. To compensate for this, cognitive dissonance predicts that one increases the perceived value of the chosen alternative and/or decreases the perceived value of the unchosen alternative(s) (Festinger, Reference Festinger1957; Morewedge and Giblin, Reference Morewedge and Giblin2015). Stakeholders' participation forces more support for the institution because stakeholders mentally justify that they made the right choices. This effect is stronger the more equal the alternatives are valued, pre-choice.
The following presents an experimental study aiming to test the effect of choice on institutional performance in a specific context, and to test each of the three posited psychological mechanisms described above through framed field experiments.
4. Experimental study
4.1 The context: REDD+ and Tanzania
REDD+ (Reducing Emissions from Deforestation and forest Degradation, and enhancement of forest carbon stocks in developing countries) is an umbrella term for actions aimed at reducing carbon emissions from forest use in developing countries (Angelsen, Reference Angelsen2009). It is seen as a relatively quick and cheap approach to mitigate climate change (Stern, Reference Stern2006), with increasing attention and funds being received at the global level since its launch in 2007. More than US$ 9.8 billion was pledged in aggregate between 2006 and 2014, with Norway as one of the most prominent donors (Angelsen, Reference Angelsen2016).
The core idea of REDD+ is payment for environmental services (PES), i.e., forest users (countries, communities, individuals) are compensated for reduced emissions from estimated baselines (Angelsen, Reference Angelsen2009). In implementing the forest conservation initiatives, however, a wide range of policies are likely to be applied, such as participatory forest management (PFM), where local communities are (to varying extents) responsible for the management (Angelsen, Reference Angelsen2009; Green and Lund, Reference Green and Lund2015). In fact, the UN-REDD Programme (2011, 5) emphasizes that ‘stakeholder participation and engagement is critical to developing viable REDD+ strategies and implementation frameworks.’
Tanzania has a population crucially dependent on local forest use, in particular as a source of fuel wood (World Bank, 2008; TNRF, 2009), and has one of the highest deforestation rates in Africa (FAO, 2011). The country is one of the most active REDD+ countries in Africa, with the first bilateral REDD+ agreement with Norway signed in 2008 (Angelsen, Reference Angelsen2009). Although local REDD+ pilot projects include result-based incentives, the Tanzania-Norway agreement is not performance based and has strong elements of traditional aid cooperation (Angelsen, Reference Angelsen2016). Most land in Tanzania is under the control of the government or under various community tenure regimes, from de facto open access to well-enforced customary arrangements (Treue et al., Reference Treue2014). Pilot projects, enacted through local NGOs, are therefore heterogeneous, where different payment modalities have been tried, often in tandem with PFM (URT, 2012; Lund et al., Reference Lund2017).
4.2 Study area and sampling
During the period September–November 2014, we conducted experiments and surveys in three regions of Tanzania: Geita in the tropical northwest of the country, Kilimanjaro in the northeast and Lindi in the arid southeast of the country. In each of the three regions, we selected five villages in collaboration with district authorities or local NGOs. The villages vary in key attributes, as summarized in online appendix A. The geographical and village attribute distributions aim to make the results more generalizable.
Within each village, we randomly selected 32 participants by drawing from the village registry (an already existing list of the households in the village, but which often needed to be updated). Each participant was then randomly assigned to a specific session with a predetermined treatment. The invitation stated that they would be compensated with a small, unspecified amount of money. With eight participants in 60 experimental sessions the total sample is 480 participants.
Appendix A presents descriptive statistics at the individual level and compares the mean values for the treatment and control groups. No significant differences across groups indicates successful random treatment assignment.Footnote 5 As described in related studies (Handberg and Angelsen, Reference Handberg and Angelsen2015, Reference Angelsen2016), the sampling technique creates variations among participants, including in forest use. In this sample, 84 per cent of the participants state that they collect forest products themselves.
4.3 Experimental design
We conducted a framed field experiment, meaning that the sample (local forest users) and the framing (local forest use and conservation) are relevant (Harrison and List, Reference Harrison and List2004). The design draws on Cardenas (Reference Cardenas2004) and Handberg and Angelsen (Reference Handberg and Angelsen2015). In each session, the eight participants are collectively endowed with a stock of 80 cardboard trees. Each tree, depicted in Handberg and Angelsen (Reference Handberg and Angelsen2015), privately pays TZS 100 (US$ 0.06) to the participant who harvests it.Footnote 6 In each of nine rounds, the participant privately decides how many trees to harvest, with five trees being the upper limit.Footnote 7 The participant has to physically tip each tree to indicate harvest, observed by an enumerator, who takes note and replants the harvested forest, such that the next participant faces the same forest size (decisions are made as if simultaneous). At the end of the harvest round, the aggregate number of trees harvested is revealed to the participants and removed from the forest. Thereafter, the forest grows by two trees for every ten trees standing. The sessions lasts for nine rounds, or until the forest depletes to less than eight trees. The number of rounds and all other parameters are made known to the participants.Footnote 8
4.4 Treatments
In a 2 × 3 between-groups design, the sessions were randomly selected into choice (192 participants) and no-choice (192 participants), in addition to a control group (96 participants). Under the two treatment groups, sessions were again randomly selected into three sub-group treatments, 20 per cent, 60 per cent and 100 per cent PES (with 64 participants, i.e., eight sessions, in each group).
In the choice treatment, participants were asked to choose one of two additional payment schemes (in addition to the TZS 100 per tree harvested) before the first round: (i) sell 40 of the trees immediately for TZS 50 each, implying a private payment of TZS 250 before the session starts, with an initial forest size of 40 (instead of 80) trees, or (ii) receive 20/60/100 TZS for each tree not harvested in each round, i.e., the difference between the upper limit and the harvested amount of trees. One of the three payments (20 per cent, 60 per cent or 100 per cent, measured relative to the value of a harvested tree) was presented at random (with examples for clarification), creating the three sub-treatment groups.Footnote 9 As the payments are incentives for decreasing forest use, scheme (ii) simulates a payment for environmental services (PES) scheme, and is presented as such. The participants were given three minutes to discuss privately with each other, before an anonymous referendum was held. The scheme was then chosen by majority vote.
In the no-choice treatment, participants had scheme (ii) imposed exogenously (with one of the three payment levels). These participants also communicated in private for three minutes before the first harvest round.
In the control group, the participants experienced no PES scheme, but they also communicated in private for three minutes before the first round.
Besides SoO, table 1 presents four potential mechanisms through which participation can improve institutional performance. The presented experimental design controls for these potential confounding effects. First, local knowledge was not used to select the treatment best suited, as all sessions chose the superior scheme (ii). This also ensures that choice has no direct impact on payoffs. Second, participants under choice are not given information or benefits that empower them more than others. Third, there is no additional transparency under choice than under the other two groups. Fourth, participants in the choice and no-choice treatment groups and in the control group are all allowed to communicate in private, thus creating the same conditions for dialogue.
As presented in section 3, the treatment relies on the assumption that choosing is an integral part of participation and that it is the choice that creates sense of ownership, and on the assumption that the participants perceive it as an actual choice. The latter will be discussed in section 6.
4.5 Theoretical predictions
As presented in the related designs of Handberg and Angelsen (Reference Handberg and Angelsen2015, Reference Handberg and Angelsen2016), the social dilemma is whether to maximize own payoffs by harvesting the maximum amount of trees each round, or harvest fewer trees to improve the total payoff.
Under both 0 per cent and 20 per cent PES, the optimal decision for a payoff maximizing participant is to harvest the maximum number trees which are allowed in each round, independent of the choices of the other participants in the session (the Nash equilibrium). Under 60 per cent PES, the optimal decision depends on the beliefs the participant has about the decisions of the other participants in the same session. If a participant believes the others to be either selfish (harvest close to maximum) or strongly pro-social (harvest close to nothing), the optimal decision is to harvest the maximum amount of trees allowed in each round. If the participant believes that the mean decision of the others is close to neither corner, the optimal decision is to harvest less. Under 100 per cent PES, there is no incentive to harvest any trees, independent of the choices of others. These predictions are presented in full in Handberg and Angelsen (Reference Angelsen2016).
Choosing PES scheme or having the same scheme imposed should not affect the harvest decision of the participants, as the parameters are identical. SoO, however, suggests that participants who choose the PES scheme are more positive to the intention of the scheme and should thus harvest less (increase effectiveness). The theory and previous studies presented in section 3 thus suggest that the payment schemes should perform better when chosen than when imposed. In this experimental setting:
H1. Participants harvest more trees under no-choice than under choice
Biased information processing predicts that participants who choose the PES scheme, evaluate PES more favorably than others, which makes the PES scheme more effective among participants in the choice group. In the post-experiment interview, we indirectly asked the participants about their perceptions of the PES concept (questions 24–26, appendix E). On a scale from 1 to 5, where 5 is strongly agree and 1 is strongly disagree they responded to the following statements: ‘it is right that those who benefit from the clean air that our forests produce, contribute to conserving the forest’ (normative evaluation), ‘it is not proven that paying for living trees decreases deforestation’ (factual evaluation), ‘paying for living trees makes other forest use considerations less important; like tradition, culture and religion’ (PES and crowding-out). The hypothesis is thus:
H2. Choice induces more positive participant evaluations of PES schemes and the treatment effect is stronger among these participants
Associative self-anchoring predicts that the positive effect is stronger the more positive the participant evaluates themselves. In the post-experiment questionnaire, we asked about the participants' first, intuitive reaction to four letters, which they rated on a scale from one to five (strongly dislike, dislike, ambivalent, like, strongly like) (question 32, appendix E). Participants were (orally) presented with the letters E and R, then the first letter in their stated first name and one letter not in their stated names. The theory predicts that the participants on average rate the letter in their first name more favorably than the letter not in their name. Similarly, participants with names which includes E and/or R, should on average rate the letter(s) more favorably than participants whose name does not include the letter(s). The theory predicts that participants who highly rate letters in their own name (in absolute terms or relative to the rating of other letters), should have a stronger sense of ownership. The treatment effect should thus be stronger among these positive implicit self-evaluators, implying the following hypothesis:
H3. The choice effect on harvest rates (relative to no-choice) is stronger among participants with higher implicit self-evaluation
Cognitive dissonance predicts that the positive effect is stronger the more equal the options are, as the potential for cognitive dissonance is stronger here. This implies that the effect of choice is stronger the lower the PES level is. The specific hypothesis is:
H4. The choice effect on harvest rates (relative to no-choice) is stronger among participants that experience the low PES level (20 per cent versus 60 or 100 per cent)
A plausible outcome measure for forest conservation is decrease in extractive forest use. The following will therefore consider reduction in number of harvested trees as the measure of success of the payment scheme. As the upper limit is relative to the forest size, analyses at absolute numbers could create biases. A harvest rate [0–1] indicating the number of harvested trees relative to the upper limit is therefore used.Footnote 10
5. Results
5.1 Sense of ownership
Table 2 summarizes the individual mean harvest rates by the treatment and control groups. Here and in the following, the 17 participants voting for scheme (i) are excluded.Footnote 11
Mean individual harvest rates
Session clustered standard errors in parentheses.
Without the PES scheme, the mean harvest rate is lower than the Nash equilibrium (0.57 < 1).Footnote 12 The mean harvest rates under the two treatment groups (aggregating the three PES levels) are both significantly lower than the mean harvest rate under the control group, but they are not significantly different from each other (p = 0.488). The test thus lends no support for H1.Footnote 13
5.2 Biased information processing
The biased information processing hypothesis predicts that participants who chose the PES scheme should evaluate PES more favorably than participants who had the scheme imposed. The three elicited PES evaluations – normative evaluation, factual evaluation, and PES and crowding-out – rank PES perception from 1 to 5. Table 3 summarizes the evaluations separately by the choice treatment. It reveals that there are no significant differences between the group's mean evaluation in any of the three statements. There is therefore no support for the first requirement of H2, and the hypothesis is rejected.
Participants' PES evaluation by the choice treatment
Standard errors in parentheses. p-values produced by Wilcoxon ranksum tests.
5.3 Associative self-anchoring
The associative self-anchoring hypothesis predicts that: (i) the participants' mean implicit self-evaluation should be positive; and (ii) the treatment effect is stronger for participants with positive implicit self-evaluation than for others. Table 4 presents the mean implicit self-evaluation of the participants. The table reveals that participants evaluate the letters E and R more favorably if it is the first letter in their stated first name. The participants also evaluate the first letter in their first stated name more favorably than a letter not in their name (see appendix C for the distribution of letters). The finding is not as clear in their evaluations of E and R when their stated names include the letters elsewhere as when their names do not include the letters; R is evaluated more favorably, but E is not. Taken together, the data supports previous findings in that people tend to evaluate letters in their own name more favorably than other letters, which supports prediction (i).Footnote 14
Mean implicit self-evaluations by letter
Standard errors in parentheses. p-values produced by Wilcoxon ranksum tests. Random letter refers to two letters randomly chosen among those in the participant's name and those not in the name.
Table 5 tests for associative self-anchoring by the interaction effect of the choice treatment and positive implicit self-evaluation on individual mean harvest rates. Model (1) defines positives as those who value the letter in their name positively (like or strongly like on the five-point Likert scale) while the remainders constitute the negatives and neutrals (strongly dislike, dislike or ambivalent). Model (2) defines positives as those who value the first letter in their name more favorably than the letter not in their name, and the others as negatives and neutrals.
Testing for associative self-anchoring
Dependent variable: individual mean harvest rate through the session.
Session clustered standard errors in parentheses. ***, **, *: significant at the 1, 5 or 10% level. Includes village dummies (not reported).
With no significant impact on mean harvest rates of the choice in table 5, among either the positives or among the neutrals/negatives, there is no support for H3.
5.4 Cognitive dissonance
The cognitive dissonance hypothesis predicts that the treatment effect is stronger the lower the PES level is, as this implies that the two payment schemes presented to the participants are more similar.Footnote 15 Table 6 reports the mean harvest rates by the three payment levels under no choice and choice.
Mean individual harvest rates by PES level
Session clustered standard errors in parentheses. ***, **, *: significant at the 1, 5 or 10% level.
Investigating the means by the three payment levels reveals that the participants respond to increasing the PES payment. Comparing the means within each PES level reveals no significant impact of the choice treatment at any of the three PES levels, thus leaving no support for H4. Appendix B provides robustness tests by regressing harvest rates on PES and choice treatments in interaction, with controls. Further tests, exploring first-round harvest rates to avoid the round dynamic of the experiment, are also presented in appendix B. All robustness tests produce similar results and the same conclusions as here.
6. Discussion
In the presented framed field experiment, there is no significant impact on institutional effectiveness (measured as decreased forest use) of allowing participants to choose a PES scheme relative to imposing the same scheme. Examining the data by three theories explaining the possible effect – biased information processing, associative self-anchoring and cognitive dissonance – finds little impact of the choice treatment. The following subsections discuss the results in light of being context specific, internally or externally invalid, and sensitive to the participation form.
6.1 No significant sense of ownership in this context
One explanation is that SoO does not apply in the context of incentive-based forest conservation among local forest users in Tanzania. The SoO hypothesis might underestimate the rationality of Tanzanian forest users; they support good policies and oppose what are considered bad policies, independent of the source of origin or design.
The specific context of the empirical test is important. Tanzania has a history of decentralized forest management, also at the village level (Blomley and Ramadhani, Reference Blomley and Ramadhani2006; Lund and Treue, Reference Lund and Treue2008). The sampled participants should thus be familiar with participation in the three stages of figure 1. Still, choices tend to be made by village leadership, such as the forest committee, not directly by the forest users (Lund and Treue, Reference Lund and Treue2008). Unfortunately the sub-sample of village leaders is too small for clear inference.
The treatment also investigates the hypothesis in the context of incentive-based forest conservation. The PES scheme is a particular institution that leaves more freedom to the forest user than, for instance, a command and control institution. Previous experimental studies find a dividend of democracy in more constraining institutions (Tyran and Feld, Reference Tyran and Feld2006; Dal Bó et al., Reference Dal Bó, Foster and Putterman2010; Sutter et al., Reference Sutter, Haigner and Kocher2010; Markussen et al., Reference Markussen, Putterman and Tyran2013). No observed effect in this context does not necessarily imply that there is no effect under other institutional arrangements. Furthermore, although there are REDD+ pilot projects in Tanzania, the initiative is unknown to most forest users. Inability to properly relate to the chosen PES scheme in the experiment might reduce the impact of SoO.
Lastly, the treatment seeks to test for participation in the design stage. SoO could, for instance, apply more strongly in the initiative stage than in the design stage, and could also have a stronger impact on efficiency or equity than on effectiveness.
6.2 External validity and design issues
In addition to external validity more broadly, a necessity for rejecting the SoO hypothesis in this context is treatment validity, i.e., that participants' response to the treatment is predictive for real-life responses of participation in designing PES schemes (Handberg and Angelsen, Reference Handberg and Angelsen2015).
The treatment relies on the assumptions that choice is an integral part of local participation and that the potential SoO works through choice. The latter assumption is supported by theories within psychology and behavioral economics, and these specific theories are tested. The former assumption is supported by scholars on participation (e.g., Arnstein, Reference Arnstein1969; Pretty, Reference Pretty1995; Cornwall, Reference Cornwall2008; Reed, Reference Reed2008) who emphasize that participation should involve local stakeholders in decision-making processes.
Although the participants are presented with a choice, the choice treatment could be too weak, which may lead to a type II error. Allowing the participants to choose between two payment schemes, where one is inferior to the other, has the desirable attribute that it controls for selection effects. The cost is that the treatment presents a rather easy choice; only minor cognitive efforts are necessary to infer that one is preferred over the other. The participants can feel (rightly so) that they make no real choice.
An additional caveat is that the harvest rates are low, even under no and low PES. This limits the improvement potential of the choice treatment. Still, mean harvest rates under 60 per cent and 100 per cent PES are significantly lower than the mean under 20 per cent PES, indicating that some potential exists.
6.3 Weak versus strong participation
Related to the weak treatment caveat, SoO may depend on the degree of real influence the stakeholders exercise through their choice. The ranking in figure 1 defines control as strong participation, and consent and consultation as weaker forms of participation.
The choice in the experiment is informed and free, but one option is clearly superior to the other. An obvious choice leaves little power to the stakeholder. The choice is also limited to the two payment schemes, meaning that the degree of participation is arguably closer to ‘consent’ than ‘control.’ The lack of support for the SoO hypothesis could thus be due to the treatment's weak form of participation. Under 20 per cent PES, the choice should be perceived as less obvious since the options are more similar than under 60 per cent and 100 per cent PES. A decreasing difference in mean harvest rates by choice and a reversal of difference would indicate an effect of choice when the options of the choice are more equal, but the trend here is not significant (table 6). Both cognitive dissonance and associative self-anchoring predict stronger choice effects with higher cognitive efforts.
The lack of a clear overall treatment effect (and positive findings of others, presented in section 2.2) could thus be an argument for stronger participation. Weak participation, such as participatory rhetoric without actual participation, could then be not only immoral, it could also, from a project proponent's viewpoint, be a cost with no benefits. In contrast, strong participation might not only be an end in itself, but also have positive effects on institutional performance.
Participation simply as a management tool is arguably a contradiction in terms. Strong participation implies stakeholders' opportunity to change the rules of the game, without strict constraints by proponents. SoO may require strong participation to have a real impact and improve institutional performance.
7. Conclusion
Participation of local stakeholders in initiating, designing or implementing institutions is often claimed by scholars and practitioners to improve institutional effectiveness, efficiency or equity, through at least five possible mechanisms. Sense of ownership, as one of these mechanisms, is often advocated within the development and nature conservation communities. Theoretical and empirical justifications remain scarce. Possible theoretical support is found within psychology and behavioral economics, and a few lab experiments on the effects of democracy give empirical support.
The experimental study prsented in this paper tests the effect of participation in the design stage on institutional effectiveness, through the sense of ownership hypothesis. The results lend no support to the hypothesis in the context of incentive-based forest conservation among local Tanzanian forest users.
One possible explanation for the lack of support is that the experiment represents a weak form of participation. An implication is that weak participation is not only problematic when viewing participation as an end, but also when viewing it as a means to improve institutional performance. The finding questions the view that (weak) participation is a low-cost strategy to establish sense of ownership among local stakeholder, and thus improve performance. The often-observed weak participation of local stakeholders in development and nature conservation institutions in developing countries could thus have little value except pleasing donors and the public.
At a societal level, participation is normally either advocated as an end in itself (e.g., Sen, Reference Sen1999) or as a means to achieve other ends, such as economic security (e.g., Arnstein, Reference Arnstein1969; Sen, Reference Sen1999). For project proponents on strict budgets with specific goals, these arguments might not motivate stronger participatory approaches. From the perspective of a project proponent, the latter argument is a positive externality and the former is normative. A better understanding of the sense of ownership hypothesis (and other mechanisms) can help proponents initiate, design or implement more effective, efficient or equitable institutions. In addition, revealing such positive effects of participation creates incentives for stronger and more extensive participation of local stakeholders. Research exploring positive effects of local participation on institutional performance thus incentivizes proponents to integrate participation in designing and implementing institutions, and to facilitate local initiatives.
Supplementary material
The supplementary material for this article can be found at https://doi.org/10.1017/S1355770X18000190.
Acknowledgements
The author appreciates financial support from the Center for International Forestry Research (CIFOR) and from the Climate Change Impacts, Adaptation and Mitigation (CCIAM) programme, both funded by the Norwegian Agency for Development Cooperation (Norad). The author is grateful for valuable comments by Frode Alfnes, Arild Angelsen, Ann-Kathrin Koessler, Nina Bruvik Westberg, seminar participants at Nordev15 and two anonymous referees.
