cash incentives and cooperative behavior: payments for...
TRANSCRIPT
Research Proposal
“Cash incentives and cooperative behavior: Payments for
Ecosystem Services and common property management in Mexico”
Patricia Yanez-Pagans∗
University of Wisconsin-Madison
May 7, 2012Preliminary draft
1 Introduction
The impacts of incentives on human behavior have long been at the center of economic theory. Under
the assumption that individuals are self-interested and short-term maximizers, economists have generally
predicted that higher economic incentives would lead to higher effort or performance (Gneezy et al. 2011).
These predictions have also been motivated by the fact that extrinsic incentives have been assumed to be
the main driver of behavior. It was only some decades ago, following Titmuss (1970) thesis indicating that
monetary compensation tends to undermine an individual’s sense of civic duty, and also thanks to pioneering
work in social psychology, that economists have acknowledged that in many situations human behavior is
driven by motivations that go beyond the extrinsic and that these may interact with cash incentives in unex-
pected ways. Prosocial and cooperative behavior are examples of human conducts that can be explained by
multiple motivations 1. A growing amount of literature, mostly based on experimental evidence, has shown
that individuals cooperate more than expected when confronted with social dilemmas (Ostrom 1990), and
that they devote extra time or income to help others (Ariely et al. 2009). Moreover, several authors have
discussed the roles of intrinsic motivation and social norms as drivers of these behaviors, showing that in
some contexts, economic incentives may reinforce these motivations but in others they may backfire them
(Fehr & Falk 2002). While there has been a lot of experimental work analyzing these issues, externally valid
studies are still rare.
This study will use real-life behavior data to analyze how cash incentives for forest conservation in-
fluence household’s cooperation in common property management. I will exploit the unique situation in
Mexico, which has more than 70 percent of its forest land managed by common property communities and
∗Electronic address: [email protected] authors tend to use these two terms indistinctly, but here I distinguish them. Pro-social behavior makes reference to
actions that benefit other people or society as a whole. Cooperative behavior involves the process of working or acting togetherto achieve a common goal.
1
that is currently implementing one of the largest Payments for Ecosystem Services (PES) programs in the
developing world, namely the Program of Payments for Hydrological Services (PSAH) (Munoz-Pina et al.
2008) 2. In recent years, the world community has hastened its efforts to create mechanisms that minimize
the adverse use of land to mitigate climate change. One such mechanism is PES that provide landowners
cash or in-kind incentives in exchange for changing their land stewardship practices to provide some sort
of environmental service, such as safeguarding watersheds or increasing carbon sequestration by protecting
standing forests (Wunder et al. 2008). To date, many developing countries have been experimenting with
PES to achieve goals under agreements for the Reduction of Emissions from Deforestation and Degradation
(REDD). Despite active debate about their environmental effectiveness, we still know very little about how
these monetary incentives are changing rural livelihoods in the developing world.
Within Mexican communities, unpaid community work has a long history and has been a central element
not only for forest conservation but for the provision of public goods. One characteristic of the Mexican
PSAH program is that it does not impose any rules on participants about the use of program’s funds as
long as forest cover is preserved. Field work has shown that payments received by communal properties
are increasingly being used by community leaders to provide economic incentives to households to perform
forest management activities (FMA) that were previously uncompensated, for example, opening fire breaks
or doing forest patrols. To date, the PES literature has given little attention to the question of how the
introduction of monetary incentives influences households’ decisions to cooperate in common property man-
agement. This is a relevant question given that many rural areas in developing countries are characterized
by long-standing traditions of norm-based collective action (Kerr et al. 2011) and there is the possibility
that monetary incentives provided by PES programs may interact with social norms reinforcing or under-
mining cooperative behavior. For example, cash incentives could suggest to individuals that they should not
do community work without payment any more. On the other hand, if people perceive they are being re-
warded for their effort, incentives could signal that participation in community work is good. I will claim here
that the final effect will depend on the drivers of cooperative behavior as well as on how incentives are framed.
The first part of this project will start then by answering the question: how do cash incentives for conser-
vation affect households’ willingness to participate in community work? Understanding how PES incentives,
in general, influence cooperative behavior is an important question not just for program efficiency but, most
important of all, for the sustainability of Mexican communities, given that the provision of many public goods
strongly depends on households providing free labor to the community3. The second part of this paper will
answer a more specific question: How does the framing of the incentive impacts cooperative behavior? This
is also a relevant question since case studies on the distributional arrangements after a PES contract in
Mexico have shown that community leaders choose from a range of options (Munoz-Pina et al. 2008): some
divide equally the payment among members with an informal agreement of cooperation, others use the funds
to pay wages for performing some specific FMA, and others invest in public goods. Several studies, mostly
within contract theory, have shown that people may respond positively or negatively depending on how the
incentive is given to them and how they perceive it.
2The main objective of the Mexican PSAH program is to increase the production of hydrological services by promoting forestconservation.
3For example, some empirical evidence from Mexican communities in Oaxaca indicates that 80 percent of the costs ofproviding public goods, such as opening roads, is covered by unpaid community labor, while the rest is covered by governmentor private funds.
2
To understand the mechanisms driving behavior, the relationship between community leaders and house-
holds is modelled within a principal-agent framework, where leaders decide the optimal allocation rule for the
PSAH income anticipating households’ response, and households choose the extent of their cooperation based
on the incentive they receive. The main argument behind the theoretical approach is that cash incentives for
forest conservation might change social norms about cooperation making free riding more or less acceptable.
This hypothesis is based on the findings of several empirical studies that suggest that providing economic
incentives to people for obeying social norms may weaken norm enforcement and lead to the gradual erosion
of norm-guided behavior (Fehr & Falk 2002). The predictions of the model constructed here, indicate that
implicit contracts or lump-sum transfers provided to households without clear work conditionalities may
reinforce social norms of cooperation increasing households’ participation in community work. As opposed,
when payments are given for very specific activities they could signal the market value of community labor
leading to a reduction in cooperation in unpaid community work.
To answer these questions, I will use data from a national household and community survey that was
designed and run by a multidisciplinary team of researchers from the University of Wisconsin-Madison, Duke
University, and Amherst College in the summer of 2011. I exploit the time-dimension in the survey, which
was introduced through recall questions, to identify the impacts of cash incentives on both the decision to
cooperate and the intensity of cooperation. The counterfactual case is constructed based on the behavior of
rejected applicants from the Mexican PSAH program. This allows to control for key unobservable character-
istics that may simultaneously influence program enrollment and collective action at the community level,
for example, the level of institutional capacity and/or opportunity costs. By looking at real-life behavior,
this study contributes to a growing literature, both in social psychology and economics, that looks at the
impacts of incentives on human conduct and argues that different non-pecuniary motives shaping behavior
may interact with incentives in unexpected ways. Moreover, by using the monetary compensation from the
Mexican PSAH program to identify the impacts of cash incentives on cooperation, this study also contributes
to the emerging socioeconomic literature related to PES that looks at household time allocation decisions.
This research proposal is organized as follows. In the next section, I will explain how the proposed study
fits into the literature of incentives and cooperative behavior, trying to emphasize its potential contributions.
In section 3, I discuss the setting of the study and provide an overview of the Mexican Payments for
Hydrological Services Program. In section 4, a theoretical framework is presented, which provides a formal
story about the mechanisms that may be driving behavior within Mexican communities. In section 5, I
discuss my empirical strategy. Finally, in section 6, I discuss some potential conclusions and limitations of
this study.
2 Relation to existing literature
The impacts of incentives on human behavior have long been at the center of the economic theory; however,
a narrow view of the motivations driving behavior has limited our understanding of the interactions between
economic incentives and human conduct (Fehr & Falk 2002). Traditional economic theory assumes that
individuals are selfish and rational, and emphasizes that extrinsic motivation or the interest in having more
income is what usually drives behavior. If this holds, then individuals would just follow the law of supply,
and higher monetary incentives will inevitably lead to more effort or higher performance (Gneezy et al.
3
2011). Titmuss (1971) was the first to challenge the idea that there is always a positive response to cash
incentives. He suggested that individuals are more willing to donate blood voluntarily than when they are
offered a monetary compensation. Although Titmuss did not present robust empirical evidence, his thesis
attracted substantial attention both in social psychology and economics.
Following Titmuss ideas, a growing literature has argued that different non-pecuniary motives may shape
human behavior and interact with cash incentives in unexpected ways. A first group of studies discuss,
both theoretically and empirically, the crowding-out of intrinsic motivation by extrinsic incentives. Within
these studies, individuals undertake certain activities because they like them or they derive satisfaction from
doing them. Findings in a variety of social interactions show that in the presence of intrinsic motivation
monetary compensation may reduce the effort or the time devoted to a specific task (e.g Deci 1971, Lepper
et al. 1973, Andreoni 1990, Frey & Oberholzer-Gee 1997, Frey & Jegen 2001). Many of the authors within
this literature have used reduced-form models assuming a direct link between incentives and motivation. As
Benabou & Tirole (2006) argue, a more discriminatory analysis is needed, as it is difficult to always assume
that incentives will crowd-out motivation, and there are many examples, particularly in the labor literature,
showing that incentives do work (Prendergast 1999, Lazear 2000).
A second group of studies propose that social norms might be the ones driving behavior. Within this
literature, several authors have pushed forward the construction of a new behavioral theory where individuals
are boundedly rational, there is moral behavior, and concerns about social approval (Akerlof 1980, Selten
1990, Rabin 1993, Ostrom 1998, Benabou & Tirole 2006). We can think of multiple social norms, and also
rules derived from them, that interact with economic incentives in unexpected ways. Some empirical studies
discuss norms of reciprocity in a principal-agent framework and show that if the agent perceives the actions
of the principal as kind then she values the pay-off positively. On the contrary, if the actions are perceived
as hostile then she values the pay-off negatively (Fehr et al. 1997, Fehr & Falk 2002). Other authors discuss
concerns of social reputation or self-image and how they interact with economic incentives. In this con-
text, individual behavior should follow closely what the society defines as appropriate and rewards (Fehr &
Gachter 2000, Ariely et al. 2009, Carpenter & Myers 2010).
Based on field observation and the history of community work in Mexico, my study focuses on the
second group of studies that propose that social norms drive behavior. This theory of behavior is particu-
larly important for collective action problems (Ostrom 1998, Vatn 2009) and characterizes natural resource
management in many rural areas of the developing world (Baland & Platteau 1996). The main argument
behind my theoretical approach is that cash incentives for forest conservation might change social norms
about cooperation making free riding more or less acceptable. This hypothesis is based on the findings of
several empirical studies. A survey of the literature presented by Fehr & Falk (2002) suggests that rewarding
people monetarily for obeying social norms may weaken norm enforcement and lead to the gradual erosion
of norm-guided behavior. More specifically, there is evidence presented by Gneezy & Rustichini (2000) that
shows that introducing a monetary fine for late-coming parents at day-care centers increased the number of
late-coming parents. In this case, the fine was changing the rules of behavior making it more acceptable to
leave your child until late. Finally, Fuster & Meier (2010) present laboratory experimental evidence, based
on public goods games, and show that free riders are punished less harshly when incentives were in place,
which in fact led to reductions in the average contribution to public goods.
4
This study is one of a number of papers that are not only interested in the impact of monetary incentives
per se but that explore how the framing of the incentive plays a role in the behavioral response. Within
this literature, several studies have looked at how the nature of the incentive (kind versus cash) affect peo-
ple’s behavior. For example, Heyman & Ariely (2004) propose that in situations that are framed as social,
such as helping someone to move, monetary incentives might diminish the perceptions of the interaction as
social reducing the help provided. Very recently, and using a field experiment, Ashraf et al. (2012) found
that non-financial rewards are more successful at eliciting effort in pro-social tasks when compared to finan-
cial rewards. The main explanation provided by the authors is that non-financial rewards facilitate social
comparisons among individuals and increase their motivation. The framing of the incentive has also been
discussed within the literature related to contract theory. Some studies have shown that, in the presence
of reciprocally fair actors, implicit contracts work better than explicit contracts in promoting higher levels
of cooperation (Fehr & Schimdt 2000, Fehr & Gachter 2002) 4. A closely similar finding is presented by
Rand et al. (2009) who show that reward outperforms punishment in repeated public goods games. This
second group of studies are probably most closely related to this research project, as it will be looking at two
different types of monetary incentives: lump-sum transfers, which could be conceived as implicit contracts,
and wages, which are more closely related to explicit contracts.
As opposed to the studies reviewed before, most of which have used laboratory or field experimental
evidence, this study looks at real-life behavior. Although experimental evidence has the advantage of of-
fering a clean identification of the effects, there is a lot of discussion around the external validity of these
findings. There are two main arguments against the experimental evidence. First, the populations that are
being studied (usually college students in laboratory experiments) might behave very differently from the
populations of interest. Second, even if populations are very similar, such as the case of field experiments,
the behavior of people in games can be totally different to their real-life behavior. In this sense, this study
will use more accurate data about cooperative behavior in a real-life situation, making its external validity
more acceptable; however, this may come at the cost of having a less ideal identification strategy. It is also
important to mention that in most experimental evidence incentives are exogenously given and set by the
researcher. In this study, I will exploit the unique situation in Mexico where community leaders design the
incentive anticipating households’ behavior. Therefore, understanding the interaction between the principal
and the agents will be very important not only for the predictions of the theoretical framework but for the
empirical strategy of this paper.
To the best of my knowledge, this study is also one of the first to formally test the effects of PES on
cooperation decisions in common property management. Given the limited availability of household-level
data from beneficiaries and non-beneficiaries of PES programs, we still know very little about the socioe-
conomic impacts of these programs. There is an emerging literature, mostly using data from the Chinese
Sloping Lands Conversion Program that shows that, by releasing credit constraints, cash incentives from this
program promote more off-farm labor (Groom et al. 2007, Uchida et al. 2009). There have also been some
few studies that discuss how the distribution of PES incentives within communities influences households’
perceptions about these benefits and their use of environmental resources. These studies show that social
4In the explicit contract the principals explicitly conditioned a fine on the agent’s deviation from a desired effort level. In theimplicit contract they promised a bonus after the effort was observed. The promise was not binding and it was just consideredcheap talk.
5
and economic factors are important to determine perceptions, and that payments could be reducing house-
holds’ willingness to cooperate by changing their logic from doing what is considered appropriate to start
thinking about what is individually best to do. Although these last studies seem to be close to my approach
they do not offer robust empirical evidence. Sommerville et al. (2010), who discuss about perceptions, use
semi-structured interviews in Madagascar but do not include outcomes of cooperation in their analysis. Vatn
(2010) discusses cooperative behavior but does not use any particular data to prove his argument.
A very recent study by Kerr et al. (2011) suggests the possibility that incentive payments coming from
PES might influence collective action to manage common property. This study runs experiments 5, both
in Mexico and Tanzania, and concludes that providing cash incentives raises participation where people are
otherwise uninterested, but that participation is always high when social norms about cooperation are strong.
My study differs from the previous one in 3 ways. First, although Kerr et al. (2011) is looking at real-life
behavior, there is an important element about principal-agent interaction that is missing. In particular, in
their study the researcher sets the incentive and interacts with households directly; moreover, the payment
is offered for a single day of work. In real-life, community leaders are the ones who design the incentive
and cooperation is repetitive. There are reasons to believe that households might behave differently when
confronted to an outsider and also when cooperation is only related to a single day. A second distinction,
is that Kerr et al. (2011) is looking only at participation decisions while this study looks at both the par-
ticipation decision and the intensity. Looking at both outcomes is important in contexts where community
work has a long history, as it might be difficult for households to shirk completely so only changes in the
intensity of their participation are expected. Finally, as opposed to the experiments presented by Kerr et al.
(2011) and many others in the literature, this study looks not only at how incentives change participation in
activities where monetary compensation is provided but, most important of all, how they affect participation
in other activities that remain unpaid but that are of similar nature. In settings with high levels of poverty,
we might expect to see an increase in households’ participation in activities that are being paid; however,
an important question is what happens with those tasks that remain unpaid. As it was mentioned before,
there is the risk that by providing payments for some activities households will not be willing to perform
other types of community work for free any more.
3 The setting
Mexico has been facing both high deforestation and severe water scarcity in recent years. According to the
Mexico’s National Water Commission, in 2003 two thirds of the country’s most important aquifers were
over exploited (CNA 2003); moreover, between 1990 and 2000, approximately, 8 million hectares of forest
were transformed into agricultural fields or pastures (Velasquez et al. 2002). In the past decades, several
policy instruments have been used by the Mexican government to reduce deforestation, such as regulations
of land-use change in natural or protected areas, subsidies for sustainable agricultural practices, and police
action to stop timber theft, among the most important. The Payments for Hydrological Services program
(PSAH) was first implemented by the National Forestry Commission (CONAFOR) in 2003 with the intention
5For example, in their experiments in Mexico, they provide some randomly selected households living in communitiesparticipating in a PES program 3 different treatments. In the first one, a payment was never offered or mentioned; in thesecond one, an individual cash payment was offered for their cooperation; in the third one, a payment to support the villagewas offered for each participant.
6
of affecting the numerous low-value small-scale land-use changes that usually went beyond the government’s
capacity to regulate.
The main objective of the PSAH program is to increase the production of hydrological services by pro-
moting forest conservation. For this, five year renewable contracts are signed with both individual and
communal landowners and payments are made annually. The program classifies forests according to their
importance for aquifers and watersheds, and currently pays annually US 27.3 per hectare of forest enrolled
and US 36.4 for cloud forests 6. The minimum amount of land required to enroll in the program is 50
hectares and the maximum is 4000 hectares. Participants need to maintain the forest cover in the land
enrolled, and monitoring is made by satellite image analysis and ground visits. Landowners are removed
from the program if there are signs of deforestation in the enrolled land; moreover, payments are reduced if
there is forest loss caused by natural causes, such as fires or pests (Munoz-Pina et al. 2008). To comply with
program requirement, participants are encouraged to perform some forest management activities (FMA),
such as constructing firebreaks to prevent fires or doing forest patrols to avoid any type of land-use change.
The program does not impose specific requirements on the type of activities or the intensity in which they
should be done. This creates a lot of heterogeneity across beneficiaries in terms of the effort they put into
FMA.
Common property communities are an important part of the pool of beneficiaries in the PSAH program.
In 2008, approximately 45% of program recipients were common property communities. This gives the
program not only a unique feature in terms of its poverty reduction potential 7 but offers the opportunity
of understanding how cash incentives interact with common property management decisions. Authority in
Mexican communities is well defined and is divided into 3 bodies. The first one is the asamblea, which is
the principal decision-making body and where all households with land-use rights participate and vote. The
second one is the comisariado, which is the executive body and is composed by a president, a secretary, and
a treasurer. The third one is the consejo de vigilancia, which monitors the activities of the comisariado and
it is composed by a president and two secretaries. Households living in communities are divided into two
groups according to whether they have or not land-use rights. The community usually assigns those with
land-use rights, called ejidatarios, a parcel of land that they use for both residential and production purposes.
Although these households legally own these parcels they are only allowed to rent and sell their land between
ejidatarios (de Janvry et al. 2012). The forest land is part of what they consider to be common property.
Households without land-use rights are usually those that came to the ejido from other parts of the country,
they could be also sons or daughters of ejidatarios, since land-use rights can only be transfered to one children.
In most communities there is an old tradition of families performing community work, which consists of
non-paid activities that benefit all (VanWey et al. 2005). Mostly households with land-use rights allocate
time to community services, but there are also communities where households without these rights contribute
as well. Some examples of community work are cleaning roads, painting schools, or building communal
6Payment rates were originally based on approximate calculations of the average opportunity cost of land conversion fromforest to maize crops (Munoz-Pina et al. 2008)
7PES programs in many countries benefit private landowners which are not necessarily at the bottom of the income distribu-tion. In the Mexican case, communal property allows very poor households to access to the benefits of this program. Accordingto data presented by CONAFOR, in 2006, 78 of payments went to forests owned by people living in population centres withhigh or very high marginalization rates. Moreover, according to data from the National Institute of Statistics, in 2004, 31 ofthe PSAH recipients had incomes below the extreme poverty line.
7
infrastructure. For communities that possess large amounts of forest land, forest management activities
(FMA) are usually the most important components of community work. Fieldwork has shown that there
is a lot of heterogeneity in the amount of community work that households perform both within as well as
across communities. In some communities, households work for the community once a week, while in others
they just help twice in a year.
4 Theoretical framework
The relationship between community leaders and households is modelled within a principal-agent framework.
In particular, a simple static and partial equilibrium model is constructed where community leaders choose
the optimal rule of distribution for the income coming from the PSAH program anticipating households’ co-
operative behavior, and households choose the extent of their participation in community work in response
to the received cash incentives. A central element of the model is that the distributional rule adopted by
leaders could change the social norms of cooperation, making free-riding behavior more or less acceptable.
Before presenting the formal logic of the model, I discuss the basic intuition. Every year community leaders
receive the money from the PSAH program and need to decide how to allocate it. In order to continue in the
program they need to satisfy the requirements explained in the previous section and, for this, they need to
have households’ support in order to perform some FMA. Therefore, leaders need to anticipate households’
responses in order to offer the best incentive that, ideally, allows them to obtain not only their support for
FMA but their support in all other types of community work. Given the way authority is organized within
communities, I assume that leaders cannot get private gains from the program and that they maximize the
utility of the community.
As long as program requirements are met, leaders are free to decide how to allocate program funds.
Fieldwork has shown that there are two distributional arrangements. In the first one, leaders use PSAH
income to provide lump-sum transfers to households with no specific work conditionalities but rather an
implicit agreement of cooperation in FMA. In the second one, leaders use the money to pay wages for some
very specific FMA. What determines which distributional rule is adopted is an important question of this
research project. It is important to acknowledge that it is assumed that households cannot individually
manipulate the distributional rule choice. Although households participate in asambleas and have the right
to vote, there is heterogeneity in their involvement and participation in these meetings. In general, leaders
seem to be the ones proposing and taking the most important decisions for the community. Fieldwork has
shown that most households within communities are not familiar with the program and they don’t have
direct contact with outside organizations. Usually, what happens is that leaders take the decision about the
distributional rule, then they present their proposal in the asamblea, and most households will accept it.
4.1 Community leaders’ problem
Community leaders need to make two choices when maximizing community’s net benefit from the PSAH pro-
gram. First, they need to choose the optimal amount of forest land (F ) that will be enrolled in the program.
Second, they need to decide how to allocate the program’s income, which means they need to choose the
optimal proportion of funds that will be distributed directly to households, as lump-sum transfers, without
specific working conditionalities (γ) or, inversely, the proportion of program’s income that will be used to
8
provide wages for very specific forest management activities (1 − γ) 8 It is important to acknowledge that
there are several advantages and disadvantages resulting from the distributional rule. Given that land own-
ership is clearly established within communities, lump-sum transfers are usually given to households with
land-use rights (i.e. ejidatarios). Since there are no clear work conditionalities, these transfers could help
to increase cooperation not only in FMA but in other types of community work; however, they could also
create the opposite effects on households that are not receiving them (i.e. non-ejidatarios) (Alix-Garcia et al.
2005). In terms of wages, they could be given to anyone in the community, regardless of their land-use rights
and these will probably increase participation in paid activities, however by increasing the number of paid
activities there is the risk that they could weaken social norms about cooperation. If we put community
leaders decisions in the context of contract theory, we can think of them as deciding between providing
some sort of implicit contract (lump-sum transfers) versus an explicit contract (wages)9. As this theoretical
framework pretends to show, the behavioral implications of each of them could be very different.
Let pf be the payment per hectare of forest that is set aside for conservation and that is given by the
PSAH program. g(∑Ni=1 l
ui +
∑Ni=1 l
pi ) is a benefit function that indicates that the community gains from
the aggregate amount of labor provided by the N households in the community both in paid (lp) and un-
paid (lu) activities. Community costs of program implementation are represented by c(γ, x, F, θ). This cost
function is decreasing in the proportion of program’s income that goes directly to households (∂c(.)∂γ < 0)
since designing contracts and closely monitoring each paid activity is more costly. The cost function is
decreasing in the visibility of households’ actions (∂c(.)∂x < 0), since this facilitates monitoring both in paid
and unpaid community work, and it is increasing in the amount of forest land that is enrolled in the program
(∂c(.)∂F > 0). Finally, θ is a vector of community characteristics that may increase or decrease the costs of pro-
gram implementation and may influence the distributional rule. For example, geographical characteristics,
such as the size or the type of forest, or population characteristics, such as the average level of education
or the age of households with land-use rights. This parameter is crucial in the model as it helps to under-
stand why leaders may prefer one distributional rule over the other. As part of this research project, I will
need to analyze carefully this decision-making process. For this, I will use results from case studies as well
as survey data. As it will be explained later, this is also essential for the empirical identification of the effects.
The leaders maximization problem can be expressed as:
maxF,γ
{pfF + g(
N∑i=1
lpi +
N∑i=1
lui )− c(γ, x, F, θ)} (1)
s.t. F ≤ F̄
γ ≤ 1
F, γ ≥ 0
8In reality, this is a simplified description of the community leaders problem. I rule out the possibility of investing in publicgoods or of leaders keeping the money for themselves. As it was explained before, authority in Mexican communities is verywell structured and is composed by different bodies that continuously monitor each others work; therefore, I assume communityleaders’ decisions cannot be driven by private interests.
9We can think of explicit contracts as those that clearly specify the rights and obligations of the worker and impose finesin case of deviating from those actions. When we think about wages, the fine is just not receiving the payment if you do notwork. Implicit contracts can be seen as cooperation promises that are not binding and are just cheap talk (?).
9
But leaders anticipate that households’ cooperation decisions will depend on the distributional rule they
choose. This means that the aggregate amounts of labor supplied by households, both in paid and unpaid
community activities, are a function of the proportion of program’s income that goes to lump-sum transfers.
More specifically,∑Ni=1 l
p(γ) and∑Ni=1 l
u(γ). Then, with no binding constraints, the first order conditions
(FOC) are:
pf =∂c(γ, x, F, θ)
∂F(2)
N [∂g
∂lp∂lp
∂γ+∂g
∂lu∂lu
∂γ] =
∂c(γ, x, F, θ)
∂γ(3)
The FOC indicate that leaders will choose to distribute the proportion of income that will allow them to
equalize the benefits from community participation with the costs of implementing this rule. Solving those
two equations should give us the optimal distributional rule:
γ∗ = γ(x, θ, pf , N) (4)
4.2 Household problem
A household i decides how much time to allocate to paid community activities (lpi ), unpaid community
work (lui ), and own production activities (loi ). The total endowment of time is given by T and no leisure
is assumed to exist 10. Participating in paid and unpaid community activities entails a cost of c(lpi ) and
c(lui ), respectively. Paid community activities yield a monetary reward w, which is equal for all households.
This incentive rate is set by community leaders and is a decreasing function of the proportion of PSAH
income that is distributed directly as lump-sum transfers. This means that w(γ) and ∂w∂γ < 0. Households
get benefits from aggregate community work, and this is represented by a function ag(∑Ni=1 l
pi +
∑Ni=1 l
ui ),
where a is the marginal benefit a household gets from community work. For example, we can think of these
benefits as increased water supply or water quality, which come as a result of more time devoted to forest
management activities. For the moment, I assume a is constant across all households11.
The production function of households’ own activities is given by q(loie) and only uses labor as an input.
This function is smooth and satisfies q′(loi ) > 0 and q′′(loi ) ≤ 0 for all loi . We can think of owned production
activities as work that is done in agriculture, livestock activities, or off-farm employment. Households can
sell their production for a unit price of p. Since this is a partial equilibrium model, I assume prices remain
fixed despite any changes observed in labor markets. The average amount of unpaid community work pro-
vided by all households is given by l̄u =
∑N
i=1lui
N . The visibility of a households’ actions or effort within
the community is given by a parameter x 12, and B is the amount of lump-sum transfers that households
get from community leaders, and this is an increasing function of the amount of PSAH income that leaders
10Here T does not have a subscript i since we assume the total endowment of time is equal for all households. If we think ofli as the total amount of working hours within a day then T would be 24 hours. If we think of li, as the total number of daysworked within a month or a year then T would be equal to 30 or 365 days.
11We could think about introducing heterogeneity in the marginal benefits households get from community work. For example,households whose main income comes from agriculture could gain more from forest protection activities, or households whowork off the farm could gain more from community work related to clearing or opening roads.
12This idea is similar to the one presented by Benabou & Tirole (2006), where the reputational payoff of performing a prosocialactivity is amplified or minimized by the salience of a person’s actions.
10
decide to distribute directly to households. This means, B(γ) and ∂Bγ > 0.
Households face punishments when they deviate from the average amount of unpaid community work.
This is represented by a punishment function θR(l̄u − lui , x, w,B), where θ captures the degree to which a
household is exposed to punishment. For the moment, I assume θ is constant across households, but we
could also have heterogeneity in households’ exposure to punishment. This will be the case, for example, if
we expect households with land-use rights to be more exposed to punishment in case they deviate than those
without land-use rights. In terms of the punishment function R, this is increasing in downward deviations
of households’ unpaid community work decisions from the average ( ∂R∂(l̄u−lu
i)> 0). It is also increasing in
the visibility of actions (∂R∂x > 0), decreasing in the incentive rate (∂R∂w < 0) and increasing in the lump-sum
transfer ( ∂R∂B > 0). These last two elements are central for the predictions obtained from this model. The
basic intuition is that when cash incentives are provided for very specific community activities, this means
through w, households might be receiving a signal from leaders indicating that their work can be valued
and that they should not perform other community activities without compensation. This change in the
information available reduces the capacity and/or incentives to punish deviators 13. In terms of the lump-
sum transfers, households might perceive this income as a recognition for their cooperation and feel more
obligation to reciprocate by participating in different kinds of community work. If this is the case, deviation
becomes more costly.
Formally, the household maximization problem can be represented as follows:
maxloi,lpi,lui
{pq(loi ) + w(γ)lpi − c(lpi )− c(lui ) +B(γ) + ag(
N∑i=1
lpi +
N∑i=1
lui )− θR(
∑Ni=1 l
ui
N− lui , x, w,B(γ))} (5)
s.t. T = loi + lpi + lui (6)
The first-order conditions from this problem are:
w(γ) + a∂g
∂lpi= p
∂q
∂lpi+∂c
∂lpi(7)
a∂g
∂lui+ θ
∂R
∂lui[1− 1
N] = p
∂q
∂lui+∂c
∂lci(8)
The previous equilibrium conditions show that the optimal amount of labor allocated to both paid and
unpaid community work will be the one that equates the marginal benefits to the marginal costs of par-
ticipation. We can see from equation 8 that the marginal benefit from doing unpaid community work is
derived not only from the marginal benefits that households get from doing community work in general, but
also from the reduction in punishment that results from increasing unpaid work whenever it was below the
average value. This means that whenever the household is working an equal amount than the average there
are no extra gains from doing additional unpaid community work. If we use some very simple functional
13I borrow this intuition from experimental evidence presented by ?, who show that after introducing a monetary fine forlate-coming parents at day-care centers, the number of late-coming parents increases. They argue that incentives could bechanging the information agents have and this leads to unexpected behavior. Also, Fuster & Meier (2010), using laboratoryexperimental evidence based on public goods games, show that free rides are punished less harshly in treatments with incentives,which leads to reductions in the average contribution to public goods.
11
forms, without violating the main assumptions of the model, we can get some preliminary predictions about
the effects of different variables on the decisions to cooperate in paid and unpaid community work. Assume
the following forms:
q(loi ) = loi
c(lu,pi ) =(lu,pi )2
2
g(
N∑i=1
lpi +
N∑i=1
lui ) =
N∑i=1
lpi +
N∑i=1
lui
R =B(γ)x
w(γ)(
∑Ni=1 l
ui
N− lui )
The maximization problem can then be expressed as:
maxlpi,lui
{p(T − lpi − lui ) + w(γ)lpi −
(lpi )2
2− (lui )2
2+B(γ) + a(
N∑i=1
lpi +
N∑i=1
lui )− θ[B(γ)x
w(γ)(
∑Ni=1 l
ui
N− lui )]} (9)
The solutions to this problem are given by:
lp∗i = a− p+ w(γ) (10)
lu∗i = a− p+θB(γ)x
w(γ)(1− 1
N) (11)
We can see here that increases in the return to own production activities will decrease the time allocated
to both paid and unpaid community work (∂lp∗
i
∂p < 0,∂lu∗
i
∂p < 0). We can think of this as an opportunity cost
effect, since p is simply measuring the opportunity cost of participating in community activities. We can also
verify that increases in the marginal return to aggregate community work will increase the time allocated to
both types of community work (∂lp∗
i
∂a > 0,∂lu∗
i
∂a > 0). This is also intuitive. If households perceive they can
gain from contributing to community work, such as helping to clear roads, then they will devote more time
to do it. As it was mentioned before, heterogeneity in the parameter a could help to capture differences in
gains from specific types of community work based on household profile.
In terms of the different compensation schemes, the increase in payments for very specific community ac-
tivities leads to an increase in time allocated to community work where these payments are given (∂lp∗
i
∂w > 0).
This will be true as long as payments w are higher than the returns a household can get from own production
activities p. This is the reason why we might expect poor households to participate more in paid activities,
since usually the returns they get from own production activities is very low. We can see from equation 11
that increases in payments for specific activities reduce the time allocated to unpaid activities (∂lu∗
i
∂w < 0).
12
The main argument here is that higher payments reduce the punishment that comes from deviating from the
average behavior. In contrast to these findings, increases in lump-sum transfers have no effects on paid com-
munity work but the increase unpaid community work (∂lu∗
i
∂B > 0). Here the argument is the exact opposite,
receiving transfers without clear working conditionalities but with an implicit agreement of cooperation in-
creases the punishment from deviation. Again, for those households whose opportunity cost of participating
in unpaid community work is very high it will be better to allocate more time to own production activities
and confront the punishment costs.
Other important parameters in the model are the degree of exposure to punishment (θ)and the visibil-
ity of actions (x). As expected, results from equation 11 show that increases in the degree of exposure to
punishment increase time allocated to unpaid community work (∂lu∗
i
∂θ > 0) and increases in the visibility of
actions also increase the time allocated to community work (∂lu∗
i
∂x > 0). These two results are very intuitive.
For example, for households with land-use rights we might expect to see more punishments for deviation so
we might also expect to see more unpaid community work. Also, in communities where actions are clearly
visible, for example, in small communities or that live in very compact areas, then we might expect to see
less deviations in behavior in order to avoid punishment. As Olson (1965) argues, larger communities have
greater number of free riders, since the impact of each individual defector is smaller.
4.3 Further extensions
So far, I constructed a very simple model that allows to understand some of the possible mechanisms driving
cooperative behavior within Mexican communities when incentives for forest conservation are implemented.
There are still several improvements or extensions that could be incorporated to the previous model. First,
more accurate functional forms should be defined in order to obtain better predictions about behavior. For
example, the relationship between the distributional rule (γ) and lump-sum transfers (B) and wages for
compensated community activities (w) should be well specified. Second, I am currently assuming that all
households face the same unit costs for community work c, the same opportunity costs p, gain equally from
community work (a), and face punishments in the same way (θ). As it was mentioned before, one could
probably introduce variation in these parameters by considering different types of agents. Third, right now
there is no uncertainty in the model, but we could think about introducing it if we consider that there
is a positive probability of not receiving program payments in the next period if FMA are not enough to
satisfy the specified requirements. If this is the case, today’s cooperative behavior will determine the future
payments the community and households can get. For this, I will need to construct a dynamic model that
captures the impact of this anticipation effect on both leaders and households’ behavior.
5 Data
This study uses data from a national household and community-level survey, which was designed by a mul-
tidisciplinary team of researchers from the University of Wisconsin-Madison, Duke University, and Amherst
College, and collected between June and August of 2011 in Mexico. The survey covers both beneficiaries
and non-beneficiaries from the 2008 PSAH cohort. A stratified random sampling strategy was applied both
by region and land-use rights. In a first step, 4 regions were selected (north, center, south west, and south
east) based on dominant ecosystem type and socio-economic groupings. Within each region, and based on
13
the availability of good quality past satellite images, to monitor for deforestation over time, several Land-
sat footprints (areas of 180X180 sq km) were randomly selected in each region. Within the geographical
areas covered by these footprints, a sample of beneficiary communities that entered the program in 2008
was randomly selected. Then, a nearest neighbor covariate matching estimator was used to select the non-
beneficiary communities from the pool of those that applied to the program in 2008 but were rejected 14. In a
second step, surveyors further stratified the sample within common property communities by land-use rights.
Based on lists provided by program officers or community leaders, surveyors randomly selected 5 households
with full land-use rights and voting power (”ejidatarios”) and 5 without them (”non-ejidatarios”). The final
sample is composed of 1086 households (567 beneficiaries) and (519 non-beneficiaries) distributed over 116
common property communities 15. Figure 1 shows how the properties surveyed are distributed over the
country.
Both household and community surveys are quite comprehensive. In order to have baseline measure-
ments, surveys included recall questions from 2007, which is the year previous to program implementation.
By having information from two different points in time (i.e. 2011 and 2007) we are able to construct a panel
data set. The household survey was responded by the head of the household. Most questions are related
to household-level information, such as household assets, access to land, and production decisions, among
the most important. We also collected detailed individual-level information about education, migration, and
employment decisions. Questions related to community work and participation in forest management activ-
ities were done at the household level. The community survey was applied to a group of community leaders
and included questions about community characteristics as well as questions about decisions related to the
use and distribution of PSAH income. The community survey also included questions about the number of
households that participate in community and forest management activities. This information will be used
to construct some of the explanatory variables and could potentially be used to perform some robustness
checks. Entry and analysis of much of the survey data is still in progress; here I only use information from the
section covering household participation in forest management activities and in uncompensated community
work. I also use only a part of the information available in the community survey which is relevant to this
study.
6 Empirical strategy
To look at the impacts of economic incentives on cooperative behavior, I exploit the time variation in com-
munity access to Payments for Ecosystem Services. In particular, I compare changes observed over time
in cooperation between households that live in beneficiary and non-beneficiary communities of the PSAH
program in Mexico. As opposed to the standard prediction of self-centered behavior, cooperative behavior
exists whenever individuals maximize a joint pay-off instead of their own pay-off (Brosig 2002). In this study,
the household will be the unit of analysis and I will look specifically at households’ participation in forest
management activities (FMA). In the survey, we included a list of different FMA, such as constructing fire
14According to Alix-Garcia et al. (2012), some of the most important reasons for rejection are limited funds from theprogram (35%), having less than the required minimum forest cover (50%), being outside of the eligible zone (6%) and missingdocumentation (9%)
15Despite the land-use rights stratification strategy applied at the community level, the final number of households in thesample is not exactly divisible by 10. The reason is that in some very small communities there were less than 10 households.In other cases, communities only had households with land-use rights.
14
breaks, doing forest patrols, participating in reforestation activities, and some others, and we asked the head
of each household if any of its members participated in each of these activities in the past 12 months as
well as in 2007. Moreover, we asked them whether they received any direct payment for their work and how
many days per year they allocated to each activity. This means I will be able to distinguish between paid
and unpaid activities, and also to look at changes in participation both in the extensive and intensive mar-
gins, which implies considering not only participation decisions, but also the intensity in participation. The
analysis will also take into account cooperation in other types of community work, such as opening roads or
helping to build community infrastructure. Unfortunately, there is no time variation in this variable; there-
fore, identification here will have to exploit matching techniques and rely on cross-sectional comparisons
between beneficiaries and non-beneficiaries.
Taking into account this information, I will estimate the following baseline regression for FMA using
probit, logit, or linear models, as needed, depending on the outcome variable considered:
Yijt = β0 + β1Tt + β2Pjt + β3(Tt ∗ Pjt) + αXi + γZj + εijt (12)
Where, Yijt is the cooperation outcome of interest for individual i, living in community j, in time t. Tt
is a dummy variable capturing time that takes the value of 1 in the year 2011 and 0 in 2007, Pjt is a dummy
variable that takes the value of 1 if the individual lives in a community that participates in the PSAH pro-
gram, and Xi are household characteristics that may be important to control for, such as number of members
in the household, employment in the baseline year, level of education of the head of the household, etc. Zj is
a vector of community characteristics, such as population size, access to basic services, distance to cities, etc.
The coefficient β3 is the parameter of interest, the Difference-in-Difference (DID) estimator, and measures
the excess of growth or the decrease in cooperation for those households living in beneficiary communities
(treated) compared to those living in communities that do not participate in the program (control). The
main assumption here is that living in a beneficiary community implies a greater exposure to payments or
monetary compensation in exchange for community work. Data from the survey shows that in communities
that participate in the program approximately 52% of the households participated in a paid activity in the
past year while in those that do not participate in the program only 28% received some type of compensation.
Since standard errors may be correlated within households living in the same community, I will allow the
standard errors to be clustered at the community level.
Up to this point, there is not a clear prediction coming from the theoretical framework that could tell
us what the sign of β3 will be. As it is claimed, the final effect seems to depend on how the money was
distributed by community leaders. Moreover, the final effect will also depend on how the other drivers of
behavior interact with incentives. For example, it seems that households with less income will have lower
opportunity costs and therefore will cooperate more when monetary compensation is given to them. To
look at this and other heterogeneous effects, I will estimate a model with a triple interaction term, as the
following one:
Yijt = β0 + β1Tt + β2Pjt + β3Ii + β4(Tt ∗ Pjt) + β5(Tt ∗ Pjt ∗ Ii) + αXi + γZj + εijt (13)
Now, β5 is the coefficient of interest and measures the differences in impact across different groups of
the population distinguished by the variable Ii. We can think of this variable as the income group or as
15
land-use rights. The DID approach presented before allow us to difference away any unobservable time-
invariant characteristic. There are, however, two potential weaknesses with this approach. The first one
is that we cannot control for unobserved temporary individual-specific or community-specific components
that may influence the participation decision and that may also influence the outcome. The second one is
that some macro effects can have differential impacts across the two groups. This may happen if households
in beneficiary communities and households in non-beneficiary communities have some (possibly unknown)
characteristics that make them react differently to shocks. To overcome these limitations, I will use a DID
matching approach. As Blundell and Costa Dias (2002) indicate, this approach may significantly improve
the quality of non-experimental evaluation, as it allows us to difference away time-invariant unobservable
variables as well as time-variant factors that have parallel trends. I could then match households based on
some vector of variables X that may influence their cooperation decisions. The effect of the treatment on
the treated will be then estimated over the common support of X and the coefficient of interest will be then
defined as:
β̂ =∑iεT
[Yi1 − Yi0]−∑jεC
Wij [Yj1 − Yj0]
wi (14)
Where T and C are treatment and control groups, respectively, Wij is the weight placed on comparison
observation j for individual i and wi accounts for the re-weighting that reconstructs the outcome. Table 1
presents preliminary statistics comparing some characteristics for households in the treatment and control
groups. As it can be seen, the balance in the original sample is very good and there are no significant differ-
ences among households in both groups. There is still work that needs to be done to compare community
characteristics; moreover, the matching strategy will be also very important for the analysis of other types
of community work not directly related to forest conservation. In particular, since we only have information
about participation in other types of community work for a single year (i.e. 2011), I will match treatment and
control observations based on some household and community characteristics that may impact cooperation
in the baseline. Although we will need to be more careful with the results coming from this analysis, the
cross-sectional matching estimator constitutes an improvement over the simple between estimator.
The second part of this study, tries to understand not only whether cash incentives affect cooperative
behavior, but how the framing of the incentive affects participation in community work. For this, I exploit
the heterogeneity in the distributional rules used by community leaders related to the use of program’s
money. More specifically, within communities that participate in the program, I compare whether providing
lump-sum transfers, with an implicit agreement for cooperation, or paying wages for very specific FMA is a
more effective strategy to promote households’ cooperation. The baseline regression in this analysis is the
following one:
Yijt = β0 + β1Tt + β2Djt + β3(Tt ∗Djt) + αXi + γZj + εijt (15)
As before, Yijt is the cooperation outcome of interest for individual i, living in community j, in time t. Tt
is a dummy variable capturing time that takes the value of 1 in the year 2011 and 0 in 2007, Djt is a dummy
variable that takes the value of 1 if the individual lives in a community where lump-sum transfers are given
using the PSAH income, and Xi and Zj are household and community characteristics, respectively. The
coefficient β3 is the parameter of interest and is expected to be positive, particularly for unpaid community
16
work. Estimating equation 15 as a reduced form equation will result in biased estimates, the reason is that
the distributional rule or the type of contract adopted by community leaders is a function of the level of
cooperation in the community. At the same time, a household’s level of cooperation is a function of the
distributional rule observed in the community. This means there is a simultaneity problem, which should be
potentially corrected through the use of Instrumental Variables (IV).
Finding the right IV constitutes one of the main challenges within this research project. In particular,
using information from case studies implemented in some of the surveyed sites, I will carefully analyze the
decision-making process that takes place at the community level. Hopefully, this will help me to find an
adequate IV that predicts the distributional rule adopted by leaders but that does not affect directly cooper-
ation decisions. There is some possibility that I could also exploit some historical information coming from
community censuses, which could give some information about community leaders characteristics, religious
affiliation, indigenousness, or the presence of other ethnic groups that could be used to construct an IV.
Also, I could try to see if some geographical characteristics, such as forest type, slope or elevation affect this
decision by changing the costs of program implementing. Once this process is completed, I will be able to
use information from communities in the control group in this analysis. In this case, I could predict which
communities that are non-beneficiaries would adopt one distributional rule over the other and then compare
the changes in behavior between those that participate in the program.
To conclude, it is important to mention that although the availability of panel data helps to improve the
identification strategy, there are some problems that may arise from using a retrospective survey. In partic-
ular, there is the possibility of having retrospective recall bias if a non-random sample of households did not
provide past information. I will need to check the characteristics of households and communities where there
is past missing data and compare them with the rest of the households in order to rule out this possibility.
Based on field work observation, I believe that any missing information is just random. It is important to
keep in mind that I will still have greater noise in the data because of random memory errors. According
to the literature, this noise will bias p-values upwards, which means it is possible that some coefficients that
appear to be statistically insignificant would in fact be significant if there was less measurement error. In my
case, recall bias will mostly affect the dependent variable of cooperation, specifically when I use the number
of days per year that households worked on community activities. It will not affect the main explanatory
variables, which are program participation or the distributional rule.
7 Potential conclusions and limitations
The UN-REDD Program (2011) estimates that future financial flows for REDD programs, including PES,
may be up to US $ 30 billion a year. Given the increasing importance of these initiatives not only from an
environmental but also from a development perspective, it is essential to understand how these incentives
are changing rural livelihoods in the developing world. Using one of the largest PES program in the world,
namely the Mexican Payments for Hydrological Services Program (PSAH), this study contributes not only
to the emerging literature on the socioeconomic impacts of PES programs, but also exploits a unique set-
ting that allows to understand how monetary compensation changes households’ willingness to cooperate in
common property management. By looking at real-life behavior, this study pretends to highlight not only
17
the ambiguous relationship between economic incentives and the multiple drivers of cooperative behavior,
but also how the framing of the incentive matters. In contexts with very strong social norms, I expect to see
no differences in behavior no matter how the incentive is given to households; however, when social norms
are weak, incentives could promote changes in very old community work traditions by pointing out that
some activities should not be done for free anymore. This will be the case when incentives are distributed
to households in the form of wages for very specific community activities. As opposed, when incentives are
provided as lump-sum transfers to households without clear work conditionalities, then cooperation could be
reinforced. I also expect to see heterogeneity in households’ reaction to monetary incentives, based on their
income and land-use rights. Regardless of how the incentive is given to them, low income households will
always increasingly participate in community work once they know incentives are being provided. The main
reason is that they face lower opportunity costs and they need this extra income. For households with land-
use rights, my preliminary prediction is that they will cooperate more when lump-sum transfers are given to
them, since these are usually distributed only among those with clear property rights and households’ may
perceive them as a recognition for their ownership and their work for the community.
It is important to mention that this is not an experimental study, so it carries some limitations that
need to be acknowledged. First, the design of the incentive is not exogenously driven but it is determined
endogenously within the relationship of community leaders and households. To the extent that this study
is able to find an adequate instrumental variable for the distributional rule, I will be able to comment on
the causal relationship between the framing of the incentive and cooperative behavior. Another important
limitation is that it might be difficult to clearly separate or identify the mechanisms driving behavior. For
example, if we observe that cooperation decreases when wages are provided, can we conclude that this is
because incentives weakened social norms or is it intrinsic motivation that was crowded-out or a reciprocity
response? So far, the theoretical model I constructed provides a story about social norms, which is mostly
based on fieldwork experience. I am conscious, however, that in the empirical section I will need to provide
robust evidence to rule out some of the possible alternative explanations. Since I do not have experimental
measures that could allow to distinguish between one explanation or the other, I will try to exploit the data
to give some tentative answers to these questions. For example, I could try to see if there is heterogeneity
in the deviations from average behavior in contexts with stronger social norms as measured by community
rules and punishments. I could also try to construct a proxy measurement for household intrinsic motiva-
tion to help others, by using information about a household’s participation in several community activities
(parties, committees, authority positions) as well as information on how much they help other people in the
community by giving gifts or transfers.
References
Akerlof, G. (1980), ‘A theory of social custom, of which unemployment may be one consequence’, Quarterly
Journal of Economics 37, 291–304.
Alix-Garcia, J., A, d. J., Sadoulet, E., Torres, J., Brana, J. & Zorilla, M. (2005), An assessment of mexico’s
payment for environmental services program, Technical report, Food and Agricultural Organization of the
United Nations.
18
Alix-Garcia, J., Sims, K., Yanez-Pagans, P., Radeloff, V. & Shapiro, E. (2012), Two dimensional evaluation:
The environmental and socioeconomic impacts of mexico’s payments for hydrological services program.
Unpublished manuscript.
Andreoni, J. (1990), ‘Impure altruism and the donations to public goods: a theory of warm glow giving’,
The Economic Journal 100, 467–477.
Ariely, D., Bracha, A. & Meier, S. (2009), ‘Doing good or doing well? image motivation and monetary
incentives in behaving prosocially’, American Economic Review 99(1), 544–555.
Ashraf, N., Bandiera, O. & Jack, K. (2012), No margin, no mission? a field experiment on incentives for
pro-social tasks. Unpublished manuscript.
Baland, J. & Platteau, J. (1996), Halting Degradation of Natural Resources: Is There a Role for Rural
Communities, FAO and Clarendon Press, Oxford.
Benabou, R. & Tirole, J. (2006), ‘Incentives and prosocial behavior’, American Economic Review
96(5), 1652–1678.
Brosig, J. (2002), ‘Identifying cooperative behavior: Some experimental results in a prisoner’s dilemma
game’, Journal of Economic Behavior and Organization 47, 275–290.
Carpenter, J. & Myers, C. (2010), ‘Why volunteer? evidence on the role of altruism, image, and incentives’,
Journal of Public Economics 94, 911–920.
CNA (2003), Determinacion de zonas criticas para la recarga de acuiferos, Technical report, Diario Oficial
de la Federacion (DOF), Mexico.
de Janvry, A., Emerick, K., Gonzalez-Navarro, M. & Sadoulet, E. (2012), Certified to migrate: Property
rights and migration in rural mexico.
Deci, E. (1971), ‘Effects of externally mediated rewards on intrinsic motivation’, Journal of Personality and
Social Psychology 18(1), 105–115.
Fehr, E. & Falk, A. (2002), ‘Psychological foundations of incentives’, European Economic Review 46, 687–724.
Fehr, E. & Gachter, S. (2000), ‘Cooperation and punishment in public goods experiments’, American Eco-
nomic Review 90(4), 980–994.
Fehr, E. & Gachter, S. (2002), Do incentive contracts undermine voluntary cooperation?
Fehr, E., Gachter, S. & Kirchsteiger, G. (1997), ‘Reciprocity as a contract enforcement device - experimental
evidence’, Econometrica 65, 833–860.
Fehr, E. & Schimdt, K. (2000), ‘Fairness, incentives, and contractual choices’, European Economic Review
44, 1057–1068.
Frey, B. & Jegen, R. (2001), ‘Motivation crowding theory’, Journal of Economic Surveys 15(5), 589–611.
Frey, B. & Oberholzer-Gee, F. (1997), ‘The cost of price incentives: An empirical analysis of motivation and
crowding-out’, The American Economic Review 87(4), 746–755.
19
Fuster, A. & Meier, S. (2010), ‘Another hidden cost of incentives: The detrimental effect on norm enforce-
ment’, Management Science 56(1), 57–70.
Gneezy, U., Meier, S. & Rey-Biel, P. (2011), ‘When and why incentives (don’t) work to modify behavior’,
Journal of Economic Perspectives 25(4), 191–210.
Gneezy, U. & Rustichini, A. (2000), ‘A fine is a price’, Journal of Legal Studies 29(1), 1–17.
Groom, B., Grosjean, P., Kontoleon, A., Swanson, T. & Zhang, S. (2007), Relaxing rural constraints: a
win-win policy for poverty and environment in china? Environmental Economy and Policy Research.
University of Cambridge.
Heyman, J. & Ariely, D. (2004), ‘Effort for payment: A tale of two markets’, Psychological Science
15(11), 787–793.
Kerr, J., Vardhan, M. & Jindal, R. (2011), ‘Prosocial behavior and incentives: Evidence from field experi-
ments in rural mexico and tanzania’, Ecological Economics 73, 220–227.
Lazear, E. (2000), ‘Performance pay and productivity’, American Economic Review 90(5), 1346–1361.
Lepper, M., Greene, D. & Nisbet, R. (1973), ‘Undermining children’s intrinsic interest with extrinsic rewards:
A test of the “overjustification” hypothesis’, Journal of Personality and Social Psychology 28, 129–137.
Munoz-Pina, C., Guevara, A., Torres, J. & Brana, J. (2008), ‘Paying for the hydrological services of mexico’s
forests: Analysis, negotiations and results’, Ecological Economics 65, 725–736.
Ostrom, E. (1990), Governing the Commons, Cambridge University Press, Cambridge, UK.
Ostrom, E. (1998), ‘A behavioral approach to the rational choice theory of collective action: Presidential
address, american political science association, 1997’, The American Political Science Review 92(1), 1–22.
Prendergast, C. (1999), ‘The provision of incentives in firms’, Journal of Economic Literature 37(1), 7–63.
Rabin, M. (1993), ‘Incorporating fairness into game-theory and economics’, American Economic Review
83(5), 1281–1302.
Rand, D., Dreber, A., Ellingsen, T., Fudenberg, D. & Nowak, M. (2009), ‘Positive interactions promote
public cooperation’, Science 325.
Selten, R. (1990), ‘Bounded rationality’, Journal of Institutional and Theoretical Economics 146, 649–658.
Sommerville, M. J. J., Rahajaharison, M. & Milner-Gulland, E. (2010), ‘The role of fairness and benefit
distribution in community-based payment for environmental services interventions: A case study from
menabe, madagascar’, Ecological Economics 69, 1262–1271.
Titmuss, R. (1970), The gift relationship, Allen and Unwin, London.
Uchida, E., Rozelle, S. & Xu, J. (2009), ‘Conservation payments, liquidity constraints, and off-farm labor:
Impact of the grain-for-green program on rural households in china’, American Journal of Agricultural
Economics 91(1), 70–86.
VanWey, L., Tucker, C. & Diaz, E. (2005), ‘Community organization, migration, and remittances in oaxaca’,
Latin American Research Review 40(1), 83–107.
20
Vatn, A. (2009), ‘Cooperative behavior and institutions’, The Journal of Socio-Economics 38, 188–196.
Vatn, A. (2010), ‘An institutional analysis of payments for environmental services’, Ecological Economics
69, 1245–1252.
Velasquez, A., Maas, J., Bocco, G. & Ezcurra, E. (2002), ‘Patrones y tasa de cambio de uso del suelo en
mexico’, Gaceta Ecologica 62, 21–37.
Wunder, A., Engel, S. & Pagiola, S. (2008), ‘Taking stock: a comparative analysis of payments for environ-
mental services programs in developed and developing countries’, Ecological Economics 65, 834–852.
21
Figure 1: Centroid points for each property surveyed
Table 1: Household characteristics
Mean Beneficiaries Mean non-beneficiaries DifferenceHousehold size 4.884 4.578 -0.306∗
Wealth Index 2007 -0.109 0.054 0.163Speaks indigenous language 0.479 0.484 0.005Time to nearest market (min) 72.642 68.396 -4.247Knows how to read and write 0.815 0.793 -0.022Male head of the household 0.802 0.802 -0.001Age of head of the household 47.649 49.355 1.706No education 0.177 0.169 -0.008High education 0.255 0.209 -0.046Farm employment 2007 0.634 0.633 -0.001Off-farm employment 2007 0.215 0.181 -0.035Number of rooms in house 2007 1.888 1.955 0.067Had electricity in the house 2007 0.772 0.796 0.024Observations 567 519
* p<0.10 ** p<0.05 *** p<0.01
22