Protecting the Environment

Abstract

The role of extrinsic versus intrinsic motivation in environmental decisions remain unresolved. We exploit data from a lab-in-the-field experiment to analyze the role of extrinsic and intrinsic incentives in shaping individual demand for a payment for environmental services (PES) program in São Paulo, Brazil. The lab-in-the-field experiment is a theoretical incentive program that offers annual payments to landholders in vulnerable watersheds for either conserving and/or restoring trees surrounding springs on their land to preserve and improve local water quality. Our findings suggest that, in contrast with predictions from rational choice theory, individuals’ responses to incentives are not monotonic. Landholders who took part in our lab-in-the-field experiment were randomly assigned to four offer levels and asked a double-bounded contingent valuation question to elicit a willingness to accept value. Landholders were less likely to accept the higher offers compared to the lowest offers. Given that the rational choice model fails to fully account for the role of incentives in shaping demand for PES, we then look at the interaction of the randomized incentive offers and individuals’ initial intrinsic motivations. We find that, while high monetary incentives crowd in demand of progovernment landholders, they crowd out demand of proenvironment and prosocial landholders. Overall, we find much evidence of heterogeneous responses.

Keywords

intrinsic motivation extrinsic incentives payment for environmental services field experiment

In neoclassical economics, rational choices based on the benefits and costs of alternative actions explain behavior. Carrots (rewards) and sticks (sanctions) can alter behavior by affecting these trade-offs. Offering monetary rewards to perform a certain action will increase the benefits relative to its costs and thus make the action more likely. In the context of environmental issues, policies have recently been designed to nudge agents to internalize the societal cost of degradation and adopt environmentally desirable behavior through subsidy programs, commonly known as payment for environmental services (PES). The underlying assumption of PES programs is that monetary rewards provide extrinsic motivation to conserve the environment by influencing the cost–benefit analysis that determines behavior.

However, the pure rational choice paradigm has increasingly been called into question by behavioral economics, which allows for alternative motivations for individual action. In particular, individuals’ actions may respond to intrinsic motivations independently of external (or extrinsic) incentives (Deci and Ryan 1985). In the realm of environmental policies, individuals may have varying levels of intrinsic motivation to protect the environment. Critically, the effect of extrinsic monetary incentives can depend on individuals’ intrinsic motivations. For example, a number of studies in the economics and psychology literature suggest a crowding-out effect whereby monetary incentives may actually reduce intrinsic motivation to behave a certain way (Frey 1992; Benabou and Tirole 2003). Frey (1992) shows that, as individuals perceive a policy as an external instrument to control or regulate intrinsic motivation, it may diminish their self-determination and, therefore, welfare associated with ecological behavior. On the other hand, intrinsic and extrinsic motivations may act as complements, and offering higher levels of subsidy may leverage preinstalled conservation behavior (crowding-in). Agents may interpret the subsidy payments as recognition of “good” behavior, raising their self-evaluation and welfare derived from conservation (Frey 1992).

Another mechanism through which monetary incentives may undermine conservation is framing. Evidence from the lab suggests that external incentives may shift an individual’s decision from a social frame to a monetary frame (Heyman and Ariely 2004); once the framing on the activity changes, the level of the monetary incentive matters above and beyond the intrinsic motivations, suggesting that levels of conservation may in fact decrease after the phasing out of a PES program (Gneezy and Rustichini 2000; Lepper and Greene 1978).

The literature thus suggests that the impact of a monetary incentive on conservation behavior is a priori ambiguous and may be conditioned by preexisting levels of intrinsic motivations. Monetary incentives to increase conservation of private land may not have the effect expected under rational choice accounts.

We contribute to this literature by shedding light on the interplay of extrinsic and intrinsic motivations in the context of a payments for environmental services (PES) program. Using a lab-in-the-field experiment in Brazil, we vary extrinsic incentives by randomizing subsidy offers to landholders in vulnerable watersheds to conserve and/or restore trees surrounding springs on their land. We capture intrinsic motivations through a detailed survey instrument on individual preferences regarding the environment and society. In an earlier working paper (see De Martino et al. 2015), we used factor analysis to construct latent indices which allowed us to develop a taxonomy of intrinsic motivations to conserve. We use the constructed indices of latent motivations in this article.

Our findings suggest that, contrary to rational-actor expectations, demand for conservation and restoration programs does not increase with higher subsidies. Landholders who took part in our experiment were randomly assigned to four offer levels and asked a double-bounded contingent valuation question to elicit a willingness to accept (WTA) value. We find that landholders are less likely to accept the payment if they were randomly assigned to the high offer treatments than if they were randomly assigned to the low offer treatments.

Our analysis of the interaction of extrinsic and intrinsic motivations suggests that monetary incentives can undermine intrinsic motivation in certain cases, making individuals less likely to accept a PES program. Specifically, we found that while high monetary incentives elicit higher demand for conservation behavior among prolegal and progovernment landholders, they crowd-out demand of proenvironment and prosocial landholders. This crowding-out effect explains why progressively higher monetary offers did not increase landowners’ likelihood to accept. From a policy perspective, this result suggests that program administrators should, to the extent possible, take individual motivation into consideration when designing PES programs. This is particularly important for ensuring that such policies have additionality, since we find those who are compliant to existing conservation laws are those accepting the payments.

Our study builds on a new strand of lab-in-the-field experiments that seek to analyze the interaction of motivation and incentives in relation to environment protection (Vollan 2008; Kerr, Vardhan, and Jindal 2012). Using trust and common pool resource games in South Africa and Namibia, Vollan (2008) finds that a monetary incentive’s crowding-out effect on prosocial behavior is dependent on three conditions: strong existing norms of trust and reciprocity; a low degree of self-determination in the individual; and perceptions of the external regulation as controlling rather than supportive. Kerr, Vardhan, and Jindal (2012) run lab-in-the-field experiments to isolate the impact of incentives on weekend volunteerism in Mexico and Tanzania. In Mexico, group payments made through village authorities yield lower participation where people distrust leaders. Payments did not undermine participation in Tanzania, but they reduced satisfaction from the task.

In the PES literature, many studies have analyzed land and demographic characteristics of participants in the programs (Sierra and Russman 2006, Rios and Pagiola 2009; Pagiola, Rios, and Arcenas 2010; Arriagada 2009; Pfaff, Robalin, and Sanchez-Azefeifa 2008; Robalino, Pfaff, Sanchez-Azofefifa, Alpizar, Leon, and Rodriguez 2008), yet the role of intrinsic motivation in the context of demand for PES has not been thoroughly studied. One exception is a recent paper by Zanella, Schleyer, and Speelman (2014) who use ex post data to show that information, environmental concern, and participation positively affect take up of PES programs in Brazil. We are not aware of any studies that analyze the interaction of extrinsic and intrinsic incentives in the context of demand for PES programs.

The remainder of this article is as follows: we introduce payments for environmental services and its relevance in Brazil; we then describe the experimental design conducted in Brazil, followed by descriptive statistics and our empirical strategy; and we then present the results and conclude.

A Brief Introduction to Payments for Environmental Services Programs and the Brazilian Context

Payments for Environmental Services (PES) seek to correct externalities by using monetary subsidies to incentivize individuals to preserve or restore public goods (such as forests) or common goods (such as water). A variety of market failures impede individuals from allocating socially optimal levels of conservation effort. Social benefits of conservation may disproportionally accrue to certain groups. For instance, land-degrading activities upstream may have large adverse effects on those living downstream, with little direct impact on upstream users. While tree planting is socially optimal, upstream individuals will not internalize the benefits and, therefore, risk not investing in conservation activities. In this setting, a direct subsidy to upstream landholders will reduce the wedge between private and public marginal benefits. This is the idea behind PES programs. Typically, landowners will be offered a payment conditional on changing or maintaining environmentally desirable land and resource management practices.

Understanding the benefits and pitfalls of this approach is crucial, as the number of PES programs has steadily increased over the last decade, especially in Latin America. In Brazil, several states have launched PES pilots,¹ mainly with a focus on preventing deforestation and its side effects, such as erosion. While the monetary incentives are expected to help landowners overcome barriers to invest in inputs needed to transition from their current land use, it is unclear that they would be sufficient to trigger behavioral change in the absence of other incentives, information campaigns, and additional trainings.

The present study adds to the current state of evidence on determinants of demand in the context of a government-led PES program in the state of São Paulo (SP), Brazil. Around 40 percent of the state is at risk of erosion, which is a major contributor to the state’s worst water crisis to date affecting 20 million people. Against this background, the state policy on climate change was adopted in 2009 to promote large-scale restoration. The program we study (Mina d’ Água) was one of the policy instruments piloted under this initiative. It is the first PES project implemented directly by the SP state government and was developed in partnership with twenty-one municipalities.

Analytical Framework

Determinants of Demand

From the landholder’s perspective, there are many potential drivers for investment in a PES program. We group these motivations in two categories: extrinsic motivations and intrinsic motivations. According to the rational choice model, the landholder responds only to extrinsic incentives: he will conserve if the monetary benefits exceed the costs. Conservation beyond the expected level of a strict profit-maximizing landholder suggests the individual may hold intrinsic motivations that lead him or her to conserve some portion of land irrespective of costs and benefits.

We consider three extrinsic drivers of demand for investment in a conservation or restoration program, which are not mutually exclusive, along with their expected effect on behavior:

Level of monetary incentive (the “carrot”): the higher the incentive offered, the more likely a landholder will accept the payment and conserve his property.

Opportunity cost of land: the higher the opportunity cost embodied in alternative income generating activities, the less likely a landholder will conserve his land.

Prolegal motivation (the “stick”): the landholder will conserve all land required under the conservation law in order to avoid fines. Under the Brazil Forest Code, a landholder is responsible for the full preservation of “permanent forest preservation areas” (APP), which includes areas adjacent to rivers and ponds, steep hillsides, and springs.²

We consider four main intrinsic drivers of demand for investment in a PES program:

Proenvironment motivation: the landholder values the existence and importance of the environment.

Prosocial motivation: the landholder values protection of the environment for his community members and society as a whole, current, and/or future (Eckel and Grossman 1996; Meier 2007).

Progovernment motivation: the landholder believes the government is responsible for paying to protect water sources on private, and separately, public property.

Social norms: the landholder values the opinions and perceptions of their neighbor on conservation matters.

As discussed, we are especially interested in the interaction of extrinsic incentives on intrinsic motivations. While the expected effect on conservation behavior of each driver in isolation is straightforward, the effect of the interaction between extrinsic and intrinsic motivations is a priori ambiguous given the possibility of the monetary incentive supporting or undermining the landholder’s intrinsic motivation.

Data and Experimental Setup

Data Collection

We ran a lab-in-the-field experiment within the baseline survey for a pilot PES program in the state of SP. The baseline survey was carried out in 2013 in two municipalities, Ibiúna and Guapiara, both located in the southeast region of SP state. Each municipality established priority areas for conservation and restoration, which typically consist of water basins exploited by local water companies that distribute domestic water to the surrounding communities. In our study area, we list all landowners in the catchment of these priority water basins and their adjacent areas. Within that group, we sampled all agricultural landowners with at least one spring on their parcel. In total, we surveyed 350 landowners. The survey captures patterns of land and water use, agricultural production, and income. We also elicited WTA for PES programs using randomly assigned offer levels and conducted an exhaustive survey of landholders’ attitudes toward environmental issues. The descriptives section discusses the characteristics of the surveyed landholders.

Data Generation for Extrinsic and Intrinsic Motivations

We now detail the data generation strategy we employ to measure extrinsic and intrinsic motivations in our survey.

Eliciting responses to extrinsic incentives

To understand how landholders respond to extrinsic motivations, we introduce a monetary incentive and present landowners with two hypothetical PES projects, one for conservation and one for restoration. As environmental services are difficult to monetize, contingent valuation models such as our lab-in-the-field experiment are used to capture opportunity costs of the landholder. If the landholder is motivated only by extrinsic motivations, the expectation is that he would not accept a payment for conservation that is less than his opportunity cost and would not accept a payment for restoration that is less than the sum of his opportunity and restoration costs. Any finding contradicting these assumptions would prompt us to consider alternative motivations for the landholder’s behavior.

We use a double-bounded dichotomous contingent valuation model to elicit individual willingness-to-accept compensation for a PES program. If the landholder accepted the first offer, then no second offer was made. If the first offer was turned down, a second and final offer was made, which was 50 percent greater than the first bid (table 1).

Table 1.

Offers per Hectare (in R$).

	Conservation		Restoration
Payment	Bid 1	Bid 2	Bid 1	Bid 2
Level 1	150	225	2,000	3,000
Level 2	200	300	3,000	4,500
Level 3	250	375	4,000	6,000
Level 4	300	450	5,000	7,500

The incentive offer was randomized across four different levels among the surveyed landholders. The randomization was stratified across the two municipalities and in the priority and adjacent areas. Please refer to the descriptives section for information on balance tests.

The conservation project offered a yearly payment per hectare of conserved APP in compliance with the law. The payments were proportional to the size of the forest cover in APP areas. The first offer ranged from R$150 to R$300 per year per hectare. The upper bound of R$300 per year per hectare is equivalent to the average annual return of keeping livestock on degraded pasture, according to informal discussions with SP state secretariat of environment.

The restoration project offered a one-time payment to restore any uncovered degraded APP area. As opposed to the annual conservation payments, this was a one-off, nonrenewable payment to the household. Given the large costs of restoration, payments were higher than in the conservation program. Similar to the conservation offer, payments were proportional to the size of the uncovered APP areas. The first offer ranged from R$2,000 to R$5,000 per hectare and was determined using the costs borne to the government for similar restoration activities (R$4,000 per hectare).

As mentioned, the lab-in-the-field experiment was carried out within a baseline survey for a pilot PES program in the state of SP. We use data from the survey on land and water use, agricultural production, and income as proxies for opportunity costs in the regression analysis.

To measure legal compliance, we consider landholders to be “prolegal” if they conserve land where it is required by law in order to avoid fines. To verify if the landholder is in full compliance with the law, we documented existing levels of conservation on each hectare of the landholder’s property when the survey was administered, including hectares under APP and outside of APP. We devise an indicator taking the value of 1 if the landholder is currently in full compliance with the APP law³ and 0 if not.

2. Eliciting intrinsic motivations: survey responses

To elicit intrinsic motivations, we use data from the detailed survey questions on perceptions of conservation and society. The survey responses allow us to identify landholders as motivated by proenvironment, prosocial, progovernment, or social norms attitudes as described in the analytical framework. Using factor analysis, we construct indices to reduce the dimensionality of the proxies for motivations in an earlier working paper (De Martino et al. 2015). See the Appendix for more details on the construction of the indices. We then test if stated preferences as captured in the indices determine revealed preferences, as measured by existing levels of conservation at the time of the survey. This step is crucial to establish that the indices provide a valid measure of latent intrinsic motivations. Our results, reported in an earlier working paper (De Martino et al. 2015), show that stated preferences are in fact consistent with reveled preferences once we control for opportunity costs, and we refer readers to this working paper for further discussion.

Descriptive Statistics

Extrinsic Drivers: Monetary Incentives

The percentage of households who accepted the first and second bid, respectively, is reported in table 2. Independent of the treatment level, half of the sample accepted the first bid for the conservation program, and one-third of the sample accepted the first bid for the restoration program. We observe monotonic responses from those who rejected the first offer but then accepted the second, higher, offer. Twenty to thirty percent accepted the second and final bid for the conservation program whereas only 11 to 17 percent of this sample accepted the second and final bid for the restoration program (table 2).

Table 2.

Bids Accepted.

	Percentage accepted first bid (%)	Percentage accepted second bid of those who rejected first bid (%)
Conservation offer (in R$)
150; 225	50	21
200; 300	45	31
250; 375	45	26
300; 450	48	18
Restoration offer (in R$)
2,000; 3,000	37	11
3,000; 4,500	33	13
4,000; 6,000	33	12
5,000; 7,500	36	17

Balance tests on the four randomized treatments were carried out on observable characteristics. Small yet significant differences across treatments arise from the age of the head of household, credit use, area of property, amount of erosion on property and conservation within APP zones (table 3)⁴, and are controlled for in all estimation models.

Table 3.

Property Characteristics.

Property characteristics	Total	Guapiara	Ibiuna	Difference
Size of the property (Ha)	12.2	12.2	12.3
Household lives on the property (%)	81	91	72	***
Has no area designated to agriculture (%)	20	7	30	***
Agriculture is used for own consumption (%)	75	81	69	**
Agriculture is only used for own consumption (%)	32	29	34
Area of property covered with forest (%)	43	31	53
Has erosion on part of the property (%)	11	17	6	***
Has livestock (%)	61	74	50	***
Number of springs	1.9	1.7	2.1	**
Observations	350	188	162

Note: “Difference” denotes t-tests of means (Guapiara vs. Ibiuna).

**p < .05. ***p < .01.

Take-up of the conservation bid was 60 percent of the sample, whereas take-up of the restoration bid was 43 percent of the sample.

Extrinsic Drivers: Opportunity Costs and Legal Compliance

Our average respondent is fifty-five years of age and holds a little over twelve hectares of land—a small-scale producer by Brazilian standards (table 3). About 15 percent of the sample uses credit. For those who do not use credit, 76 percent of the sample stated it was because they did not need credit. Thus, we are analyzing a sample that is not credit constrained or not undertaking investments. Thirty percent of the sample experienced a bad agricultural yield in the previous year. We find that Guapiara experiences higher levels of erosion than Ibiuna. Although over 80 percent of our sample uses at least part of their land for agriculture (Ibiuna 70 percent and Guapiara 93 percent), properties in our sample have high forest cover (Ibiuna 53 percent and Guapiara 31 percent). At the time of the survey, 52 percent of landholders were in full compliance with the APP law (Ibiuna 56 percent and Guapiara 48 percent). Ibiuna landholders mostly engage in horticulture, while Guapiara producers mainly produce annual crops and use significantly more livestock. In our estimation, we account for these differences using municipality fixed effects.

Intrinsic Drivers: Motivations

Table 4 presents the breakdown of the sample by intrinsic drivers: proenvironment, prosocial, progovernment, and social norms. The average respondent answered positive to 52 percent of the proenvironment questions, whereas the average respondent answered positive to only 8 percent of the prosocial questions. When looking at the descriptives of those who accepted the restoration bid, the average respondent answered positive to 56 percent of the proenvironment questions and to only 4 percent of the prosocial questions.

Table 4.

Indices.

Intrinsic motivation index	Number of questions in additive index (standardized in regressions)	Mean number of “Yes” replies (n = 351)	Standard deviation	Mean number of “Yes” replies for those who accepted conservation bid (n = 210)	Mean number of “Yes” replies for those who accepted restoration bid (n = 151)
Proenvironment	7	4.72	2.35	4.79	5.07
Prosocial	1	0.08	0.27	0.06	0.04
Progovernment	3	1.84	1.22	1.98	2.12
Social norms	2	0.87	0.38	0.95	0.96
Access to information	3	1.04	.80	1.06	1.07

The average landholder responded positively to 61 percent of the progovernment questions and 87 percent of the questions on social norms influence. On average, the landholder responded positively to 52 percent of the questions on knowledge of PES and the MdA program in particular. Those who accepted the restoration bid responded positively to 71 percent of the progovernment questions, 95 percent of the questions on social norms influence, and 54 percent of the questions on access to information.

The descriptives are consistent with the regression results later: landholders with prosocial motivations are categorically less likely to accept the monetary incentive, while proenvironment landholders are less likely to accept the monetary incentive only at high offer levels. Landholders with progovernment motivations, social norms influence, and access to information, on the other hand, are categorically more likely to accept the monetary incentive.

Experimental Results

This section presents our regression results. First, we determine the average WTA value of the extrinsic (monetary) incentive. Second, in order to gain a richer understanding of the effect of opportunity costs, intrinsic motivations, and the interaction of intrinsic motivations and the extrinsic randomized incentive offer, we use a probit model to determine the main drivers of demand for investment in PES.

Maximum Likelihood Estimation (MLE) of WTA Model

The average WTA value is analyzed using a double-bounded dichotomous model. We construct a likelihood function and use MLE to obtain the coefficients. We estimate the average monetary amount required to comply with the requirements of the conservation and, separately, the restoration program as observed in the sample independent of the randomized offer level. We represent the first bid amount by t¹ and the second bid by t². Thus, the landholder will belong to one of the following three categories:

The landholder answers yes to the first bid: the probability is Pr(WTA ≤ t¹).

The landholder answers no to the first but yes to the second bid: the probability is Pr(t¹ < WTA ≤ t²).

The landholder answers no to the first bid and the second bid: the probability is Pr(t² < WTA < ∞).

Employing a WTA model, we calculate the average using a double-bounded model (Lopez-Feldman 2012) with dichotomous choices under the assumption that there is a single valuation function behind both answers. $y_{i}^{1}$ and $y_{i}^{2}$ are the dichotomous variables that capture the response to the first and second questions.

To obtain estimates for the average WTA (represented as β and σ), we construct a likelihood function and use MLE programming to obtain the coefficients⁵:

\begin{array}{l} \sum_{i = 1}^{N} d_{i}^{n n} \ln (Φ ({z'}_{i} \frac{β}{σ} - \frac{t^{2}}{σ})) + d_{i}^{s} \ln (1 - Φ ({z'}_{i} \frac{β}{σ} - \frac{t^{1}}{σ})) \\ + d_{i}^{n s} \ln (Φ ({z'}_{i} \frac{β}{σ} - \frac{t^{1}}{σ}) - Φ ({z'}_{i} \frac{β}{σ} - \frac{t^{2}}{σ})) \end{array}

where $d_{i}^{n n}$ , $d_{i}^{s}$ , and $d_{i}^{n s}$ are variables that take the value of one or zero depending on the response to $y_{i}^{1}$ and $y_{i}^{2}$ . First we estimate the WTA without covariates and then we control for opportunity costs as defined by the Demographics′_i and LandCharacteristics′_i vectors (see table 5).

Table 5.

Willingness to Accept (WTA) Estimates.

Program	Direct WTA	WTA with covariates
Conservation	R$244	R$233
Restoration	R$7,030	R$6,853

Result 1: Average WTA value is low for the conservation program but high for the restoration program.

The average WTA for the conservation program lies between the first and second offer level, whereas the average WTA for the restoration program is much higher and lies between the third and fourth (highest) offer level. The higher average WTA for the restoration program is indicative of the program’s requirement to reallocate time, labor, and money to reforest the property and affect potential farming land. By contrast, the conservation program requires less effort.

The complementary model below further explores the characteristics of households who accepted an incentive for the conservation, and separately, the restoration program to understand the determinants of a relatively low WTA, and highlights results that corroborate the main finding that there exists large heterogeneity in response to incentive payments.

Interacting Intrinsic Motivations and Extrinsic Incentives

We estimate landholder-level probit regressions where the dependent variable Accept _i is a dichotomous variable equal to 1 if landholder i accepted the theoretical incentive to participate in the conservation program and, separately, the restoration program, and zero otherwise. We assume that the probability of acceptance is a function of opportunity costs, indices of intrinsic motivations, the incentive offer, and the interaction of indices and incentives, or:

\begin{array}{l} {Accept}_{i} = α + β_{1} {Demographics′}_{i} + β_{2} {Land characteristics′}_{i} \\ + β_{3} \Pr {olegal}_{i} + β_{4} {Proenv}_{i} + β_{5} {Prosocial}_{i} \\ + β_{6} {Progovt}_{i} + β_{7} {Social norms}_{i} + β_{8} {Acess info}_{i} \\ + β_{9} i {. incentive}_{i} + β_{10} i {. incentive}_{i} \times indices + ∊_{i} \end{array}

Demographics _i is a vector of covariates including household size, education, age, gender of the household head, income, and credit access and use. Land characteristics _i is a vector of covariates conventionally used in analyses of PES demand and are comprised of property size in hectares, number of people working on the land, possession of legal documents for ownership or renting of the property, soil characteristics (sand, clay, mix, and red soil), steepness of land, evidence of erosion on property, number of springs on property, if the landholder experienced a bad yield in the previous year, (log) profits from agriculture, and a binary variable indicating whether the property is used for agriculture.

Prolegal _i is a dichotomous variable equal to 1 if the landholder is in full compliance of the APP law at the time of the survey and 0 otherwise.

To go beyond the strict homo economicus landholder, we are interested in empirically documenting the role of intrinsic motivations in determining demand for PES. As outlined earlier, we consider four main intrinsic drivers of investment in the program: Proenv _i , Prosocial _i , Prosovt _i , and Social norms _i .

We also consider access(to)information _i an important and final covariate to determine if those who have access to and understand information on PES and the MdA program of interest have higher demand for PES programs.

i.incentive is a vector of dummy variables for the four randomized levels of payments, using the lowest offer as the base. The variable i.incentive × i.indices captures the effect of interacting the randomized monetary incentive with the various motivations. The motivations of interest in the heterogeneity analysis are prolegal, proenvironment, prosocial, progovernment, social norms, and access to information.

Finally, ∊ _i is an independently distributed error term assumed to be normally distributed with zero mean and a constant variance.

Marginal effects are computed for continuous and dichotomous explanatory variables. Standard errors are clustered at the municipality level.⁶ Estimation results are reported in table 6. As we are measuring elasticities of demand, the interpretation of coefficients is in percentage change, except when analyzing the standardized indices, which are interpreted as increases in standard deviations above the mean.

Table 6.

Estimation Results.

	(I)	(II)	(III)	(IV)	(V)	(VI)	(VII)
Accept conservation program incentive	Means	Offer level	Opp costs	Offer level + Opp costs	Opportunity costs + indices	Offer level, opportunity costs + indices	Fully interacted model
Size of the HH	3.915		0.008	0.008	0.005	0.006	0
	0.145		0.013	0.012	0.007	0.006	0.004
Gender HH head	0.853		−0.163	−0.162	−0.111	−0.107	−0.12
	0.019		0.146	0.143	0.175	0.169	0.168
Age of HH head	55.425		−0.002	−0.002	−0.001	−0.001	−0.001
	0.753		0.003	0.003	0.003	0.003	0.002
Education of HH head	2.469		0.019	0.019	0.026***	0.025***	0.025***
	0.063		0.016	0.016	0.004	0.005	0.008
Income from agriculture—number of minimum wage	1.326		0.008	0.008	0.002	0.002	0.001
	0.087		0.015	0.016	0.009	0.01	0.01
HH uses credit	0.147		−0.022	−0.015	−0.045	−0.037	−0.047
	0.019		0.092	0.088	0.091	0.088	0.102
ln(profit)	12.126		−0.097	−0.108	−0.055	−0.067	−0.045
	0.008		0.12	0.112	0.073	0.063	0.049
Ibiuna	0.54	0.080***	0.057	0.058	0.057	0.057	0.057
	0.027	0	0.055	0.055	0.055	0.055	0.049
Experienced bad agricultural yield in previous year	0.29		0.111	0.111	0.084	0.083	0.091
	0.025		0.162	0.165	0.157	0.161	0.135
Total area of property (Ha)	12.506		0.003*	0.003*	0.003**	0.003**	0.003**
	0.713		0.002	0.002	0.001	0.001	0.001
Number of members that participate in agriculture activity	1.962		−0.038***	−0.039**	−0.028*	−0.029*	−0.023
Number of members that participate in agriculture activity	0.117		0.014	0.016	0.016	0.017	0.022
Has legal documents for property	0.883		0.050***	0.052***	0.039	0.043	0.044
	0.017		0.011	0.008	0.042	0.037	0.027
Sand	0.267		0.059	0.06	0.093	0.094	0.092
	0.024		0.131	0.138	0.101	0.109	0.103
Clay	0.258		0.007	0.009	0.054***	0.059***	0.042***
	0.024		0.016	0.02	0.014	0.009	0.007
Sand-clay	0.226		0.014	0.011	0.031	0.03	0.014
	0.023		0.041	0.039	0.024	0.021	0.028
Terra Roxa	0.437		0.054***	0.058***	0.107***	0.112***	0.106***
	0.027		0.006	0.003	0.02	0.016	0.015
Proportion of total steepness over total property	0.754		−0.024	−0.019	−0.027	−0.02	−0.033
	0.015		0.061	0.054	0.11	0.099	0.08
# of properties eroded	0.15		0.013	0.007	0.056***	0.050***	0.052***
	0.027		0.015	0.014	0.005	0.004	0.002
Land used for agriculture	0.804		−0.146***	−0.144***	−0.131***	−0.128***	−0.122***
	0.022		0.047	0.048	0.049	0.049	0.043
Chance of deforesting around APP spring	0.044		0.111***	0.102***	0.172*	0.168*	0.145
	0.011		0.028	0.024	0.104	0.098	0.102
% of non-APP land conserved	0.381		−0.127***	−0.131***	−0.100***	−0.106***	−0.122**
	0.018		0.013	0.014	0.026	0.024	0.051
Prolegal	0.525		0.078**	0.078**	0.110***	0.108***	0.104***
	0.027		0.033	0.032	0.034	0.032	0.021
Access to information on PES (std)	0.009				0.018	0.016	0.050***
	0.054				0.042	0.04	0.016
Social norms (std)	−0.01				0.122***	0.123***	0.088***
	0.054				0.011	0.008	0.019
Proenvironment (std)	0.003				−0.016***	−0.014***	0.001
	0.054				0.004	0.004	0.052
Prosocial (std)	0.008				−0.016***	−0.019**	−0.013
	0.055				0.006	0.009	0.01
Progovernment (std)	−0.009				0.071***	0.069***	0.030***
	0.054				0.015	0.014	0.005
High offer treatment	0.49			−0.048***		−0.053**	−0.061
	0.027			0.018		0.027	0.042
Second payment offer level		0.006
		0.095
Third payment offer level		−0.02
		0.015
Fourth payment offer level		−0.04
		0.153
Interaction high offer and proenvironment (std)							−0.04
							0.108
Interaction high offer and prosocial (std)							−0.039
							0.062
Interaction high offer and progovernment (std)							0.073***
							0.01
Interaction high offer and prolegal							0.019***
							0.006
Interaction high offer and access to information (std)							−0.068
							0.052
Interaction high offer and Social Norms influence (std)							0.088
							0.083
Observations	348	351	348	348	348	348	348

Note: APP = permanent forest preservation areas; HH = household.

*p < .10. **p < .05. ***p < .01.

Result 2: Demand for conservation and restoration is nonmonotonic with respect to extrinsic monetary incentives.

If landholders are profit-maximizing agents, the prediction is that a higher subsidy offer will yield higher demand for the PES program. Our first result contradicts this monotonicity assumption (tables 6 and 7). We pool the four randomized treatment offer levels into two groups, high offer and low offer. Landholders are 5 percent less likely to accept the conservation offer if assigned to the high payment treatment compared to those who received the low payment treatment. This is consistent across all specifications. For the restoration program, the likelihood to accept a payment also decreases with a high monetary offer, and the magnitude is even greater: landholders are 26 percent less likely to accept the payment if assigned to the high offer versus the low offer.

Table 7.

Estimation Results.

Accept Restoration program incentive	(I)	(II)	(III)	(IV)	(V)	(VI)	(VII)
Size of the HH	3.915		0.003	0.004	−0.001	−0.001	−0.005
	0.145		0.009	0.01	0.014	0.014	0.012
Gender HH head	0.853		−0.171*	−0.170*	−0.118	−0.117	−0.116
	0.019		0.09	0.09	0.117	0.117	0.098
Age of HH head	55.425		0.001	0.001	0.001	0.001	0.001
	0.753		0.003	0.003	0.002	0.002	0.003
Education of HH head	2.469		0.042***	0.042***	0.042***	0.042***	0.038***
	0.063		0.015	0.015	0.003	0.003	0.004
Income from agriculture—number of minimum wage	1.326		−0.002	−0.002	−0.009***	−0.010***	−0.012***
	0.087		0.001	0.002	0.001	0	0
HH uses credit	0.147		−0.004	−0.001	0.003	0.005	−0.006
	0.019		0.125	0.119	0.114	0.109	0.123
ln(profit)	12.126		−0.199*	−0.204*	−0.184***	−0.188***	−0.187**
	0.008		0.107	0.11	0.07	0.072	0.083
Ibiuna	0.54	0.211***	0.103***	0.103***	0.110***	0.110***	0.122***
	0.027	0.003	0.021	0.021	0.012	0.013	0.001
Experienced bad agricultural yield in previous year	0.29		0.119	0.119	0.094**	0.094**	0.100***
	0.025		0.078	0.079	0.044	0.046	0.03
Total area of property (Ha)	12.506		0.002***	0.002***	0.002**	0.002**	0.002*
	0.713		0.001	0.001	0.001	0.001	0.001
Number of members that participate in agriculture activity	1.962		−0.035	−0.036	−0.018	−0.019	−0.014
	0.117		0.026	0.027	0.027	0.027	0.023
Has legal documents for property	0.883		−0.112***	−0.111***	−0.116***	−0.115***	−0.121***
	0.017		0.003	0.003	0.028	0.028	0.022
Sand	0.267		0.094	0.095	0.120***	0.121***	0.110***
	0.024		0.074	0.076	0.017	0.02	0.002
Clay	0.258		0.085	0.087	0.115***	0.116***	0.099***
	0.024		0.055	0.056	0.025	0.025	0.004
Sand-clay	0.226		0.015	0.014	0.027	0.027	0.022
	0.023		0.012	0.012	0.025	0.025	0.042
Terra Roxa	0.437		−0.063	−0.061	−0.019	−0.017	−0.001
	0.027		0.064	0.064	0.037	0.036	0.015
Proportion of total steepness over total property	0.754		0.029	0.03	0.041	0.043	0.033
	0.015		0.061	0.06	0.11	0.11	0.085
# of properties eroded	0.15		−0.046***	−0.048***	−0.020***	−0.022***	−0.016**
	0.027		0.002	0.005	0.002	0.005	0.006
Land used for agriculture	0.804		−0.095***	−0.095***	−0.089***	−0.089***	−0.100***
	0.022		0.012	0.011	0.008	0.007	0.011
Chance of deforesting around APP spring	0.044		−0.133	−0.136	−0.11	−0.111	−0.132
	0.011		0.141	0.146	0.112	0.114	0.118
% of non-APP land conserved	0.381		−0.119***	−0.122***	−0.077**	−0.079**	−0.079***
	0.018		0.027	0.029	0.038	0.04	0.026
Prolegal	0.525		0.049***	0.049***	0.064***	0.063***	−0.069*
	0.027		0.015	0.015	0.004	0.003	0.04
Access to information on PES (std)	0.009				0.032**	0.031*	0.052***
	0.054				0.015	0.016	0.004
Social norms (std)	−0.01				0.101***	0.102***	0.093*
	0.054				0.029	0.029	0.049
Proenvironment (std)	0.003				0.024	0.024	0.040*
	0.054				0.023	0.024	0.022
Prosocial (std)	0.008				−0.029	−0.03	0.006
	0.055				0.032	0.033	0.047
Progovernment (std)	−0.009				0.083***	0.082***	0.092***
	0.054				0.014	0.015	0.025
High offer treatment	0.49			−0.02		−0.017	−0.256***
	0.027			0.018		0.017	0.007
Second payment offer level		−0.031
		0.085
Third payment offer level		−0.041**
		0.019
Fourth payment offer level		0.017
		0.165
Interaction high offer and proenvironment (std)							−0.030***
							0.005
Interaction high offer and prosocial (std)							−0.424***
							0.019
Interaction high offer and progovernment (std)							−0.024
							0.1
Interaction high offer and prolegal							0.264***
							0.053
Interaction high offer and access to information (std)							−0.037
							0.041
Interaction high offer and social norms influence (std)							0.038
							0.069
Observations	348	351	348	348	348	348	348

Note: APP = permanent forest preservation areas; HH = household; PES = payments for environmental services.

*p < .10. **p < .05. ***p < .01.

Result 3: Landholders with high opportunity costs are less likely to accept the extrinsic incentive.

In our third specification, we regress demographics and land characteristics on accepting a conservation payment, controlling for the high offer level, which is highly significant and negative.⁷ The results consistently show that various measures of high opportunity cost are associated with lower demand for the programs. As an example, landholders who use most of their land for agriculture are 14 percent less likely to accept the incentive for the conservation program and 10 percent less likely to accept the incentive for the restoration program (tables 6 and 7). As the number of members that participate in agricultural activity increases by one worker, the landholder is 3.8 percent less likely to accept the conservation incentive. However, having experienced a bad agricultural yield in the previous year is associated with a 10 percent greater likelihood to accept the incentive for the restoration program. If the property suffers from any erosion problems,⁸ the landholder is 5 percent less likely to accept the offer for restoration. Restoration of these properties may be too expensive for the landholder and the payments may not fully compensate the costs. In general, the income elasticity of demand was significant and negative for restoration program. As income from agriculture increases,⁹ the landholder is 20 percent less likely to accept the payment for the restoration program.

Result 4: Policies may be redundant.

In our context, a purely profit-maximizing landholder would include the expected legal cost of noncompliance with the APP zoning laws in the utility-maximization function. We find that being prolegal is associated with an 8 percent increase in landholders’ probability of accepting the conservation offer. When interacted with the randomized offer levels, prolegal landholders are 12 percent more likely to accept the offer if in the high offer treatment relative to the low offer treatment.

When analyzing the results from the restoration program, we find that prolegal landholders are 5 percent more likely to accept the offer. In the fully interacted model, the prolegal landholders are 20 percent more likely to accept the offer if they are assigned to the high offer treatment relative to the low offer treatment.

According to Frey’s model, prolegal landholders may perceive the incentive, and especially higher levels of the incentive, as supporting their self-evaluation of legal compliance. An alternative explanation is that the program’s administration costs are perceived to be too high to make it worth it for these landholders to accept the subsidy at low levels of subsidy.

It may not be immediately clear that increasing the level of the subsidy to elicit prolegal landholders to take up the program makes any fiscal sense if these landholders are already incentivized by the law; nevertheless, excluding participants on the basis of their ex ante compliance may be even more problematic if exclusion changes their behavior ex post of the intervention. However, this question is outside the scope of this analysis.

Results from Analysis of Intrinsic Motivations

We now explore the possibility that the demand response to the randomized offer levels may vary across levels of intrinsic motivations.

Result 5: For proenvironment and prosocial motivated landholders, demand for conservation and restoration programs is undermined by higher monetary incentive levels.

We find evidence that higher monetary incentives crowd-out demand for proenvironment and prosocial landholders. In the fully interacted model for the restoration program, a one-unit increase in standard deviation above the mean in the proenvironment index leads to a 4 percent greater likelihood of accepting the incentive; however, this effect is dampened by 3 percent if the landholder is assigned to the high offer treatment. Thus, interaction with the higher offer treatment crowds-out demand for the restoration program. Following Frey’s model, the high incentive may be undermining the self-determination of proenvironment landholders, as it disengages their interest in the program.

This finding is consistent with evidence from landholders’ existing levels of conservation prior to the experiment. Using the percentage of non-APP land conserved as another proxy for proenvironment behavior, we also find a negative and highly significant association, when controlling for opportunity costs, between the percentage of non-APP land conserved and the likelihood that the landholder accepts the payment for either program.

We find similar but even stronger dynamics in the case of prosocial landholders. A one-unit increase in standard deviation above the mean in the prosocial index leads to a 42 percent lower probability of accepting the higher offer for the restoration program (significant at the 1 percent level). The higher offers are significantly disengaging prosocial motivated landholders. In an earlier analysis, we found a positive association between the prosocial index and existing levels of land conservation under APP and outside of APP jurisdiction prior to introduction of the monetary incentive (De Martino et al. 2015). Thus, the introduction of a monetary incentive and especially the interaction with the higher offer disengages demand from those with prosocial motivation. According to Frey’s model, the incentive may be seen as undermining their self-determination.

Result 6: The principal matters.

Perception of the government is a significant determinant of accepting the payment. When controlling for opportunity costs and the offer level, an increase in one standard deviation above the mean in the progovernment index is associated with a 7 percent higher probability of accepting the payment for the conservation program (see table 6). If the landholder is assigned to the high offer treatment, a one-unit increase in standard deviation above the mean increases the probability of accepting the payment to 10 percent. For the restoration program, a one-unit increase in standard deviation above the mean in the progovernment index is associated with 8 to 9 percent higher probability of accepting the payment (see table 7).

In our sample, landholders who believe it is the government’s responsibility to pay to protect public resources on private and public property have greater demand for PES programs. Higher subsidies further increase demand for PES programs. On the other hand, in more qualitative analysis of the landholders who refused the monetary incentive, we found that 24 percent and 36 percent of the sample offered the conservation incentive and restoration incentive, respectively, stated they were not confident the government program would be implemented, they did not want the government to control their property, or they did not want to be burdened with the bureaucracy. Thus, we learned that the principal matters and eliciting support for, and confidence in, government programs is crucial for increasing demand of PES programs.

Result 7: Monetary incentives can override social norm concerns.

Support for Result 7 comes from tables 6 and 7. The payment offer for conservation and restoration incentivizes the landholders who are influenced by social norms. Social norms is an index derived from factor analysis and captures those landholders who plan to discuss the project with their neighbor and will enroll in the program if their neighbor enrolls.

However, since the monetary instrument incentivized them to accept before their ability to discuss the project with their neighbor, and without them having knowledge of the decision of their neighbor, the extrinsic incentive may have overridden their social norm concerns. An increase in one standard deviation above the mean in the social norms index leads to a 12 percent higher probability the landholder will accept the incentive for the conservation program, and a 10 percent higher probability the landholder will accept the offer for the restoration program.

Result 8: Information matters.

Support for Result 8 comes from tables 6 and 7. Those who have access to general information on PES, information on the specific Mina d’ Água PES scheme, and understand the APP law are 5 percent more likely to accept the offer for both the conservation and the restoration program. The results are in line with the findings of Zanella, Schleyer, and Speelman (2014); however, Zanella, Schleyer, and Speelman conclude that access to information is the most significant determinate of participation over environmental concern. We find otherwise; various motivations and the interaction of motivations with varying levels of incentives are more significant determinants of demand for PES than access to information.

Conclusion

In this study, we employ a lab-in-the-field experiment to shed light on the determinants of demand for a PES program in the state of SP, Brazil. Our results cast some doubt on a pure rational choice model. We find that demand for PES does not always monotonically increase with the level of subsidy provided. As the offer level increases, the average landholder becomes less likely to accept the payment. Other factors than pure monetary increases are likely at play.

Next, we consider the role of intrinsic motivation in determining demand for PES. Specifically, we use a taxonomy to separate multiple sources of intrinsic motivation: proenvironment, prosocial, progovernment, and social norms. We find evidence of important differences in how types of intrinsic motivation interact with experimental changes in monetary incentives. Landholders who believe the government is responsible for paying to protect water sources (progovernment) are more likely to accept the monetary offer, and their demand increases with the size of the subsidy. In contrast, landholders with prosocial motivations are categorically less likely to accept the monetary incentive, and refusal increases with higher levels of subsidy. Landholders with proenvironment motivations are less likely to accept the monetary incentive at high offer levels.

While the present study provides credible evidence on the interplay of intrinsic and extrinsic motivations in determining demand for a PES program, additional work is needed to understand the dynamic impact of a PES program on conservation behavior and on preferences and perception over time.

Our findings also indicate that higher levels of knowledge on conservation technologies are associated with higher take up of the PES offer. Financial constraints and undermining the importance of intrinsic motivations may not be the sole sources of inefficiencies in understanding conservation behavior, and program administrators may consider supplementing their outreach with education campaigns.

Overall, we show that considerable heterogeneity underlies demand for PES programs. Some of our results suggest that, at certain levels of subsidization and among certain subgroups, payments may not have additionality. This implies that in order to achieve fiscal efficiency, such programs may consider running randomized controlled trial pilots, varying the subsidy levels and eliciting motivations, to establish the price elasticity of demand in their target population. Such studies, complemented with education campaigns, can be powerful ex ante instruments to improve the design of the subsidy at the pilot stage and ultimately, to improve take-up of well-targeted recipients for the program.

Footnotes

Appendix

Acknowledgments

We would like to thank Stefano Pagiola, Caio Piza, Richard Tol, and Pedro Rosa Dias for their helpful comments. We would also like to thank participants at the 2014 “Expanding the Frontiers of Behavioral Public Economics” conference at Tulane University and at the 2014 Latin American and Caribbean Economic Association in São Paulo, Brazil, for their comments and feedback. We would also like to thank the São Paulo State Secretaria do Meio Ambiente and in particular Helena Carrascosa, Araci Kamiyama, Caroline Vigo Cogueto, and Ana Carolina Dalla Vecchia.

Declaration of Conflicting Interests

The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Funding

The author(s) disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: This research was funded by the Spanish Fund for Latin America and the World Bank Brazil Country Management Unit.

Notes

References

Arriagada

Rodrigo

. 2009. “Payments for Environmental Services and Their Impact on Forest Transition in Costa Rica.” Working Papers, Latin American and Caribbean Environmental Economics Program, Turrialba, Costa Rica.

Benabou

Roland

Tirole

Jean

. 2003. “Intrinsic and extrinsic motivation.” The Review of Economic Studies 70:489–520.

Deci

Edward L.

Ryan

Richard M.

. 1985. Intrinsic Motivation and Self-determination in Human Behavior. New York: Plenum.

De Martino

Samantha

Kondylis

Florence

Pagiola

Stefano

Zwager

Astrid

. 2015. “Do They Say As They Do? Stated versus Revealed Preferences and Take Up in an Incentives for Conservation Program.” PES Learning Paper No. 2015 -1. Washington, DC: World Bank.

Eckel

Catherine

Grossman

Philip J.

. 1996. “Altruism in Anonymous Dictator Games.” Games and Economic Behavior 16:181–91.

Frey

Bruno S.

1992. “Tertium Datur: Pricing, Regulating and Intrinsic Motivation.” KYKLOS 45:161–84.

Gneezy

Uri

Rustichini

Aldo

. 2000. “Pay Enough or Don’t Pay At All.” The Quarterly Journal of Economics 115:791–810.

Heyman

James

Ariely

Dan

. 2004. “Effort for Payment. A Tale of Two Markets.” Psychological Science 15:787–93.

Kerr

John

Vardhan

Mamta

Jindal

Rohit

. 2012. “Prosocial Behavior and Incentives: Evidence from Field Experiments in Rural Mexico and Tanzania.” Ecological Economics 73:220–27.

10.

Lepper

Mark R.

Ed Greene

David

. 1978. The Hidden Costs of Reward: New Perspectives on the Psychology of Human Motivation. Hillsdale, NJ: Lawrence Erlbaum.

11.

Lopez-Feldman

Alejandro

. 2012. “Introduction to Contingent Valuation Using Stata.” MPRA Paper 41018, University Library of Munich, Munich, Germany.

12.

Meier

Stephan

. 2007. “A Survey of Economic Theories and Field Evidence on Pro-social Behavior.” In Economics and Psychology. A Promising New Cross-disciplinary Field, edited by Frey

Bruno S.

Stutzer

Alois

, 51–87. Cambridge, MA: MIT Press.

13.

Pagiola

Stefano

Rios

Ana

Arcenas

Agustin

. 2010. “Poor Household Participation in Payments for Environmental Services: Lessons from the Silvopastoral Project in Quindío, Colombia.” Environmental & Resource Economics European Association of Environmental and Resource Economists 47:371–94.

14.

Pfaff

Alexander

Robalin

Juan A.

Arturo Sanchez-Azefeifa

. 2008. “Payments for Environmental Services: Empirical Analysis for Costa Rica.” Working Paper Series SAN08-05, Terry Sanford Institute of Public Policy, Duke University, Durham, NC.

15.

Rios

Ana

Pagiola

Stefano

. 2009. “Poor Household Participation in Payments for Environmental Services in Nicaragua and Colombia.” MPRA Paper 13727, University Library of Munich, Munich, Germany.

16.

Robalino

Juan

Pfaff

Alexander

Arturo Sanchez-Azofefia

Alpizar

Francisco

Leon

Carlos

Rodriguez

Carlos Manuel

. 2008. “Deforestation Impacts of Environmental Services Payments: Costa Rica’s PSA Program 2000–2005.” Discussion Papers dp-08-24-efd, Resources for the Future. Accessed May 30, 2013. http://www.rff.org/files/sharepoint/WorkImages/Download/EfD-DP-08-24.pdf.

17.

Sierra

Russman

. 2006. “On the Efficiency of Environmental Service Payments: A Forest Conservation Assessment in the Osa Peninsula, Costa Rica.” Ecological Economics 59:131–41.

18.

Vollan

Bjørn

. 2008. “Socio-ecological Explanations for Crowding-out: Effects from Economic Field Experiments in Southern Africa.” Ecological Economics 67:560–73.

19.

Zanella

Matheus

Schleyer

Christian

Speelman

Stijn

. 2014. “Why Do Farmers Join Payments for Ecosystem Services (PES) Schemes? An Assessment of PES Water Scheme Participation in Brazil.” Ecological Economics. Accessed June 25, 2014. http://dx.doi.org/10.1016/j.ecolecon.2014.06.004.