Payment for Environment Services to Promote Compliance with Brazil’s Forest Code: The Case of “Produtores de Água e Floresta”

Payments for ecosystems services (PES) can promote natural resource conservation by increasing compliance with environmental laws. Law enforcement and PES proponents assume that individuals make decisions about compliance based on expectations of gains, likelihood of being caught in non-compliance, and magnitude of sanctions. Brazil’s Forest Code, characterized by low levels of compliance, includes incentive and disincentive mechanisms. We interviewed landowners in the Atlantic Forest to understand their motivations to participate (or not) in a PES project, the effects of knowledge and perceptions of environmental regulations on compliance, and how both environmental regulations and PES affect land management decision-making. We found that neither expectations of financial gains nor PES payments drive behavioral change and that the perception of systemic corruption reduced compliance with environment regulations. There were important behavioral differences between long-term residents for whom the land is their main source of income and recent residents with little dependence on land-generated income.


Introduction
Natural conservation can be promoted both by compulsory (laws) and voluntary-mixed tools like payments for ecosystems services (PES) (e.g., [1]). PES are administered in places already subject to other environmental policies, and in many places PES provide incentives to comply with existing policies [2,3]. Regulations and incentives are important elements to signalize societal norms and expectations [4]. Mixing compulsory regulatory disincentives with voluntary incentive-based instruments can potentially improve ecosystem services and increase incomes [5]. Policy mixes or combinations can improve the quantity and quality of ecosystem services provided [6]. Evidence of the efficacy of policy mixes of PES with existing regulations has grown [7][8][9]. Most analyses are theoretical, based on little empirical evidence [10]. This study contributes to this literature using qualitative data that reveal facets of landowner decision-making about compliance with existing environmental laws and participation in PES programs. We used a case-study to investigate how forest conservation instruments using incentives and disincentives influence land use decisions.
Environmental behaviors may result from interactions between emotions, attitudes, beliefs, identities, knowledge, worldviews, and values embedded in social and cultural contexts combined with skills and opportunities to act [11,12]. Environmental behaviors are all types of behavior that change  -never attended school: 2 (11%) -elementary school: 11 (58%) -high school: 4 (21%) -technical school: 1 (5%) -college: 1 (5%) We conducted semi-structured interviews (34-150 min, Table A1 provides summary of interview questions) to explore why participants chose to join or not join PAF and why people complied or not with the FC (University of Florida IRB201701354). The interviews covered four topics: (1) general information about properties and socio-economic characteristics of the landowners; (2) perceptions about forests and ecosystem services, including explanations of the distribution of forest and regrowth areas on the land; (3) perceptions of and knowledge about environmental regulations; and (4) knowledge about the project and their motivations for joining or not joining PAF [41]. Procedures were used to develop the protocol, which was reviewed by three researchers with experience in conservation and sustainable agriculture. We revised the interview guide to accommodate local language and culture after testing the interview instrument with five Rio Claro landowners of various socio-economic backgrounds.
Data analysis involved four steps ( Figure A2). Steps 1 and 2 are commonly used in many types of qualitative data analysis [42,43].
Step 1, topical coding, identifies specific ideas in the individual responses and captures each respondent's comments based on transcripts of the interviews and the researchers' notes and case summaries [44,45]. In Step 2, thematic coding, we grouped similar concepts generated in Step 1 into themes and identified relationships among them [46,47]. These larger frames helped us understand respondents' more global views about relevant topics and to identify broader commonalities among participants than the specific concepts in Step 1. In Step 3, analytic coding created mental models of respondents' representations of a condition or process and the relationships among the concepts [48]. We identified components in the respondents' mental models that help Guandu Watershed agency (AGEVAP) provided a geo-referenced database showing property boundaries of PAF participants and areas designated for reforestation and provided descriptive information about the participants. We contacted all but six of 67 PAF participants, two of whom were unwilling to respond. Six were unresponsive to ten phone calls and two visits to their properties on different times and days. We sampled to match participants and non-participants on socio-economic traits and opportunity cost proxies (land characteristics). Owners were not in residence on many properties and we therefore used intercept sampling at other locations frequented by landowners, the local rural labor union and the local technical assistance and rural extension office (EMATER). We interviewed 19 landowners who chose not to participate in PAF and two who enrolled but dropped out of the project before receiving any payment (Table 1 provides demographic traits of the respondents in the two comparison groups). We conducted semi-structured interviews (34-150 min, Table A1 provides summary of interview questions) to explore why participants chose to join or not join PAF and why people complied or not with the FC (University of Florida IRB201701354). The interviews covered four topics: (1) general information about properties and socio-economic characteristics of the landowners; (2) perceptions about forests and ecosystem services, including explanations of the distribution of forest and regrowth areas on the land; (3) perceptions of and knowledge about environmental regulations; and (4) knowledge about the project and their motivations for joining or not joining PAF [41]. Procedures were used to develop the protocol, which was reviewed by three researchers with experience in conservation and sustainable Sustainability 2020, 12, 8138 5 of 51 agriculture. We revised the interview guide to accommodate local language and culture after testing the interview instrument with five Rio Claro landowners of various socio-economic backgrounds.
Data analysis involved four steps ( Figure A2). Steps 1 and 2 are commonly used in many types of qualitative data analysis [42,43].
Step 1, topical coding, identifies specific ideas in the individual responses and captures each respondent's comments based on transcripts of the interviews and the researchers' notes and case summaries [44,45]. In Step 2, thematic coding, we grouped similar concepts generated in Step 1 into themes and identified relationships among them [46,47]. These larger frames helped us understand respondents' more global views about relevant topics and to identify broader commonalities among participants than the specific concepts in Step 1. In Step 3, analytic coding created mental models of respondents' representations of a condition or process and the relationships among the concepts [48]. We identified components in the respondents' mental models that help explain how participants think about participation in PAF and to understand commonalities and differences among respondents' views of the roles of the FC and PAF in their lives. We printed statements of the themes and arranged them into representations that reflect thought processes and interactions among themes. In Step 4, we assessed the degree of model agreement or "fit" with the majority of cases [46,49] and created explanatory models. Finally, we re-examined our theoretical model and modified it based on our results [50,51], creating a model that contributes to the theoretical understanding of socio-environmental systems.

Results
The interviews produced 267 first-level (Step 1) codes for the 59 participants and 158 for the 21 non-participants. We modified our initial theoretical model of the relationships between law enforcement, participation in PAF, and environmental additionality of PAF to reflect our findings. The explanatory models of decision-making ( Figure 2) highlight (1) important distinctions between outsider and insider groups; (2) the impacts of corruption and lack of trust in government on decision-making processes; and (3) pathways that reflect added constructs in decision-making among insiders and outsiders and between PAF participants and non-participants. explain how participants think about participation in PAF and to understand commonalities and differences among respondents' views of the roles of the FC and PAF in their lives. We printed statements of the themes and arranged them into representations that reflect thought processes and interactions among themes. In Step 4, we assessed the degree of model agreement or "fit" with the majority of cases [46,49] and created explanatory models. Finally, we re-examined our theoretical model and modified it based on our results [50,51], creating a model that contributes to the theoretical understanding of socio-environmental systems.

Results
The interviews produced 267 first-level (Step 1) codes for the 59 participants and 158 for the 21 non-participants. We modified our initial theoretical model of the relationships between law enforcement, participation in PAF, and environmental additionality of PAF to reflect our findings. The explanatory models of decision-making ( Figure 2) highlight (1) important distinctions between outsider and insider groups; (2) the impacts of corruption and lack of trust in government on decision-making processes; and (3) pathways that reflect added constructs in decision-making among insiders and outsiders and between PAF participants and non-participants.
Among both participants and non-participants there are two broad groups with divergent approaches to land-use decision-making. We labelled one group "insiders," referring to landowners born in the area and raised as farmers, and the other "outsiders," property owners with non-agricultural backgrounds. An individual's background affects land use decision-making and objectives for property management. We henceforth distinguish between insiders and outsiders because it was the most consistent distinction among respondents, including differences between PAF participants and nonparticipants. There were more outsiders among the sample of participants (61%) than that of nonparticipants (29%). Many PAF participants were relatively wealthy and well-educated, and included large landholders who lived in urban areas and did not rely on farm income. We built models for insider participants, outsider participants, insider non-participants, and outsider non-participants (Figure 2a   The insider-outsider groups differ fundamentally in their reasons for owning property in the area. Insiders, both participants and non-participants, use their property mainly for income whereas outsiders use it for leisure or supplementary income. Several PAF participants' comments capture the importance of income: "We already worked in the rural area, then there was an opportunity to buy a property and we did it… We want to keep using the property for our family livelihood" (PAF11, male, 79-year-old insider). Insider non-participants shared similar ideas. For example, one said: "All that we Among both participants and non-participants there are two broad groups with divergent approaches to land-use decision-making. We labelled one group "insiders," referring to landowners born in the area and raised as farmers, and the other "outsiders," property owners with non-agricultural backgrounds. An individual's background affects land use decision-making and objectives for property management. We henceforth distinguish between insiders and outsiders because it was the most consistent distinction among respondents, including differences between PAF participants and non-participants. There were more outsiders among the sample of participants (61%) than that of non-participants (29%). Many PAF participants were relatively wealthy and well-educated, and included large landholders who lived in urban areas and did not rely on farm income. We built models for insider participants, outsider participants, insider non-participants, and outsider non-participants (Figure 2a-c).
The insider-outsider groups differ fundamentally in their reasons for owning property in the area. Insiders, both participants and non-participants, use their property mainly for income whereas outsiders use it for leisure or supplementary income. Several PAF participants' comments capture the importance of income: "We already worked in the rural area, then there was an opportunity to buy a property and we did it . . . We want to keep using the property for our family livelihood" (PAF11, male, 79-year-old insider). Insider non-participants shared similar ideas. For example, one said: "All that we obtained (could buy) in life came from this property" (N1, female, 49-year-old insider). Another said: "I like to do what I do, the necessity made me like it. I was born in a rural area and do not know how to do anything else" (N3, male, 50-year-old insider). Outsiders, in contrast, sometimes justified the way they thought about their property by emphasizing their non-rural backgrounds: "I do not perceive any drawback from having forest in my land because I do not depend on my property [for a living]" (PAF22, male, 65-year-old outsider). Another participant bought the land to conserve forest (PAF64, male 90-year-old outsider). Outsiders often mentioned that their family histories were not related to the rural lifestyle. "My father was an intellectual, not a farmer. I was raised in an apartment, but I always desired a rural life so I moved to the countryside" (PAF68, female, 67-year-old outsider). Another said: "My goal is to have a property that pays for itself because now I have to use resources from other sources of income to keep the farm" (PAF79 male 54-year-old outsider). Insiders typically used their land to generate income, whereas outsiders used it for leisure but were not averse to using it to generate income.

Motivations
Intrinsic pro-environmental motivations and the ease of joining largely drove individual landowners' decisions to join PAF and economic motivations and barriers to joining were the main reasons non-participants did not join. Financial gain was not the main reason for joining for most PAF participants. Over 80% of participants said that they would participate without the money although they commented that: "Any extra money is always good" (PAF50, Male, 54-year-old outsider). For outsider non-participants, an emerging theme was "heard about it but nobody offered it directly" and some said they might join if asked. Non-participant N11 (Male 52-year-old outsider) gave a typical explanation: "We found out about the program on TV but did not look for details. We do not intend to deforest anyway, so we do not need the incentive [to conserve forest]. But if the project directly asked us to join the program, we could join." Many non-participants outsiders said that PAF looked like a good opportunity, but that they did not trust the government. They thought that the government might stop payments once the forest grows back because forest land is legally protected. This justification was also provided by some insider non-participants when we asked about the minimum level of economic incentives that would make them join a conservation program and increase their pro-environmental behavior.
Economic motivations revealed in the interviews clearly included the opportunity costs discussed in the PES literature. For example: "Decision-making is based on what improves our situation, but we always focus on cattle because it seems to provide us more profit" (N22, male 32-year-old insider). Another said: "if I had more forest it would affect my income" (N26, male 56-year-old insider). The perception of restrictions on land use influenced economic motivations on land use decisions. Participant PAF68 (Female 67-year-old outsider) stated: "In the 1980s everyone started to plant brachiaria grass when rural laborers started to migrate to nearby cities . . . brachiaria spreads Sustainability 2020, 12, 8138 8 of 51 because it does not require much labor." The high opportunity cost of allowing the pasture to become forest was the most common reason non-participants gave for not participating. Non-participants went so far as to describe their individual "opportunity cost math" to show that the payments were not enough and said that they would not participate because pastures are more useful than forests. Participant NP17 (male 58-year-old insider) said: "I heard PAF's proposal, but I thought it [the payment] was too little . . . I could get more planting yams. And in the future, what will I leave for my children and grandchildren if I sign? It is permanent." Diverse ideas emerged along with these common themes. Two non-participants pointed out that the effort and cost to society to restore forest to increase water quality and availability would have no impact compared to the addressing lack of sewage treatment facilities in cities. This implies that they believed reforestation would require a major individual effort but produce little benefit for society. One respondent indicated that he would join PAF if everybody in the region did so, suggesting that a collective effort would justify the individual cost.
Enrolment of communal land by quilombolas, slave descendants, deserves special attention because community-owned and private owned land are often treated differently in PES projects. The quilombola community in Rio Claro includes about 240 people in 55 families, 85% of whom depend entirely on their land for their livelihoods. The community association decided to join PAF because they had once made and sold charcoal to the steel industry, which resulted in extensive forest degradation, and they believed that participation in an environmental project would improve their public image. When the community joined the project, its members were not aware that the project would provide payments. By the time of this study, PAF had become the main source of financial income for the association and provided members with jobs in reforestation on private properties.
ELR compliance was another motivation for PAF participation. Many participants indicated their awareness of FC requirements and believed that most landowners could comply without compromising their livelihoods. Participants expressed different perceptions of the likelihood of enforcement of environmental laws in general, even though most knew someone who had been sanctioned for non-compliance. Many perceived increased environmental awareness and law enforcement over the previous two decades.
The theoretical approach in Figure 1 suggests that sanction size and likelihood of detection would be the most important factors in reducing the utility of acting against the law. This was true in Rio Claro where many people had been jailed for deforestation and the perceived probability of getting caught was high. These conditions decreased motivation to violate this legal requirement, showing that laws can drive pro-environmental action. Nonetheless, fines were perceived as less effective than jail terms in reducing illegal actions and there is a perception that the restoration required by the FC is not enforced. Landowners understood that authorization was required to change land-use if they allowed trees to regrow. In consequence some landowners voiced the idea of a "negative opportunity cost," willingness to pay to suppress forest regrowth even if they do not plan to use the pasture in the near future.
Corruption of environmental enforcement officers and "the government" as an institution was a predominant emergent theme that greatly affected perceptions of the fairness of environment laws. Many landowners remarked that their decision to comply with environmental regulations included consideration of the inequitable enforcement of sanctions. Many did not believe that enforcement was equal for small and large landowners, saying that the latter could bribe officers or even influence the creation of laws.

Relationships between Themes
PAF's modest fiscal incentives did not appear to reduce other motivations for FC compliance. One reason cited for compliance was the importance of forest, especially the hydrological benefits it provides. Participants learned about these benefits from the project, which could have increased motivations to conserve. We did not investigate changes in perception due to the project, but many Sustainability 2020, 12, 8138 9 of 51 landowners said that they traditionally conserved forest near springs suggesting that information from the project only reinforced their prior knowledge.
The reason for owning property influences how landowners perceive FC restrictions on land uses. Increased restrictions were widely perceived as negative because they reduce landowners' sense of self-determination. Many insiders perceive FC compliance primarily as a restriction, one saying, "The forest is untouchable" (PAF 78, male, 65-year-old insider) and many felt that part of their property was "not really theirs to manage." PAF does not include landowners in choosing species for reforestation. Some argued for change because some landowners prefer species for aesthetic value or potential income generation.
Labor constraints influenced land-use decisions. One non-participant insider said that: "There are not enough people to work the land. Our children grew up and left and we can't afford to hire other people" (N10, male, 78-year-old insider). Some outsiders who initially planned to use the property as a source of income pointed out that labor availability limited this option: "In the beginning I was thinking about raising cattle, but since it is hard to find labor in the region, I gave up" (PAF51, male 66-year-old outsider). Other limitations also emerged: "The worst here are the roads. It's hard to maintain production or any other activity" (PAF05, male 73-year-old outsider). In contrast, some landowners who valued their land for leisure did not want paved roads "because it would make me lose my privacy" (PAF40, male 73-year-old outsider).
Intrinsic and social motivations for conservation indicate reasons for owning property which influences motivations for conserving forest. Conserving forest and its aesthetic value and a desire to fulfil a family dream were motivations for owning land. Owning property seemed to strengthen intrinsic and social motivations for pro-environmental action. Respondents frequently mentioned beauty and tranquility as forest ecosystem services, discussed their desire to maintain forest for future generations or for wildlife, and indicated that maintaining forest was a lesson that their family had learned. Intrinsic motivations prevailed among outsiders whereas insiders were more driven by perceived economic benefits.
Use of PES in a policy-mix context may be problematic. Most of PAF's money went to a minority of landowners with large properties [52] even though most PAF participants have small holdings by Brazilian standards. This bias has implications for project effects on real and perceived equity. Some participants said that they received meagre benefits from PAF while other people received substantial benefits to protect forest that law requires them to preserve. PAF sometimes describes its accomplishments by the number of participating properties, but care is needed when using this measure because PAF does not have a minimum property size requirement. Four participants owned < 1 ha and three < 5 ha. Further, in addition to PAF payments, five landowners received a second and much larger PES payment as an incentive to transform part of their land into a Private Reserve of Natural Heritage (RPPN from Portuguese acronym), a category in the national protected area system.
We modified our theoretical model to reflect findings ( Figure 3). For the Rio Claro context, the type (fines or jail) of sanction resulted in different decreases in utility of acting within the rule, where jail reduced illegality, but fines seems to have no or small implications in the decreased motivation to violate legal requirements. The perception of corruption is important to explain this and also how people perceive the fairness, size of sanctions and likelihood of detection, therefore influencing causal path of rule enforcement to reduce utility of acting against the rules. Furthermore, in Rio Claro PES enrolment was not mainly motivated by increasing compliance with the rules. So, although it does increase the utility of acting within rules, we removed this box from the model. Finally, we highlighted the feedback loops between behavior decision, behavior change and environmental additionality.
violate legal requirements. The perception of corruption is important to explain this and also how people perceive the fairness, size of sanctions and likelihood of detection, therefore influencing causal path of rule enforcement to reduce utility of acting against the rules. Furthermore, in Rio Claro PES enrolment was not mainly motivated by increasing compliance with the rules. So, although it does increase the utility of acting within rules, we removed this box from the model. Finally, we highlighted the feedback loops between behavior decision, behavior change and environmental additionality.

Contributions to the PES Literature
Although transaction costs accrue substantially when PES projects include numerous smallholders [53], PAF was accessible to landowners large and small, poor and wealthy. In contrast to some PES projects [19,54,55], property size did not seem to influence the likelihood of participation in PAF. Our finding that

Contributions to the PES Literature
Although transaction costs accrue substantially when PES projects include numerous smallholders [53], PAF was accessible to landowners large and small, poor and wealthy. In contrast to some PES projects [19,54,55], property size did not seem to influence the likelihood of participation in PAF. Our finding that outsiders were prevalent among PAF participants is similar to [17] findings for Costa Rica. Compared to non-participants, smallholders and larger landowners enrolled in a PES tended to be older and wealthier, had access to non-farm salaries, and participated only marginally in agriculture. In contrast, PAF participants included some insiders who depend on farm income and had limited formal education. Our interviews reflect findings by [56] that willingness to participate was negatively associated with availability of family labor and with the fear of changing production patterns characteristic of low-income, farm-dependent landowners.
The data we collected support findings that perceptions toward conservation and intrinsic environmental motivations drive participation in PES [19,23,[57][58][59]. Other researchers [60,61] who question the rational choice theories that underlie many PES schemes report that the opportunity cost principle is often only loosely relevant. Nevertheless, similar to findings by [62] in Mexico, many PES participants "like forest but also like cash" and the payments provided an important additional incentive. PAF participants' recognition of the importance of forest for essential environmental services, especially hydrological, could stimulate long-term behavioral changes and promote more conservation [63].
Monitoring and enforcing conditionalities of pro-environmental behavior that were agreed upon in exchange for enrolment in PES are necessary incentives for effective conservation [64]. PAF monitors compliance but lacks clear protocols for dealing with non-compliance. PAF's contracts state that non-compliant landowners can receive only partial payments or be excluded from the program, but as of 2018 no PAF participants were penalized for non-compliance (AGEVAP, personal communication). PES implementers often tolerate some non-compliance and only one-fourth of PES projects described in the literature report sanctions for non-compliance [64]. The authors in [64] point out that local politics and budgetary constraints often have greater influence on PES enforcement than budgetary constraints. This appears true for PAF, which tolerates some non-compliance in recognition of the time needed to build trust with participants.

Insights from ELR
Landowners' personal financial gains can compensate for their potential financial losses from illegal actions, which influences the behavior decision of whether to deforest or suppress forest regrowth in ecological buffer zones. This situation reveals an inherent tension between individual and social assessments of ecosystem service value. Many factors affect the likelihood of collective action when faced with such social dilemmas [65], but enforcing regulations is often critical insofar as it influences individual perceptions of costs [66,67]. PAF participants and non-participants were aware of people being fined or jailed for deforesting, but perceived differences in the sanctions. People feared jail and would avoid deforesting but did not fear sanctions for failure to reforest. A study in the Amazon Biome found that only 6% of the Rural Environmental Registry (CAR) registered producers were taking steps to restore illegally cleared areas on their properties; and suggested that full compliance with the FC offered few economic benefits from the landowner's perspective [68].
Rio Claro property owners who depend on land for their livelihood value cleared land more than forest. As noted earlier, some landowners pay the "negative opportunity cost" to suppress forest regrowth to protect their descendants' ability to make decisions about how to use the land in the context of tough enforcement of prohibitions on deforestation. This example shows how policies may intersect with behaviors and norms to result in unanticipated outcomes [69]. This viewpoint may also reflect the higher selling price of cleared land, which could be a vestige of decades of governmental incentives to clear land in Brazil [70]. Cleared land in Rio Claro sells for two to five times more than forested land despite its out-of-compliance status. Riparian cleared land would sell for less than forested land if the FC were enforced because the purchaser would have to pay for reforestation. The value of cleared land could also reflect a cultural tendency of Latin Americans to value pasture over forest [71].
The nature and magnitude of the effects of political corruption on compliance with environment regulations remains understudied [72] but corruption was an important theme in many interviews. Perceived corruption and unfairness, such as large landowners having fewer "real" obligations for legal compliance, are recognized as faults in the Brazilian legal system. Perceptions of corruption, unfairness, and impunity help explain low levels of compliance with environmental law in Rio Claro. Similarly, [68] highlighted that the perception of impunity severely weakens environmental policies to control deforestation in the Amazon. We therefore concur with [3] that incentive instruments cannot offset weak governance resulting from limited state capacity to enforce or corruption. Loss of respect for and confidence in environmental laws generated by perceived corruption deserves more attention.

Implications Emerging from the Empirical Findings
Brazil has extensive experience with diverse approaches to environment protection, PES being among the most recent. As in many regions where PES interventions are implemented, our study occurred in an area characterized by weakly enforced environmental regulations [73] but that is of tremendous importance to conservation [74]. The authors of [75] hypothesize that landowners often burn or cut early successional growth even when they do not need the area for production to avoid restrictions in the future. In this setting, PES could help tip the balance toward allowing natural forest regeneration. It would be useful to examine how price differences between cleared and forested land in Brazil have evolved in response to perceptions of enforcement of the FC.
Debates about the relative benefits of disincentive and incentive policies seem to be moving toward policy mixes [75]. It is crucial to understand how actors, who can alter the services provided, perceive the instruments and their interactions. The use of a pragmatic carrot-and-stick approach can promote pro-environmental behaviors in the context of weakly enforced laws [76], help reduce the perception that regulations are unfair, and increase compliance. A clear assessment of the time needed to transition from response to incentives to voluntary legal compliance is important because PES cannot compensate indefinitely for regulatory insufficiencies [2]. Furthermore, PES interventions could exacerbate inequalities, especially where income from land use is highly asymmetrical [77].
Enforcement is important for any regulation to be effective and FC enforcement should apply to restrictions on land use and to the conditionality of PES. Calculated and normative motivations as described by [28] seem to explain compliance intentions in Rio Claro, but lack of trust in the government and perceptions of corruption affect landowner decisions. Regulations like the FC can reduce landowners' sense of self-determination and diminish their intrinsic motivation to protect the environment [78]. Similarly, funds provided by PES can both reduce intrinsic and increase extrinsic motivations for conservation [22]. Moreover, other researchers [79] argue that the focus on rewards and punishments has led to neglecting other ways of supporting smallholders to achieve conservation objectives in the longer term. They suggest focusing on local heterogeneities and capacities and the need to promote trust, altruism and responsibility towards others and future generations.
PES and ELR approaches can be justified to the extent that landowners generate positive externalities through conservation practices that deserve rewards and generate negative externalities through deforestation that justify penalties. However, the approaches differ fundamentally with regard to who pays the costs of conservation. The limited funds for conservation could exacerbate inequities if the incentive approach primarily benefits a few large landowners. It is important to understand how PES influences the cost-effectiveness of achieving desired conservation outcomes in diverse contexts. In Rio Claro, PAF did bring environmental additionality but at a relatively high cost [80]. If law enforcement is stringent and the legal system perceived to be just, enforcement alone will motivate pro-environmental behavior, and PES payments are unnecessary. If these conditions do not hold, PES payments should cover opportunity costs and sanctions should be harsh enough to deter non-compliance. PES substitutes for environmental regulation and should target areas with the highest potential service values with high PES payments and enforced conditionality. Using PES to achieve environmental goals where legal restrictions are enforced is inefficient at best. An alternative would be to use PES to increase equity by PES payments sufficient to offset costs for poor people who could not otherwise comply. It may be fruitful to treat PES as a transitional mechanism to generate behavioral change when the primary objective of payment is to promote compliance with environmental regulations, but PES is likely not a permanent solution to non-compliance. The authors of [4] argued that environmental benefits that arise as a result of compensation or regulations require an on-going flow of payments or compliance checks and, if removed, there is a risk that these benefits will disappear.

Conclusions
Three major conclusions emerged from our study. First, the differences between insiders (farmers, mostly born in the area) and outsiders (non-farmers) are more important than the distinction between participants and non-participants regarding compliance with the Forest Code and willingness to conserve forest or reforest. Second, cost-benefit calculations are not the primary driver of decisions about PES participation. Most PAF participants were outsiders whose pro-forest decisions were largely based on perceived intrinsic values. Insiders, in contrast, were more likely to invoke financial considerations in their decision-making. Third, perceptions of systemic corruption in the enforcement process contributes to respondents' not treating environmental regulations as important in decision-making. PAF's design does not generate the maximum potential benefit from interactions between the incentive and disincentive instruments within the FC because of failure to recognize these three factors. Instead, PAF serves more as compensation for prior pro-environmental behaviors than as an incentive for behavior change.
Perception of the probable stability and longevity of any existing set of regulations, particularly those that limit landowners' ability to make land management decisions, have a critical effect on compliance. Landowners often do not allow forests to regrow and refuse to participate in projects that require reforestation when they expect that reforested areas can never be cleared. Worse yet is when perceptions of in perpetuity loss of the right to clear forest provokes aggressive anticipatory deforestation.
Efforts to save and restore ecosystems require a deep understanding of the efficacy of the tools employed to encourage pro-environmental behaviors. Many instruments are potentially useful and can sometimes be combined beneficially. Nonetheless, limited funds for conservation require decisions about which instruments to employ. These decisions should be based on an analysis of likely interactions between context and the underlying assumptions needed for each type, design, and implementation of instrument to produce the desired results. We also thank all the people that helped us in the field and that were willing to participate in the research, without them this research would not have been possible. And reviewers for their time and comments.

Conflicts of Interest:
The authors declare no conflict of interest.

Detailed Coding Procedure
We created a case file for each interview. A field assistant transcribed the interviews. The researcher paid close attention to the respondent during the interview and noted change in tone, pauses or other signs of hesitation, emotion-laden responses and other non-verbal cues to meaning. Case files were comprised of both the transcriptions and the notes taken by the researcher. The data analysis process began with coding the open response sections of the interviews. A code assigns a summative or evocative phrase that captures the essence of a respondent's comments to a question [81]. The objective of coding is to allow the researcher to identify patterns in an individual's responses and across cases, in this case patterns that may explain why landowners participated or did not participate in PAF and complied or did not comply with Forest Code.
Data analysis involved four steps, three levels of coding ( Figure A2) (1) topical, (2) thematic and (3) analytic, followed by (4) development of an overall conceptual model. The analysis process commenced with identification of specific ideas or themes that emerged from the data in response to each question and topics in the interviews. We used both the transcription and case summaries to capture the information provided by participants, a commonly used procedure to ensure reliability in coding [44,45,82,83]. We then grouped the themes identified into larger conceptual frames that reflect similar themes and relationships among themes that emerged from the interviews [46,47]. These larger frames helped us understand respondents' views about multiple topics. Even if subconsciously, human beings create these larger mental models or scaffolds in many aspects of their lives [48,84,85]. This part of the data analysis process allowed us to identify at least a portion of the participants' mental models as they considered participating in PAF and to understand the commonalities among respondents' views of the roles of the Forest Code and PAF in their lives. Our main goal was not to provide statistical generalizations, but to characterize different perspectives among the landowners. Our final step in data analysis was to develop a model, based on an analysis of the individual codes, the emerging themes, and the interactions between them. We printed all themes and emerging themes and manually arranged them into a model that reflects the learning process. After we checked for agreement of models with the majority of cases (as per [46]), we also highlighted potential findings from outliers.
Topical or descriptive (Level 1) coding assigns a code to the specific comments made by each respondent ( [81]: pp. 83-92). Coding proceeded topic by topic or question by question for each respondent, and codes were developed independently for each comparison group, participants and non-participants in PAF. However, topical codes do not necessarily include only the specific subjects posed in the researcher's questions. It is common for respondents to make comments that are only tangentially related to a question or are seemingly not related at all. These emergent topics are also coded and often provide insights into the respondent's ways of thinking about a topic or the associations a topic brings to his/her mind. Topical codes are highly specific. For example, several participants commented on problems associated with labor. Some commented that their children leave the farm while others commented that there are few people willing to work as agricultural laborers for hire in the region due to low wages. These comments illustrate two aspects of the topical coding process. First, we did not ask specifically about labor. These comments were made in response to questions about other topics and hence they were emergent comments. Second, the comments varied, some were concerned with out-migration and others lack of local labor.
Thematic (Level 2) coding groups the specific comments that emerge in topical coding into broader associated categories or themes ( [81]: pp. 218-223). Thematic coding typically reduces the number of codes substantially because the comments made are grouped by the broad topics included in the interview. Emergent topical codes are also grouped when possible. For example, a thematic code that emerged in this study had to do with the effort involved in meeting bureaucratic requirements. Three topical codes were identified in the first level of coding: (1) the effort associated with land registration, (2) the wasted time spent dealing with fines related to inappropriate land use, and (3) anticipated time and effort required to join PAF. Comments of the first two types were made by all respondents whereas comments about joining PAF were made only by non-participants in PAF. The overarching thematic code for all three of these specific topical comments is bureaucratic efforts for land management. Thematic coding initiates the process of analyzing the data, moving beyond description to understand how respondents organize experiences and concepts into individual mental models.
Analytic (Level 3) coding ( [81]: pp. 223-234) develops specific models of the components included based on both the topical conceptualization of the researcher and emergent patterns that group the specific mental models expressed by respondents. These models include both abstract concepts and proposed explanatory relationships between those concepts. The proposed relationships, typically indicated by flow lines between concepts, are referred to as "propositions" because they are proposed explanations of the relationships among complex concepts. These models focus on specific components of the overall theoretical basis of the research. We developed four of these specific models in this study. For example, we based one model on the socio-economic concepts related to participation in PES described in the literature [86][87][88] and emergent concepts such as "insiders and outsiders" in our study. The overall conceptual model we developed draws upon the individual models developed in Level 3 analytic coding to create a model that we offer as a proposed theory-based explanation of how the participants in PAF and non-participants perceive the connection between payments for ecosystems services and legal compliance. Like most explanatory models, this model contributes to theory by incorporating the mental models of respondents to create a more robust understanding of a complex decision-making process. Table A2 provides first-level participant codes, Table A3 shows first-level non-participant codes, Table A4 displays themes from participants and non-participants in PAF; and Table A5 presents themes decision process, i.e., the process of combining themes together to be included in the models.
services and legal compliance. Like most explanatory models, this model contributes to theory by incorporating the mental models of respondents to create a more robust understanding of a complex decision-making process. Table A2 provides first-level participant codes, Table A3 shows first-level non-participant codes, Table A4 displays themes from participants and non-participants in PAF; and Table A5 presents themes decision process, i.e., the process of combining themes together to be included in the models.   Figure A2. Qualitative methods summary. Summary of quantitative data analysis process. The data acquired were organized into transcripts and field notes, first-level coding created condensed, descriptive phrases that summarized the ideas expressed, level two coding grouped the individual level one codes into themes, and level three coding produced abstract concepts to allow creation of models summarizing the findings. Figure A2. Qualitative methods summary. Summary of quantitative data analysis process. The data acquired were organized into transcripts and field notes, first-level coding created condensed, descriptive phrases that summarized the ideas expressed, level two coding grouped the individual level one codes into themes, and level three coding produced abstract concepts to allow creation of models summarizing the findings.       Reduce cost of required reforestation The legislation requires a forest reserve in order to allow property to be divided into a condominium PAF55 Help to avoid tax fine Had a tax fine because the auditor did not believe the amount of production declared in relation to the size of land PAF54 Be able to produce something in the forest area Aimed to use the forest area, to get benefit from it. PAF23 Wanted to restore and could not do it alone Joined because wanted to restore part of the property and could not do it alone PAF88, PAF45, PAF66, PAF73, PAF81 Became aware of the importance of forest and stopped deforesting The project increased environment awareness and led stopping deforestation PAF89 The change in the property did not impact production The project protects reforested land and therefore did not have an impact on the productivity of the property PAF42, PAF66, PAF10, PAF08 Stopped people from taking wood from the forest The project would prevent taking wood from the forest PAF14  Was informed about PPA requirements by the project When the project was trying to enroll people, the staff informed them that what they were proposing was required in the law

PAF04
Prevents profiting from the property The environment regulation restricts the producer too much, it makes the forest of little use PAF23 The government itself does not do anything The government creates all the laws but does not do anything to improve environment awareness PAF57 Overlap of legislation There are so many overlapping environment regulations that it is hard to keep track PAF19 Park area Property is within the protected area and therefore sees more enforcement PAF78, PAF21 Some rigor in environment regulation enforcement Does not seem to perceive strong rigor, but saw environmental agents or knows people that were fined  Depends on political will CAR seems to be a good instrument but its application will depend on the will of politicians; corruption is instituted PAF40, PAF66, PAF54, PAF23, PAF73, PAF22 Nothing will change CAR will not change anything in terms of land management or enforcement PAF78, PAF86, PAF88, PAF42, PAF66, PAF57 Only bureaucracy Nothing will change, it is just another bureaucracy PAF86, PAF23, PAF57, PAF81 Is increasing real restrictions This is a movement to increase real restrictions and enforcement of environmental regulation PAF68, PAF03, PAF01, PAF72 Is increasing perception of restrictions CAR is making people think they will have to comply PAF77 Made people more aware of environmental legislation The registration process in CAR made people more aware of what was required in the Forest Code PAF56, PAF68, PAF33, PAF50, PAF03, PAF18, PAF88, PAF77, PAF17, PAF66, PAF64, PAF55 Was too much information, already forgot While registering in CAR advisors gave too much information and the interviewee even forgot it PAF89 Did because goes together with everybody Everybody did it, so the landowner did it too PAF18, PAF63 CAR can help legalization CAR has instruments to help to legalize the property PAF31 Table A2. Cont.

Emergent
Necessity drives decision-making process Necessity is the main driver of decision-making process PAF82 Does not use the forest I never walk in the forest PAF82 Television as a source of environmental information Television teaches about environment and environmental regulations PAF82 Lack of sewage treatment in the municipality Sewage treatment has an important impact water and the municipality does not take care of it PAF71 People need to understand for themselves The environmental campaigns have led to people to think and to understand the reasoning for conservation PAF71 Participates in the rural labor union because s/he has employees Main reason to participate in the labor union is that the individual has employees PAF78 Criminal fires set People always make fires in the road, and nobody knows who did it PAF78, PAF03, PAF10 Used network to obtain public benefits Used the network to obtain benefits PAF56 Property is the first place to release animals in Rio de Janeiro Private Reserve of Natural Heritage (RPPN) within property is the first place to release animals in Rio de Janeiro Was harder was to convince the family to participate The family did not want to participate because they believed it was unnecessary PAF04 River is very dirty There is a chicken producer that seems to pollute the river intensely PAF73, PAF14 Birds are coming back with the prohibitions of cages Perceived increase in bird population and believes this is due to the increased prohibition of cages PAF08, PAF34 Expansion of the cities, land division into condominium It is necessary to think how to stop the expansion of the cities, and land division into condominiums. The division is resulting in deforestation PAF08 Absence of government Does not work to report bad actions because the government does not do anything PAF81 Bureaucracy in excess to obtain license There is too much bureaucracy required to obtain a license to make a lake for raising fish. PAF81, PAF72 Future income Aims to get some income from the property in the future N11 Additional income Wants to keep the additional income from the property N20

Condominium
Wants to transform the property in a condominium N20 Property to leave something for the kids Property to leave something for the kids EX92 The bad roads are a problem for production The bad roads are a problem for production EX92       The enforcement of people that practice bird caging has increased and knows there are more birds now EX70 Contributed to people leaving rural areas Believes the rigor in the environmental law contributed to people moving away from rural areas EX70 The law is important because otherwise there would be more deforestation Believes the law has slowed or prevented more deforestation EX70 Will increase real requirements Believes the government will later start asking landowners to reforest N3, N30, N26, EX92

Diagnostic
Believes is a good way to the government to find out about rural areas N7 Only bureaucracy Nothing will change it is just another bureaucracy N21, N10 Depends on political will CAR seems to be a good instrument, but its application will depend on the will of politicians, and corruption is institutional N12, N16

Emergent
Participates in Rio Rural Participates in a World Bank project that requires "an environmental" action N1 Fire is normally a crime Someone has set a fire and the fire entered the property N1 Heard about the RPPN project, and did not participate because of the family Mentioned PES from RPPN and only did not join because property was in family name N3 "Clear" land is necessary It is necessary to clean the land every two years, otherwise you lose the right of using it N22 Sewage from the property goes to the river Property owner reported that the sewage from the property goes to the river N8 Have cows to maintain the pasture Cows avoid forest to grow back, and help to maintain the pasture EX70 Corruption with bureaucracy to legalize water extraction The producer gave up of a project because it had Corruption with bureaucracy to legalize water extraction EX92 Sustainability 2020, 12, 8138 40 of 51  Legal restrictions Property in a park area 2 Use property for leisure because cannot alter it 2 Too many restrictions on use of forest land 6 The landowner has more responsibility over forest land than over pasture 6 Not allowed to alter anything in forested areas 12 Law prevents owner from profiting from the property 13 Park area 14 Infrastructure restrictions Difficult access 2 Tried activities that did not work 2 Legal restrictions are not enforced in areas with bad roads 14 Social motivations for pro-environmental action Obligation as a citizen 3 Maintain forest for benefit of future generations 3 Societal collaboration 5 Everybody around was participating, so joined too 9 Project people were nice 9 The project sounded important 9 Recognition 9 Benefit for others 9 Worked in the project 10 Any law must be respected 13 Did because it brings together everybody 16 Table A4. Cont.

First-Level Code Topic
Laws promote pro-environmental action Maintain forest because is cutting is not allowed 3 Law is required 3 Looked for help to restore land in order to fulfill environmental regulation 8 Joined PAF because of legislation 9 Direct use of the forest/forest utilities Water Wood source 3 Firewood 5 Reduce the problems with fire 4 Erosion control 5 Clear air 5 Desire to restore areas 9 Environmental awareness and biospheric values Environmental awareness 3 Maintain forest because of animals 3 Believes forest is important for conservation 5 Would have joined without the money 9 Increase local environmental awareness 9 Forest conservation 1 Never worried much because s/he believes conservation is important 12 Conservation is important 12 Normative motivations: Agrees with the law or believes compliance is the right thing to do 13 It is necessary to understand the reasoning behind environmental laws 15 Intrinsic motivation for pro-environmental action Religion 3 Trees are beautiful 4 Sponsorship for increases in protection 9 Increases local environmental awareness 9 Likes forest 9 Forest conservation 1 Forest provides peace and beauty, stress relief 5 Forest provides inspiration for work 5 People need to understand for themselves 17 Table A4. Cont.

First-Level Code Topic
Recover previous damage to the land Recovered a degraded area 4 Relieved guilt for past deforestation 9 Positive environmental outcomes Land was abandoned areas that were not being used 4 Changed agricultural production to tourism 4 Wanted to restore and could not do it alone 9 Perception of positive outcomes 9 Became aware of the importance of forest and stopped deforesting 10 Stopped harvesting wood from the forest because of PAF 10 Increased water availability 10 Reduced erosion 10 Reduced the problems with fire 10 People would deforest if it did not exist 13 Changed behavior because increase the perception of enforcement 14 Illegality/impunity and lack of government as example Problems with hunters or palm extractors 6 Avoid criminal fire and hunting 9 Avoid land invasion 9 People believe that there will not be any sanction 15