cooperation makes it happen? a groundwater economic artefactual experiment
TRANSCRIPT
Cooperation makes it happen? A groundwater
economic artefactual experiment
Rodrigo Salcedo
Penn State University
Miguel Angel Gutierrez
Universidad Autonoma de Aguascalientes
This version: March 25, 2014
Abstract
We conduct economic framed experiments in order to analyze the effect of access to efficient
technology and group arrangements on groundwater management decisions. A groundwater game
was conducted with 256 farmers selected from different regions of the state of Aguascalientes,
Mexico, and with 200 undergraduate students. Significant differences were found on the effects
of the treatments between the two populations: group arrangements show an important positive
effect in the laboratory and a lesser effect on the field, whereas adoption of efficient technology
shows significantly positive effects on the field but not on the laboratory. This last result suggests
the presence of strategic behavior on technology adoption at the laboratory experiments, who do
not possess background information about irrigation technologies.
Key Words: Common pool resources; groundwater management; artefactual field experiment;
economic laboratory experiment; strategic behavior.
This work was carried out with the aid of a grant from the Latin American and
Caribbean Environmental Economics Program (LACEEP)
1
Cooperation makes it happen? A groundwater economic
artefactual experiment
1 Introduction
Economists and political scientists have devoted a great amount of effort to understand how common
pool resources (CPR) are managed, as well as the characteristics of institutions that emerge in order
to deal with the use and distribution of CPR see (McGinnis, 2000). CPR can be defined as goods
that, because of their natural characteristics, exclusion is difficult but agents can appropriate part
of it, and that part of the resource is no longer available for another person’s use (Ostrom and
Gardner, 1993; McGinnis, 2000). Because of the peculiarities of this kind of resources, traditional
microeconomic theory predicts that the rent-seeking behavior of individuals yields overexploitation
and further depredation of the good, as explained by the “Tragedy of the Commons” presented by
Hardin (1968). Besides the traditional solutions to the problem (clear property rights and government
intervention), recent scholars have developed research that analyzes the importance of collective action
and the role of the community on the management of CPR (Coward, 1976; Ostrom, 1986, 1990, 1992;
Ostrom and Gardner, 1993; Trawick, 2003). These scholars argue that rules that are internally created
and agreed by the community, along with a set of tools that ensure the enforcement of those rules,
are more effective in the provision and preservation of CPR (Ostrom, 1990; Ostrom and Gardner,
1993; Dayton-Johnson, 2000; Trawick, 2003; Swallow et al., 2006). These studies have found some
evidence that cooperation among the users of a CPR can emerge, and that face-to-face communication
between the agents is important to ensure compliance of the rules (Hackett et al., 1994; Cardenas,
2011; Moreno-Sanchez and Maldonado, 2010). On this regard, the characteristics of the system and
the way how agents interact within the system is important for the success of the community as an
institution that ensures the sustainable and responsible exploitation of the resource. Moreover, the
dynamics of the system and the way how CPR users make decisions in this setup is important. Then,
it is necessary to understand the conditions working for and against sustainability of local cooperation
in situations of general social and economic interdependence (Bardhan, 2000).
An efficient use of the resource could also be achieved with the use of efficient technology.
However, in the case of CPR, private investment in efficient technology could yield benefits to all
users, and some users will have incentives to free-ride or a war of attrition might be triggered,
preventing all users to invest in the efficient technology.
This research conducts economic framed experiments with both farmers from the State of
Aguascalientes, Mexico, and undergraduate students in order to analyze the performance of both
efficient technology and group arrangements in the improvement of the use of water and endorsement
of appropriation levels that yields a socially-desirable economic outcome. The document is organized
as follows: In the next section, we discuss current models related to groundwater management. Then,
in Section 3, we briefly mention the nature of the problems related to water in Aguascalientes, Mexico.
2
In Section 4 we present the groundwater game and parametrization of the model, and in Section 5, we
explain the experimental design and discuss the structure of the sessions conducted on both the field
and the lab. Section 6 presents the analysis, including the treatment effects and analysis of adoption
and agreement compliance, and Section 7 discusses the results and concludes.
2 Prior research: Groundwater allocation and CPR
Groundwater management was not analyzed as a problem in Economics until the seminal work of
Burt and Brown and Deacon (Burt, 1964, 1967, 1970; Brown and Deacon, 1972). These authors apply
individual privately owned non-renewable natural resource models to solve the centralized controlled
problem, in which inefficiencies from externalities in water extraction are fully internalized by users.
Later, economists started to treat the groundwater situation in a competitive (no control) setup.
Under this approach, groundwater users’ behavior is myopic, where no consideration of the “use value”
of water is taken into account, and the current marginal value of water is equalized to the current
marginal cost of water extraction (see Gisser and Sanchez (1980); Nieswiadomy (1985); Worthington
et al. (1985)). Most of these studies conclude that the optimal path in the controlled and the
competitive situations is the same, implying that there are no welfare gains from a controlled optimal
allocation of groundwater. This result is usually referred as the Gisser-Sanchez effect. However, as
pointed out by Koundouri (2004a,b), there are reasons to believe that the conclusions of Gisser and
Sanchez should be taken with caution, since several assumptions about the nature of the aquifer,
functional forms and heterogeneity of agents are made. When these assumptions are relaxed, the
features of the Gisser-Sanchez effect do not necessarily hold.
Later, economists began using game-theoretic features to develop groundwater analytical mod-
els. The major focus of this group of studies was to analyze the behavior and interactions among
water users given the strategic behavior that might arise in different situations, as well as the welfare
losses due to strategic behavior. These studies usually focus on the sources of inefficiency based on
three types of externalities generated by the appropriation of the resource (Gardner et al., 1990): i)
Stock (Cost) Externalities, that arise when changes in the stock of the resource affect the cost of
extraction of the resource to all users; ii) Strategic Externalities, related to the common-property
feature of the resource and the difficulty of property rights allocation, which might encourage users to
extract more than optimal level of the resource because of fear of appropriation of the scarce resource
from other users in the future (Negri, 1989); and iii) Congestion Externalities, related to the spatial
distribution of the points of extraction of the resource. Provencher and Burt (1993) also identifies, iv)
Risk Externalities, that arise when the uncertainty in the availability of surface water is considered,
which increases the optimal use value of groundwater for all firms, but firms fail to internalize this
value. Gardner et al. (1990) also considers “technological externalities”, that arise when the presence
of a new technology adopted by one group of users affects the extraction costs of those that did not
adopted the technology.1.
Considering this last type of externality, some researchers have focused on the analysis of
1For this study, we will not consider risk and congestion externalites, given that none of the parameters consideredin the experiments are stochastic, nor spatial or network effects are introduced.
3
investment decisions. Theory suggests that, when decisions are taken individually, each agent will
decide the optimal moment of investment, especially when investment is irreversible and is made
only once in lifetime. Some researchers have analyzed investment in efficient irrigation technologies
under uncertainty, and focus on the analysis of the option value of investment (Barham et al., 1998;
Carey and Zilberman, 2002). However, when there is interaction between users (e.g. aquifer problem),
strategic behavior might arise due to technological externalities. For instance, Agaarwal and Narayan
(2004) develop a dynamic two-stage game in which agents choose the level of initial investment for
the capacity of a well and subsequent extraction. They show that, due to strategic behavior in
investment, agents invest in excess capacity, allowing the overexploitation of the aquifer. Barham et
al. (1998) also analyze the relationship between sunk costs and strategic behavior but in a context of
a non-renewable resource.
In the problem presented in this study, the adoption of efficient irrigation technologies should
help to increase the water table over time, reducing costs of extraction. Thus, some groundwater
users could have the incentive to delay adoption given that they are already being benefited by those
farmers who already adopted. The problem presented in this study is similar to the one presented in
Moretto (2000) and Dosi and Moretto (2010), without considering uncertainty in returns. Moretto
(2000) develops a theory where irreversibility effects and war of attrition effects are compounded in
the decision of producers to adopt a new technology. They find switching trigger values at which
producers will switch from one technology to another.
3 Water problems in Aguascalientes
The State of Aguascalientes is located in a semi-arid area in Central Mexico. Water from rain is only
available during the rainy season between May and September with an annual rainfall of 500mm/year.
In 2007, 50,542 has. were irrigated, which represent 30 percent of the total agricultural area in the
state, and is also the most productive land (INEGI, 2009). From the irrigated agricultural areas, 36
percent receives water from dams and 67 percent is irrigated with groundwater2. The main problem in
Aguascalientes is the overexploitation of the aquifer Ojocaliente-Aguascalientes-Encarnacin (OAE),
which is currently the main source of water of the state. The estimated annual net use of groundwater
in the watershed of Aguascalientes, which is the most important agricultural area in the state and
also hosts the capital city - the city of Aguascalientes- is 433 million m3/year, from which 71 percent
of this volume is used in agriculture, 22 percent in urban purposes, 1.6 percent is used for industrial
purposes, 5.7 percent is devoted to other uses and natural losses (COTAS, 2006). However, the
estimated annual recharge of the aquifer (natural recharge and filtrations from dams and superficial
irrigation) is 234 million m3/year. Therefore, there is a deficit in the use of groundwater of 199
million m3/year which is not replenished, and the ratio of extracted water/recharged water is 1.9,
meaning that the water that is consumed is almost twice the water that is recovered (COTAS, 2006).
The problem is evident when looking at the depth at which water is extracted. As shown in Figure
1 the average depth-to-water increased from an average of 33 m. in 1965 to 87 m. in 1985 to 145 m.
in 2005 (Gobierno del Estado de Aguascalientes, 2009). Also, currently there are 3,285 wells from
2Percentages do not add because many farmers receive water from both sources
4
which 1,539 are for agricultural purposes (privately and collectively owned). Moreover, the supply
of water per capita is 281.6 m3/year, which is considered extremely low for international standards
(Gobierno del Estado de Aguascalientes, 2009).
A potential solution to the overexploitation of the aquifer is the installation of efficient irrigation
systems on the parcel. However, as noted by several authors, the impact of the adoption of these
systems on the reduction of the depletion of the aquifer per se is ambiguous. Moreover, there is some
evidence that the farmers of the area of analysis are reluctant to the adoption of these technologies
(Caldera, 2009). Also, the government is developing capacity building programs with the Water User
Associations (WUA) of the region. Given this setting, it is crucial for the authorities and researchers of
Aguascalientes to investigate about the attitude of farmers in Aguascalientes towards the use of water
for agriculture and the adoption of more efficient technologies, not only as mechanisms that might
improve private yields, but also as a water saving method; and to analyze the role of organization and
internal agreements among users in the improvement of groundwater management in Aguascalientes.
4 The Groundwater Management Game
In this section we briefly present the theoretical model that has been used for the experimental
design. We closely follow the model presented in Provencher and Burt (1993). In order to facilitate
the presentation of the model, we will first show the individual model of groundwater management
with no investment and analyze the solutions for three different types of behavior: Myopic, Rational
and Fully Coordinated. Then, we will present the individual model for both groundwater consumption
and optimal investment.
The functional forms that we have chosen for the experimental design follow Wang and Segarra
(2011). These authors consider a profit function that is linear on the demand of water. They argue
that crop water-related yields tend to increase linearly with the amount of water applied until they
reach a plateau, due to the natural capacity of plants to absorb the water. We also adapted the model
presented in Provencher and Burt (1993) to consider dept-to-water instead of the stock of water that
remains in the aquifer. We believe that groundwater users are more closely related to the concept of
water table and depth rather than the stock of water, since it is unknown.
4.1 Basic game (Baseline)
There are N farmers that use groundwater for irrigation. They pump water from a bathtub-type
aquifer through water wells, and they do not have access to surface water. In our model, we will only
consider the costs related to changes in the water table, not those related to well digging 3.
Farmer benefits from agriculture are denoted by:
Bit = αwith− witβ
St− k = wit
(αh− β
St
)− k
3Although we believe that these decisions and costs are extremely important, we tried to simplify the model in orderto make the application with farmers easier.
5
Where wit is the amount of water pumped by farmer i in period t, α is the marginal value of
production of irrigated land, h (0 < h ≤ 1) is the level of efficiency of the irrigation system, βSt
is
the marginal cost of water pumped from the well, St is the stock of water in the aquifer in period
t, and k is the fixed cost of production. As commonly assumed in the groundwater literature, only
production costs related to groundwater are variable. We assume that farmers already made all the
decisions regarding the use of other inputs, and their cost is already considered in the marginal cost β
and the fixed cost k. Note that the marginal cost of water pumped is inversely related to the stock of
water in the aquifer. If the stock of water increases, then the water table increases and farmers have
to pump water from a higher point in the well, which requires less electricity and reduces pumping
costs. Finally, there is no heterogeneity between farmers, so we will consider the symmetric case.
We believe that the concept of stock of water was less understood by participants than depth
of water extraction. Thus, we changed the cost function using the following transformation. We
considered the aquifer as a cylinder with a radius of R and height of D. Then, the maximum capacity
of the aquifer is denoted by the D×πR2. However, the stock of water of the aquifer will change over
time due to exploitation. Then, the stock can be observed using the difference between the maximum
depth of the aquifer, D, and the depth at which water is pumped, dt. The stock of water at time t
can be represented by: St =(D − dt
)×πR2. With this identity, we can transform our current-period
profit function to:
Bit = wit
(αh− β(
D − dt)× πR2
)− k (1)
The equation of motion of the water table is represented by the following equation:
dt+1 = dt +Wt
πR2− f
πR2
Where Wt is the demand of water of the N farms, Wt =∑N
i=1wit, and f is the natural recharge
of the aquifer in period t.
We also need a boundary constraint for the total water used, since no water beyond D can be
pumped. This is represented by:
dt +Wt
πR2≤ D
Farmers will make water demand decisions depending on the behavior assumed. We will analyze
the cases when all farmers are myopic, rational and fully cooperative.
Given the functional form used in this model, a myopic water user will pump the maximum
possible amount of water every period, as long as profits are positive, since myopic water users do
not internalize the future consequences of their water consumption today. Moreover, there are no
contemporaneous externalities from other water users in our model. Then, the demand of a myopic
6
user is represented by:
wm =
{w if dt ≤ D − β
πR2αh
0 otherwise
Where w is the maximum amount of water that can be pumped.
On the other hand, a rational water user will consider the future value of water in their decisions.
She will also consider the behavior of other groundwater users, since their decisions will affect the
stock of water in the future, and therefore the welfare of all users. Thus, the problem that the rational
water user solves is:
Vi1 = max{wit}Tt=1
[T∑t=1
δt−1Bit
]s.t
dt+1 = dt +1
πR2
N∑j=1
wjt + f
dt +
1
πR2
N∑j=1
wjt ≤ D
d1 given
Where δ is the discount rate. The experiments developed in this study do not consider het-
erogenous agents, so we can assume, for the theoretical model, that all users are the same and that
agent i can identify with certainty the other users’ choices, and therefore user i maximizes its current
value function given the other users’ best response4. Recalling the optimality principle, we can write
the Bellman equation for agent i and period t as:
Vit (dt) = maxwit
[wit
(αh− β(
D − dt)× πR2
)− k + δVi,t+1 (dt+1)
]s.t.
dt+1 = dt +1
πR2[wit + (N − 1)φ (dt) + f ]
dt +1
πR2[wit + (N − 1)φ (dt)] ≤ D
d1 given
Where φ (d) is the best decision taken by the other N − 1 firms and, as before, h < 1. The
solution of the problem will depend on the type of strategies that agents choose. Open-loop strategies
only depend on time and not on the current state (dt), given that agents take the initial information
and define an optimal path of extraction, whereas feedback or close-loop strategies will depend on
both time and current state. Negri (1989) and other authors argue that open-loop strategies are not
4Nevertheless, for empirical purposes, conjectures about other users’ decisions should vary between individuals,depending on observable and unobservable variables.
7
consistent, given that agents can correct their paths over time. Provencher and Burt (1993) show
that the solution of this problem involves two different types of externalities: strategic and stock.
Strategic externalities arise when feedback strategies are followed. These externalities negatively affect
the efficient allocation of the resource and could encourage the depletion of the resource.
A controlled solution of the problem differs from the individual problem in that externalities
are fully internalized. Assuming homogeneity of agents, the symmetric problem can be represented
as:
NVit (dt) = maxwit
N
[wit
(αh− β(
D − dt)× πR2
)− k + δVi,t+1 (dt+1)
]s.t.
dt+1 = dt +1
πR2
[wit + (N − 1) φ (dt) + f
]dt +
1
πR2
[wit + (N − 1) φ (dt)
]≤ D
d1 given
Where ∼ denotes values at the social optimum. If concavity of V and V is guaranteed,
Provencher and Burt (1993) show that the individual demand of water is greater than the socially
optimal demand and it leads to a lower steady-state equilibrium. With a proper parametrization, it
is possible to solve both problems numerically (see Section 4.3).
4.2 Investment
Now, we will consider the situation where farmers are allowed to invest in new irrigation technology.
This new technology has a level of irrigation efficiency h = 1. Thus, less water is required to achieve
the same production levels as with the old technology. This technology is adopted only once in
lifetime. However, the cost of adopting the new technology is I. Also, with the new technology, it is
necessary to spend every period an additional maintenance cost of m5. The period benefits with the
default technology are denoted as B0 and are represented by equation (1), whereas benefits with the
more efficient technology, without considering the investment cost, can be represented by:
B1it = wit
(α− β(
D − dt)× πR2
)− k −m (2)
With the two technologies available, the farmer not only has to decide the optimal path of
pumped water, but also the optimal moment at which he/she switches to the efficient technology.
These two situations can be combined as follows. In any period, say τ , the farmer will choose whether
to invest or not in the technology, in order to maximize his/her present value of the utility, Viτ , taking
5This additional cost can be interpreted as the cost incurred in hose and filter replacement for drip irrigationtechnology.
8
into account the equation of motion of the stock of water in the aquifer and the boundaries of wit:
Viτ = max{V 0iτ , V
1iτ − I
}s.t.
dτ+1 = dτ +1
πR2[wiτ + (N − 1)φ (dτ ) + f ]
dτ +1
πR2[wiτ + (N − 1)φ (dτ )] ≤ D
d1 given
Where
V 0iτ = B0
iτ + δVi,τ+1
and
V 1iτ =
T∑t=τ
δt−1B1it
Again, φ (·) denotes the best strategy of the other users. Note that the present value of the
utility of not investing in τ , V 0iτ , considers the possibility of investing in the new technology in the
next period τ + 1, whereas investment in τ is considered irreversible.
This problem can be solved using the two-step method proposed by Agaarwal and Narayan
(2004). The first stage involves the investment decision whereas the second stage solves the optimal
extraction path {wit}Tt=1. This problem is solved backwards, starting from the second stage. We can
search for the optimal extraction path conditional on the investment timing decision t, {wit(t)}Tt=1, t =
{1, 2, ..., T}. Each t will yield a lifetime utility V(t)i1
. Then, the optimal investment time, t∗, can
be represented as:
t∗ = argmax{V(t)i1}, ∀t = {1, 2, . . . , T} (3)
Myopic agents do not care about the future, so they will not invest in the new technology. On
the other hand, rational/strategic agents might be willing to invest in the new technology if every
user is willing to invest. However, there is the possibility of free-riding : If agent i believes that the
other users will invest, he/she will benefit from their water savings. Therefore, agent i does not have
incentives to invest. If this is the case, then all the agents will have the same strategy. Thus, the
Nash equilibrium will be the resulting equilibrium. In this case, two equilibria might arise, one in
which everyone invests in the first period, and the other, in which no one invests, depending on the
conjectures that agents make about other agents’ actions. Finally, at the social optimum equilibrium,
everyone invests in the first period.
9
4.3 Parametrization: myopic, rational and optimal behavior
The total number of water users for each well was N = 4. Also, in order to facilitate the decision-
making process of participants, we discretized the number of hours of water that they can use. Then,
for this experiment, wit = {0, 1, . . . , 10}. Also, we allow the efficiency level of the irrigation technology
to be h = 0.5 with the less efficient technology, and h = 1 with the highly-efficient one. In that sense,
it will be required 10 units of water to irrigate all the land with the less efficient technology, but only
5 units with the more efficient one (see Section 5.2).
The life span of participants will be of 5 production years. After year 5, they retire from farming
and receive their reward (Vi1). Thus, it is not important for them if there is water left in the aquifer6.
The initial depth of extraction is set to d0 = 170. The unit of length for the field experiments was
meters and for the lab experiments was yards7.
All the parameters that were used in the experiments are presented in Table 1. Also, tables
2, 3 and 4 present the revenues with the less-efficient technology, highly-efficient technology and the
costs of extraction. Note that revenues with the highly-efficient technology increase with the number
of units used up to five units. After five units, revenues do not change. This is because farmers posses
a fixed agricultural area and, with the parameters used, all the area that they hold is irrigated with
5 units (see Section 5.2). Also, note that the cost changes with each level of the depth of extraction
and the number of water units required.
The trajectories of units of water pumped, costs and well depth for each type of equilibrium
are presented in figures 2, 3 and 4. We also calculated the total benefits from possible “fixed arrange-
ments” in the number of water units. This exercise could represent the situation in which there is
an agreement between users. We calculated the total benefits with arrangements from 0 to 10 hours
of pumping. Nevertheless, we considered zero hours of pumping whenever the cost of extraction was
too high and no production was more profitable. Also, we consider 10 hours for the last period, since
the final well depth does not affect participant’s profits. The results are presented in figure 5. The
equilibrium total benefits without including the initial capital8 for myopic agents is 76.25; 108.75 for
rational/strategic agents; and 165.82 for fully coordinated agents. Also, we can see that the value
function is not linear for different fixed arrangements. Moreover, with arrangements of 3, 4, 5, 6 and
7 units for each period, it is possible to achieve net benefits higher than the Nash equilibrium. This
situation suggests that, in this experiment, it is possible to be better-off than the Nash equilibrium
if water users comply with those fixed arrangements.
In order to solve the model for the situation in which the efficient technology is available, we
first solve the same problem in the case when only the highly-efficient technology is available. We
set the parameters in a way that it is possible to achieve the same social optimum with the two
technologies. This was done in order to avoid the introduction of a bias from the parameters chosen.
Then, we solved the second phase that involves the investment decision, allowing the model to switch
between the two technologies, given the efficient path. As mentioned before, the model shows two
6Although we believe that it is important to give a value to the water that remains in the aquifer, it would requiredthe valuation of the ecological benefits of that stock, and this valuation goes beyond the scope of this paper.
7We decided to use yards instead of feet in order to keep the same parametrization, since the idea of distance of ayard is similar to the one of a meter (1 yard = 0.9144 meters).
8See next section for details
10
equilibria: one in which all users adopt the technology in the first period, and another in which none
of the users adopt the technology.
5 Experimental design
5.1 Recruitment of participants
For the artefactual field experiments, farmers from the State of Aguascalientes, Mexico were recruited,
and the information of 256 farmers will be used in this study9. Farmers were contacted in several
ways. The most efficient way to recruit farmers was through delegates of the sections of the Irrigation
Distric 001 and engineers that worked on the region. We also contacted farmers through professors
of the Univerisdad de Aguascalientes10. 52 farmers participated in 4 sessions between April and May
of 2012, whereas in the period between September and October of 2012, we conducted 21 sessions
with 204 farmers. The sessions were conducted in rooms provided by the Comisario Ejidal or by the
Irrigation District. Only in three occasions, we conducted the sessions on the field, where the water
well was located. Farmers use water for irrigation purposes11, and we allow the source of water to
be a water well or a dam. In many cases (37.89 percent), farmers had the two sources. As we will
show in the next section, the game that will be played is based on a situation in which farmers pump
water from a well. However, we allowed the participation of farmers that do not have a well because
these farmers are potential users of a well (given the declining rain in the region, farmers are getting
more permissions for well digging), and there might be differences in the behavior between these two
groups. Also, all the farmers were very familiar with the context.
For the laboratory experiments, Penn State undergraduate students were recruited through
the Laboratory for Economics Management Auctions (LEMA) recruitment system. A total of 200
students were recruited, and participants could attend the experiment only once. The sessions were
developed in two periods: between June 25 and July 15, and between August 30 and September 17
of 2013. All the sessions were conducted at LEMA at the Smeal College of Business of Penn State.
5.2 Framing
We framed the experiment as follows. Each participant holds H acres of land devoted to agriculture.
This land could be irrigated or not, depending on water users decisions. The amount of land is fixed
over time throughout the game and it is the same for all participants. Participants are water users
that extract water from a well. There are N water users in each well. Water users are homogenous
but they do not know for sure what the other users do (see below for a description of the session).
Each period will represent one year of production, and water users will decide the amount of
water they will use for the entire production year. Farmers’ decisions on the amount of water will
9We gathered information of 299 farmers, however the information of the first three sessions, which included 32farmers, was used as a pilot due to several changes in the experimental design. In addition, the information of onesession with 12 farmers is not included because they did not finish the session
10We also attempted to contact farmers in the Feria Nacional Agropecuaria de San Marcos, one of the most importantagricultural fairs in Mexico. The Universidad de Aguascalientes allowed us to get a kiosk in which we could developedthe session. However, this strategy did not work.
11Only two farmers had rain-fed land
11
be made in terms of hours of water/day pumped from the well, wit. Each irrigated acre requires 12h
hours of water/day from the well, where h is the level of efficiency of the technology that the farmer
has in place (0 < h ≤ 1). For instance, in order to produce on the H acres, farmers have to use
wit = H2h hours of water a day. However, they do not have to work in the whole H acres; they can
decide to irrigate H < H acres. Thus, the total irrigated land for the period will depend on both the
water requirements that they scheduled and the level of efficiency, wit2h.
As mentioned above, changes in the depth of the water well will depend on the amount of water
that farmers use. For the sake of consistency of information, we told participants that one hour of
water pumped from the well will contribute to the depth of the well in one meter/yard . Finally, each
farmer has an initial capital of C.
5.3 Treatments and sessions
Treatments
The experimental design consists of two treatments:
Investment in technology (IN): In this treatment, farmers will be allowed to invest in irrigation
technology. Farmers are free to choose the time in which the investment is done. The cost of
investment is I.
Internal agreement (AG): Before participants make any decision regarding water consumption, they
agree on a fixed amount of water that they should pump in each period. After participants make
their agreements, they individually and anonymously decide whether they comply with or deviate
from the agreement.
We called the “Counterfactual” treatment (CF) to those sessions when the two treatments are
deactivated. Table 5 presents the distribution of sessions and participants among the treatments for
both field and laboratory experiments.
Sessions
Games were played by groups of 4 participants. Each group represented a water well and the
four members of the group had to pump water from the same well. We named the wells with colors:
yellow, blue, orange, green and red. The number of wells that participated in the session varied
depending upon the number of attendants. Each participant received a card of the color of the group
membership. The card was randomly assigned, but we made sure that other participants with the
same color did not sit nearby. Also, each well played a game independently. In other words, wells did
not compete for water and there was no relationship between them. Finally, all experiments (both
field and laboratory) were paper-based.
For the case of the field experiments, we did not run sessions with more than 20 participants
to guarantee tractability. Also, we did not allow members of each well to sit together in order to
avoid collusion. Therefore, when possible, we arranged individual tables in the session, and members
of the same group sat in each table separately and in different parts of the room. However, the most
common setup was one in which we arranged four big tables and one member of each group sat in each
12
table. Depending on the number of groups, one to five people could sit in the same table. Although
members of different wells sat together, they did not take actions upon the same information, given
that wells are independent12.
For the case of the lab experiments, all the sessions had at most 12 participants. Also, partic-
ipants sat on individual cubicles and we did not allow members of the same well to sit next to each
other.
Participants of the field experiments played for 10 periods, and those of the laboratory experi-
ments for 15 periods. The first 5 periods (Round 1) corresponded to the baseline situation, in which
no investment nor agreement treatments were in place. After the first 5 periods, the following 5
periods (Round 2) corresponded to a specific treatment. We applied 3 types of sessions in the ex-
periments. The type of session (CF,IN,AG) and corresponding instructions were informed after the
end of the first stage. For the laboratory experiment, after Round 2, all the groups were randomly
reshuffled and the same game as in Round 2 was conducted.13.
Before participants make their pumping decisions for the first round, the initial well depth is
informed to all the members of the group. This information is public along the game and is the same
for all the water wells.
Each participant had a revenue table for the case of the less-efficient technology and, only for
the case when the IN treatment is in place, participants had the revenues of the two technologies, as
shown in tables 2 and 3. They also had the cost table presented in Table 4. Costs do not change
between treatments. With this information, participants can decide how many hours they pump.
They write down all their decisions for the first period of Round 1 in their account sheet. Once all
participants made their decisions, calculated their income, cost and benefits for the first period (here
the facilitator and assistants helped with these tasks), the facilitator proceeded to count how many
hours were used by each participant for each well. This exercise was done in a way that the other
members of the well did not notice who their group partners are. Then, we calculate the change in
the depth of each well (considering recharge) and make this information public to proceed with the
second period. This exercise is repeated four more periods and each time we update the depth of the
well.
For Round 2, we started again from the initial well depth (170 m./yards). If the type of session
was “IN”, participants will be able to invest in a highly-efficient technology. This is informed to all
participants before the round starts. However, there is a cost of I to acquire the new technology.
This payment is done only once during the game. Thus, every period, they have to decide whether to
invest or not in the technology and the amount of water to pump. Once the technology is adopted,
they will only calculate their benefits using the table of the highly-efficient technology, shown in Table
3.
If the type of session is “AG”, the members of each well meet for 10 minutes before the game
starts and discuss a fixed amount of water that they will request for each period throughout the
game. This agreement will be written in their accounting sheet. After the meeting, we proceed to
12However, it might be possible to find some correlation between groups. It is possible to control for that in theanalysis.
13There is a variation in application of the CF session in Round 3, but we will not use that round for the CF in thisstudy.
13
play a game similar to the baseline. Participants can anonymously decide whether to comply with
the agreement or deviate, and no penalties are applied.
5.4 Rewards
As mentioned above, each participant receives an initial endowment C which corresponds to their
show up fee. Then, the payoffs at the end of each stage will be composed by C plus the total
net benefits from production that the participant earned through the five periods minus I if the
participant adopted the technology in the second (and third round for lab experiments) of the “IN”
sessions . At the end of the session, participants of the field experiment tossed a coin marked with
“1” and “2”. If the coin shows “1”, he/she will receive the amount earned in stage 1, otherwise, he
receives the amount earned in stage 2. For the lab experiment, each participant drew a ball from a
bag with six balls, two of them marked with “1”, “2” or “3”, which correspond to the reward of each
of the rounds.
Participants of the field experiments received a show up fee of 200 MX Pesos (US$16) for their
attendance, and the total rewards that participants earned are between 200 and 441 MX Pesos. For
the lab experiments, participants received a show up fee of $10 and the total earnings that participants
got are between $10 and $2214.
5.5 Survey
We conducted a survey at the beginning of each session. The purpose of the survey is to complement
the data gathered in the experiments and analyze whether any individual characteristics determine
the behavior and attitude of participants in the experiment. For the field experiment, the survey
consisted on five sections and we asked about farmers’ land tenure, major crops and livestock, farmers’
experience and demographics, water access and use, and irrigation technology available on the field.
The facilitator read each question of the survey and each participant answered individually with the
help of assistants. Participants took between 20 and 30 minutes to fill the survey. For the laboratory
experiments, we asked about age, gender, major, minor, whether participants have taken courses in
Economics, Finances, Business Management, Psychology and Hydrology, and whether participants
had a farming background. We also included the Cognitive Reflection Test (CRT) proposed by
Frederick (2005)
6 Analysis of data and results
We conducted 25 sessions with a total number of 256 farmers in the field experiments, and 27 sessions
with a total number of 200 students in the lab experiments. The summary of sessions is presented in
Table 5.
14Some participants earned total benefits lower than the show up fee, however, we set the minimum amount to200MX/$10.
14
6.1 Experiment Outcomes
The main outcomes of the experiments are the number of pumping hours, benefits and well depth in
each period; and the total hours pumped, total net benefits, and final well depth.
Tables 6 and 7 show the average values in each treatment, round and period for the number of
hours, benefits and well depth for both field and lab experiments. Figures 6, 7 and 8 depict a graphical
version of these tables. Several comments can be derived from these figures. First, in Round 1, the
resulting paths of the field experiment are very different to those of the laboratory experiment. The
average path of pumping hours in the field experiment decreases in a smoother way than in the lab
experiments, and similar differences are observed on the average benefits and well depth. Also, we
can observe important differences between the three treated groups of the field experiments in Round
1, when the treatment has not been applied yet, whereas there are no significant differences between
the three treated groups of the lab experiments for the same round. These significant differences are
also shown in tables 6 and 7. Further, the average paths in Round 1 in both field and lab experiments
are very different to the theoretical outcomes presented in figures 2, 3 and 4.
In Round 2, where the treatments are applied, we can see important differences between the
three groups. For the field experiments, only the “IN” treatment shows an improvement in the three
variables analyzed with respect the counterfactual; whereas for the lab experiments, both the “IN”
and the “AG” treatments show significant improvements with respect to the counterfactual in most of
the periods. Finally, it is worth to note that the average path of hours of pumping of the counterfactual
group of the lab experiments is very similar to the myopic theoretical behavior presented in figure
2. This finding suggests that, after the first round, most participants notice the behavior of their
partners and decided to show this kind of behavior in opposition to a more cooperative one. This
result is not observed in the field experiments.
Turning to analyze individual decisions, Figures 10 and 9 show histograms of pumping hours
for each period, round and treatment from the field and laboratory experiments, respectively. We
can notice several things from those figures. First, the dispersion of choices from the field experiment
is much higher than from the laboratory experiment, which might reflect different degrees of game
comprehension. However, when looking at the decisions of the field experiment in the first period
of the first round, we can see that decisions are concentrated around 5 and 10 units, whereas in
the laboratory experiment they are mostly concentrated around 10 units, and a lesser proportion is
around 5 units. After period one, the distribution becomes flatter in the field experiment, suggesting
that farmers consider the potential cost of using high volumes of the resource in early periods. In
contrast, in the laboratory experiment, the concentration around 10 units increases until period 4,
where the distribution splits in two parts, one around zero units and another around ten units. This
is because it is not profitable for some groups to use more water, given the depth of the well.
The differences between the decisions of students and farmers might reflect structural differences
in behavior. Farmers appear to be more pro-social than students, or at least they consider a higher use
value of the resource. Students tend to be more myopic and selfish, and this behavior becomes stronger
on the following rounds. These results are consistent with other findings in the Behavioral Economics
literature. For instance, Belot et al. (2010) find that students are more likely to behave in accordance
to the economic theory than non-students in games that engage other-regarding preferences (such as
15
Trust Game or Public Good Game). Carpenter et al. (2005), Carpenter et al. (2008) and Anderson
et al. (2013) find similar results.
To fully account for the effects of each treatment on the use of the resource, it is necessary to
develop a deeper analysis of the different behaviors within the two populations (see Salcedo (2014)).
For the case when the investment treatment is imposed (Round 2), the distribution collapses to
5 units in the field experiment, meaning that most farmers chose to adopt the technology, whereas in
the field experiment, the distribution splits into two parts, one concentrated in 5 units and another
in 10. This evidence confirms the results from the theory, where we found two equilibria. Finally,
when the agreement treatment is imposed, we see that the distribution of the field experiment starts
concentrated around 5 and 7 units, and then becomes more dispersed and the concentration around
8 and 9 units increases. In the laboratory experiment, there are not significant changes in the
distribution until the last period, when most of participants switched to 10 units.
We are also interested on the end-of-round outcomes of the experiment. Tables 8 and 9 show the
mean values for the three end-of-round variables for both field and laboratory experiments. Difference
tests in both mean and distribution are performed between rounds for each treatment (CF, IN and
AG) in part a. of each table. As we can see, there are significant differences for the two treatments
on the laboratory experiments, while we found significant differences only for the IN treatment in
the field experiment. We also conducted difference tests between the counterfactual and each of the
two treatments within each round (Part b.). The results of the field experiment show that there are
some differences between the counterfactual and the IN treatment in the total benefits in Round 1,
and in the total hours pumped and final well depth in Round 2. The differences in Round 1 suggest
that, even though the treatments were assigned randomly, there are some unobservable differences
between the participants of the sessions that has to be considered in the analysis.
For the case of the lab experiments, there are no differences in the outcomes in Round 1, between
the CF and the treatments, which suggests that the samples are similar a priori. On the other hand,
in Round 2, all outcomes from the AG sessions are significantly different from the CF treatment in
both mean and median, whereas only the Total Benefits from the IN sessions are significantly different
from the CF sessions in both mean and median. There are no significant differences in median for the
Total hours pumped between IN and CF, whereas for the Final Well Depth, there are no significant
difference between IN and CF in either mean nor median.
Figure 11 shows the average total net benefit and the theoretical equilibrium values for each
treatment and round. We can see that, in both the field and laboratory experiments, the three groups
overcome, on average, the value of the Myopic theoretical result (275.25), but do not reach the Nash
solution in Round 1 (308.75). In Round 2, the total net benefits of the CF group do not reach the
Nash equilibrium in the field experiments, whereas the benefits for the IN group barely reaches the
Nash level, and the average net benefits of the AG groups do not even reach those of the myopic
benchmark. In the case of the laboratory experiments, again the CF group does not reach the Nash.
However, in this case, both the IN and AG are significantly higher than the Nash. A similar pattern
is observed with total hours of pumping and final well depth (Figures 12 and 13).
16
6.2 Treatment effects
A key contribution of this study is to analyze the effect of the two treatments (access to efficient
technology and communication) on water consumption. Effects of the treatments at the period
level are necessary to analyze. Given the presence of different behaviors in the population, which
can be translated into statistical terms as the presence of mixtures of distributions, the analysis of
treatment effects considering dynamics becomes more complex and is beyond the purpose of this
study. Nevertheless, the analysis of the the treatment effects on the three end-of-stage outcomes
(individual total hours pumped, individual total benefits and final well depth) is also a major goal of
the study. We will focus now on the analysis of these outcomes and leave the analysis of the period
outcomes for future research.
As mentioned before, there are some intrinsic differences between the treated groups, even in
the first stage, where no treatment is imposed. If these differences are not taken into account in the
analysis, any estimation of the impact of the treatment will be biased. In order to obtain reliable
impacts of the treatment, we will estimate a Difference-in-difference model.
In this case we will consider a reduced form approach to analyze the treatment effects. Consider
the outcome ygis observed at the end of the round without treatment (s = 1) and the round after
treatment (s = 2) for three different treated groups: “Counterfactual” (g = 0), “Investment” (g = 1)
and “Agreement” (g = 2). Recall that for the case of g = 0, there is no treatment imposed in both
s = 1 and s = 2. We can estimate the following linear equation:
ygis = α+ δ + θ1D1 + θ2D2 + φ1D1 × δ + φ2D2 × δ + βXi + υi + εis (4)
Where α is a constant, δ is a binary variable that takes the value of one if s = 1 and zero
otherwise, D1 is a binary variable that takes the value of one if g = 1 and zero otherwise, D2 is a
binary variable that takes the value of one if g = 2 and zero otherwise, Xi are fixed individual (group)
level observable variables, υi is a specific individual (group) level random term and εis is a individual
(group) and time level error term.
We are interested in the value of φg for each treatment, g = 1, 2, given that:
φ1 = E[y1i2 − y1
i1
]− E
[y0i2 − y0
i1
]and
φ2 = E[y2i2 − y2
i1
]− E
[y0i2 − y0
i1
]Estimation of this model with OLS will yield inconsistent estimates because of the presence of
υi. We used a random-effects model to estimate the equation presented above in order to account
for specific individual and treated-group effects. We also considered the full model which includes
some of the variables gathered through the survey to control for observable characteristics of partic-
ipants. Results are presented in Table 10 for the field experiments and Table 11 for the laboratory
experiments.
The coefficients of interest are “IN x Round2” for the effect of the “IN” treatment, and “AG
x Round2” for the effect of the “AG” treatment. For the field experiments, we can see that the
17
two treatments have a significant effect on the reduction of the total pumping hours and final well
depth, whereas only the IN treatment has a positive effect on final total benefits. This could be
explained by the fact that some participants deviated from the agreement. Thus, even though the
amount of water was reduced, it did not increased the benefits. Also, we can see that there are
intrinsic differences between the total water used in the groups, since the coefficients of “IN” and
“AG” are significantly different from zero in equation (1) and (2). This result validates our difference
in difference approach. We also present the full model including contextual and individual variables.
It seems that older people, people that participate more in non-agricultural activities, and with a
larger area with drip irrigation tend to use less hours of pumping within the experiment, whereas
people with larger areas tend to use more. At the same time, people that work only off-farm and
people with drip irrigation tend to earn less benefits in the game, and people with more area tend to
earn more. These results suggest that some individual characteristics might be related to the types
of behavior that the participant show during the game.
With respect to the laboratory experiments, we can observe that the two treatments only
show an effect on the total benefits earned by participants, and only the “AG” treatment shows
effects on total hours pumped and final well depth. This is because, as we will show in the next
subsection, several participants did not adopt the technology and “free-ride”, taking advantage of the
benefits generated by the adoption of technology by other participants. We also included individual
characteristics in the regressions, but none of them are significant.
6.3 Technology adoption
The results presented in the previous section suggested that participants of the field experiment show
higher rates of adoption than the laboratory experiment, since the “IN” treatment did not have any
effect on the latter, but an important effect on the former. As mentioned before, this might suggest
the presence of strategic behavior in adoption for some participants of the laboratory experiments.
Figure 14 shows the cumulative technology adoption percentages in each period of Round 2
for both field and laboratory experiments. We can see that 81 percent of participants of the field
experiments adopted the technology in the first period of Round 2, and then increased up to 90 percent
in the fifth period. In contrast, roughly 40 percent of participants of the laboratory experiments
adopted the new technology in the first period, and then slowly increased up to 60 percent in the fifth
period. This means that 40 percent of participants never adopted the new technology. Recall that
the theoretical model shows that there are two equilibria for the investment problem, one in which
all adopted the technology in the first period, and one in which no one adopted the technology. The
empirical findings suggest that the theoretical predictions hold in the laboratory, since the sample is
divided into people that adopted immediately and people that never adopted. This is also observed
on the third round of the laboratory experiment. Although in Round 3, most participants adopted
the new technology on the first period, the adoption rate slightly improved from 62 to 66 percent, as
shown in Figure 15.
We can analyze how adoption rates change with other factors, such as the current well depth
or personal characteristics of participants, using a proportional hazard model. Results from the Cox
proportional hazard model of adoption for Round 2 of the laboratory experiment are presented in
18
Table 12.
We can see that the effects of the two time-varying variables, well depth and well depth squared,
are significantly different from zero. As expected, well depth increases the hazard rate in 275 percent
whereas the squared of the well depth decreases the hazard ratio in 0.4 percent. This result shows that
the well depth has an inverted “U” shape effect on the hazard ratio of adopting the technology. With
respect to the individual characteristics, to be enrolled in the Information Sciences major increases
the hazard rate in 248 percent with respect to the Business major. Finally, differences in the results
of the CRT test also affects the hazard rate of adopting. Having a score of zero, one and two in
the CRT test decreases the hazard rate of adopting the technology in 65.6, 58.4 and 48 percent with
respect to the effect of having full score (three), respectively. This last results is interesting, since the
CRT is correlated with both patient and less risk averse people (Frederick, 2005). Thus, people with
higher CRT tend to be more thoughtful about the decision of investing, or is less risk averse and is
willing to bear the risk of investing, even if they know that other participants might free-ride.
Turning to the field experiment, we will not analyze the rate of adoption with a hazard model
since there is not enough variation on the rates. A possible explanation of the high adoption rate is the
background information that farmers hold about efficient irrigation technologies. There are several
institutions in the area that have extension programs and demonstration plots with which information
about efficient irrigation technologies is given to farmers. Also, as mentioned before, the Government
of the State of Aguascalientes is conducting several programs that aim at the improvement of water
use efficiency through the adoption of better irrigation technologies. A major constraint for adoption
is the lack of funding and budget constraints that farmers face. Only richer farmers can adopt the
technology, but most farmers are willing to adopt. In contrast, college students do not posses any
prior information about irrigation technologies, and they will only use the information in the game to
make their decisions. They will not adopt if they believe that it is not worth to adopt the technology
and, moreover, if their behavior is such that they believe that they can benefit from not adopting.
This makes an important difference between field and laboratory experiments that should be taken
into account. If the the experiment is meant to serve as a benchmark of the behavior of people on
a policy that might be applied, then people with different backgrounds will see the experiment in
different ways, and therefore will show different behavior.
6.4 Agreement
It is also worth to analyze the types of arrangements that participants make in the AG treatment, as
well as whether they deviate or not from the agreement. As mentioned above, participants from the
AG sessions were asked to meet for ten minutes before Round 2 starts in order to agree on a fixed
individual level of pumping throughout the five periods. However, no penalties are applied if the
participant deviates from the agreement. We expect that some participants will deviate and others
will comply with the agreement.
Figure 16 shows histograms with the agreed values of pumping hours. Panel a. shows the
agreements in the field experiment in Round 2. We can see that most of the agreements are of 5 and
7 units, and no group chose less than 5 hours. It is also worth to note that a significant number of
wells agreed on 10 units.
19
With respect to the laboratory experiments (Panel b.), more than 60 percent of the groups
agreed on a value of 5 hours, and 22 percent agreed on 7 hours in Round 2. Also, there are two groups
categorized as “zero” that actually agreed on a variable number of hours that involved combinations
0 and 10 hours, or combinations of 3 and 10 hours. Even though we mentioned that the agreement
was fixed, they decided to establish a variable agreement. These groups earned the highest profit of
the round. We can also observe this situation in Round 3 of the laboratory experiments (Panel c.).
In this round, after the groups were reshuffled, three groups chose combinations of 0 and 10 hours,
and 0 and 3 hours. Again, these three groups earned the highest total benefits, and the members of
one group earned $364, which is almost the value that participants would have earned at the social
optimum ($365).
We are also interested in analyzing whether participants deviate from the agreements or not,
as well as the factors that drive deviation. Figure 17 shows cumulative percentages of deviation from
the agreement by type of agreement. We have excluded those groups with agreements of “zero” since
the agrement involved variable values. It is clear that participants of the field experiment show higher
rates of deviation, in general, than participants of the laboratory experiment. This explains in part
why the treatment effect of the AG treatment was significantly higher in the laboratory experiments
than in the field ones. Moreover, participants of the field experiments whose group chose values of
six pumping hours deviate earlier than other participants. It might be possible that the well depth
in each period, along with personal characteristics might affect deviation rates. Again, we can test
the influence of other factors on deviation rates with a Cox proportional hazard model. Results from
the Cox proportional hazard model for the field experiment are presented in Table 13. We also ran
the Cox model with the data from the laboratory experiment. However, the effects of all covariates
was significantly different from zero and deviations were mostly explained by the baseline hazard rate
which depends only on time.
For the field experiment, we can see that well depth and its square do not influence the hazard
rate of defection. This result seems counterintuitive, since one would expect a higher probability of
defection when the well depth increases. Apparently, after controlling for personal characteristics,
those variables do not influence the deviation rates. Turning to the constant variables, having an
agreement of “6” and “8” significantly increases the probability of deviation in 195 and 579 percent.
However, these values are very high and should be taken with caution, since the number of groups
that chose those values is small. Further, some individual characteristics also affect the hazard rate of
defection. For instance, education have significant effects in reducing the hazard of defection, although
not in a monotonic way. Also, be part of an ejido and having alfalfa as major crop have very strong
effects on reducing the probability of defection, reducing the hazard rate in 77.6 and 46.8 percent,
respectively. An Ejido is a legally recognized agricultural community in Mexico originally created
manage agricultural land as a community. However, farmers that are part of the ejido currently
possess individual property rights over their land. Nevertheless, in many ejidos land and water use
decisions are still discussed with all their members. The fact that farmers that are part of an ejido
tend to deviate less from the agreement gives some insights about the importance of reputation and
social values when face-to-face communication is part of the negotiation.
20
7 Discussion
The results of the experiments presented in this study show that there are important differences be-
tween the behavior of the participants from the field and laboratory experiments. Some researchers
may argue that it is a matter of time and as more rounds are played, participants of the field exper-
iment will tend to behave as those from the laboratory experiment. Nevertheless, our experiments
have given some insights about the importance of background information on decisions making in
experiments. The difference in the results obtained in the IN treatment are very clear. Moreover, the
failure of the efficient technology of improving the benefits in the laboratory experiment suggests the
presence of strategic behavior of participants, who do not have background information about the
technology.
Background information is extremely important when analyzing the behavior of people in ex-
periments. Some subjects might be willing to participate in some programs that would help to
improve joint welfare, and other subjects might be less collaborative, depending on background infor-
mation. This is a key issue for policy design. When local authorities are willing to design a program
that helps to preserve a common-property resource, it is imperative to work with communities and
consider local traditions and history to build institutions. Theoretical results are important as a
benchmark, but local norms and the way they evolve with the provision of the natural resource are
fundamental. Turning to the laboratory experiment, an interesting result from the IN treatment was
that the probability of adoption in the laboratory setting was highly correlated with the results of
the Cognitive Reflection Test (CRT), which is usually correlated with patient and/or less risk-averse
participants.
Our results also show that, in this setting, participants of the field experiment are less likely to
comply with the agreement, in comparison to those of the field experiment. This is surprising, since
one would expect that social norms are more important in the field than in the lab. Our results show
that farmers that are part of an ejido are more likely to comply the agreement than other farmers.
This result partially confirms that importance of social norms in these settings. Our hypothesis to
explain the high deviation rates in the field experiment is that the process to reach an agreement was
less thoughtful in the field experiment than in the lab. In order to achieve a sustainable agreement, it
is necessary to discuss the pros and cons of each alternative. Students are more prone to have these
types of discussions in classes than farmers.
Finally, it is necessary to mention that we have not considered unobserved heterogeneity in
our analysis. As shown in several figures, there is a high degree of heterogeneity in decision making
among participants of the field experiment, and in a lesser extent in the laboratory experiment. The
presence of heterogeneity might reflect different attitudes on the use of the resource. As mentioned
before, some individuals might behave more cooperatively and other more competitively. To identify
those attitudes and analyze whether they are correlated with some individual characteristics is funda-
mental for policy design based on economic experiments. The success of institutions in common-pool
resources management relies on the matching between its design and user’s behavior.
21
8 Acknowledgements
This research would have been impossible to conduct without the support of the Latin American and
Caribbean Environmental Economics Program (LACEEP). We are in debt with them for all their
help. Also, Professors and authorities of the Universidad Autnoma de Aguascalientes were extremely
helpful in the organization of the sessions. In particular professors Joaquın Sosa, Jesus Meraz and
Jose Luna. We would also like to thank Ing. Juan Zamarripa, who was key in the recruitment of
many farmer and the development of many sessions with well owners. The work of Berenice Castillo,
Viviana del Hoyo, Gabriela Flores, AnneLiese Nachman and Charlotte Benson was crutial for the
implementation of the sessions. We also appreciate the comments receieved during the LACEEP
workshops and the European Summer School in Resource and Environmental Economics. We would
also like to thank Prof. James Shortle who kindly supervised the work and gave support to conduct
the laboratory experiments. Finally, this work would have been possible to develop without the
participation of the farmers of Aguascalientes and the undergraduate students of Penn State.
22
9 Tables
Table 1: Parameters used in experiment
Parameter Value
α 10
β 390π
h 1 1
h 0 0.5
N 4
H 10
D 250
R 1
C 200
I 150
f 20π
m 9.460879
k 3
d 0 170
23
Table 2: Revenues with less-efficient technology
Hours Revenue
0 -3
1 7
2 17
3 27
4 37
5 47
6 57
7 67
8 77
9 87
10 97
Revenues by hours of water pumped
with flood irrigation
Table A
Table 3: Revenues with highly-efficient technology
Hours Revenue
0 -12
1 8
2 28
3 48
4 68
5 88
6 88
7 88
8 88
9 88
10 88
Table B
Revenues by hours of water pumped
with drip irrigation
24
Tab
le4:
Pu
mp
ing
cost
s
12
34
56
78
910
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
Wa
ter
lev
el
24
92
48
24
72
46
24
52
44
24
32
42
24
12
40
23
92
38
23
72
36
23
52
34
23
32
32
23
12
30
22
92
28
22
72
26
22
52
24
22
32
22
22
12
20
21
92
18
21
72
16
21
52
14
21
32
12
21
12
10
20
92
08
20
72
06
20
52
04
20
32
02
20
12
00
00
00
00
00
00
00
00
00
00
00
00
00
00
00
00
00
00
00
00
00
00
00
00
00
00
390
195
130
98
78
65
56
49
43
39
35
33
30
28
26
24
23
22
21
20
19
18
17
16
16
15
14
14
13
13
13
12
12
11
11
11
11
10
10
10
10
99
99
88
88
8
780
390
260
195
156
130
111
98
87
78
71
65
60
56
52
49
46
43
41
39
37
35
34
33
31
30
29
28
27
26
25
24
24
23
22
22
21
21
20
20
19
19
18
18
17
17
17
16
16
16
1170
585
390
293
234
195
167
146
130
117
106
98
90
84
78
73
69
65
62
59
56
53
51
49
47
45
43
42
40
39
38
37
35
34
33
33
32
31
30
29
29
28
27
27
26
25
25
24
24
23
1560
780
520
390
312
260
223
195
173
156
142
130
120
111
104
98
92
87
82
78
74
71
68
65
62
60
58
56
54
52
50
49
47
46
45
43
42
41
40
39
38
37
36
35
35
34
33
33
32
31
1950
975
650
488
390
325
279
244
217
195
177
163
150
139
130
122
115
108
103
98
93
89
85
81
78
75
72
70
67
65
63
61
59
57
56
54
53
51
50
49
48
46
45
44
43
42
41
41
40
39
2340
1170
780
585
468
390
334
293
260
234
213
195
180
167
156
146
138
130
123
117
111
106
102
98
94
90
87
84
81
78
75
73
71
69
67
65
63
62
60
59
57
56
54
53
52
51
50
49
48
47
2730
1365
910
683
546
455
390
341
303
273
248
228
210
195
182
171
161
152
144
137
130
124
119
114
109
105
101
98
94
91
88
85
83
80
78
76
74
72
70
68
67
65
63
62
61
59
58
57
56
55
3120
1560
1040
780
624
520
446
390
347
312
284
260
240
223
208
195
184
173
164
156
149
142
136
130
125
120
116
111
108
104
101
98
95
92
89
87
84
82
80
78
76
74
73
71
69
68
66
65
64
62
3510
1755
1170
878
702
585
501
439
390
351
319
293
270
251
234
219
206
195
185
176
167
160
153
146
140
135
130
125
121
117
113
110
106
103
100
98
95
92
90
88
86
84
82
80
78
76
75
73
72
70
3900
1950
1300
975
780
650
557
488
433
390
355
325
300
279
260
244
229
217
205
195
186
177
170
163
156
150
144
139
134
130
126
122
118
115
111
108
105
103
100
98
95
93
91
89
87
85
83
81
80
78
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
Wa
ter
lev
el
19
91
98
19
71
96
19
51
94
19
31
92
19
11
90
18
91
88
18
71
86
18
51
84
18
31
82
18
11
80
17
91
78
17
71
76
17
51
74
17
31
72
17
11
70
16
91
68
16
71
66
16
51
64
16
31
62
16
11
60
15
91
58
15
71
56
15
51
54
15
31
52
15
11
50
00
00
00
00
00
00
00
00
00
00
00
00
00
00
00
00
00
00
00
00
00
00
00
00
00
88
77
77
77
77
66
66
66
66
66
55
55
55
55
55
55
55
55
44
44
44
44
44
44
44
15
15
15
14
14
14
14
13
13
13
13
13
12
12
12
12
12
11
11
11
11
11
11
11
10
10
10
10
10
10
10
10
99
99
99
99
98
88
88
88
88
23
23
22
22
21
21
21
20
20
20
19
19
19
18
18
18
17
17
17
17
16
16
16
16
16
15
15
15
15
15
14
14
14
14
14
14
13
13
13
13
13
13
13
12
12
12
12
12
12
12
31
30
29
29
28
28
27
27
26
26
26
25
25
24
24
24
23
23
23
22
22
22
21
21
21
21
20
20
20
20
19
19
19
19
18
18
18
18
18
17
17
17
17
17
16
16
16
16
16
16
38
38
37
36
35
35
34
34
33
33
32
31
31
30
30
30
29
29
28
28
27
27
27
26
26
26
25
25
25
24
24
24
23
23
23
23
22
22
22
22
21
21
21
21
21
20
20
20
20
20
46
45
44
43
43
42
41
40
40
39
38
38
37
37
36
35
35
34
34
33
33
33
32
32
31
31
30
30
30
29
29
29
28
28
28
27
27
27
26
26
26
25
25
25
25
24
24
24
24
23
54
53
52
51
50
49
48
47
46
46
45
44
43
43
42
41
41
40
40
39
38
38
37
37
36
36
35
35
35
34
34
33
33
33
32
32
31
31
31
30
30
30
29
29
29
28
28
28
28
27
61
60
59
58
57
56
55
54
53
52
51
50
50
49
48
47
47
46
45
45
44
43
43
42
42
41
41
40
39
39
39
38
38
37
37
36
36
35
35
35
34
34
34
33
33
33
32
32
32
31
69
68
66
65
64
63
62
61
59
59
58
57
56
55
54
53
52
52
51
50
49
49
48
47
47
46
46
45
44
44
43
43
42
42
41
41
40
40
39
39
39
38
38
37
37
37
36
36
35
35
76
75
74
72
71
70
68
67
66
65
64
63
62
61
60
59
58
57
57
56
55
54
53
53
52
51
51
50
49
49
48
48
47
46
46
45
45
44
44
43
43
42
42
41
41
41
40
40
39
39
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
Wa
ter
lev
el
14
91
48
14
71
46
14
51
44
14
31
42
14
11
40
13
91
38
13
71
36
13
51
34
13
31
32
13
11
30
12
91
28
12
71
26
12
51
24
12
31
22
12
11
20
11
91
18
11
71
16
11
51
14
11
31
12
11
11
10
10
91
08
10
71
06
10
51
04
10
31
02
10
11
00
00
00
00
00
00
00
00
00
00
00
00
00
00
00
00
00
00
00
00
00
00
00
00
00
00
44
44
44
44
44
43
33
33
33
33
33
33
33
33
33
33
33
33
33
33
33
33
33
33
33
88
88
77
77
77
77
77
77
77
77
66
66
66
66
66
66
66
66
66
66
65
55
55
55
55
12
11
11
11
11
11
11
11
11
11
11
10
10
10
10
10
10
10
10
10
10
10
10
99
99
99
99
99
99
99
88
88
88
88
88
88
8
15
15
15
15
15
15
15
14
14
14
14
14
14
14
14
13
13
13
13
13
13
13
13
13
12
12
12
12
12
12
12
12
12
12
12
11
11
11
11
11
11
11
11
11
11
11
11
11
10
10
19
19
19
19
19
18
18
18
18
18
18
17
17
17
17
17
17
17
16
16
16
16
16
16
16
15
15
15
15
15
15
15
15
15
14
14
14
14
14
14
14
14
14
14
13
13
13
13
13
13
23
23
23
23
22
22
22
22
21
21
21
21
21
21
20
20
20
20
20
20
19
19
19
19
19
19
18
18
18
18
18
18
18
17
17
17
17
17
17
17
17
16
16
16
16
16
16
16
16
16
27
27
27
26
26
26
26
25
25
25
25
24
24
24
24
24
23
23
23
23
23
22
22
22
22
22
21
21
21
21
21
21
21
20
20
20
20
20
20
20
19
19
19
19
19
19
19
18
18
18
31
31
30
30
30
29
29
29
29
28
28
28
28
27
27
27
27
26
26
26
26
26
25
25
25
25
25
24
24
24
24
24
23
23
23
23
23
23
22
22
22
22
22
22
22
21
21
21
21
21
35
34
34
34
33
33
33
33
32
32
32
31
31
31
31
30
30
30
29
29
29
29
29
28
28
28
28
27
27
27
27
27
26
26
26
26
26
25
25
25
25
25
25
24
24
24
24
24
24
23
39
38
38
38
37
37
36
36
36
35
35
35
35
34
34
34
33
33
33
33
32
32
32
31
31
31
31
30
30
30
30
30
29
29
29
29
28
28
28
28
28
27
27
27
27
27
27
26
26
26
Ta
ble
C
Pu
mp
ing
co
sts
by
wa
ter
lev
el
an
d p
um
pin
g h
ou
rs
9 10
Pu
mp
ing
ho
urs
Pu
mp
ing
ho
urs
Pu
mp
ing
ho
urs
4 5 6 7 810 0 1 2 35 6 7 8 90 1 2 3 40 1 2 3 4 105 6 7 8 9
25
Table 5: Summary of sessions: Field and laboratory experiments
Treatment Sessions Groups Participants Periods
Counterfactual 7 21 84 840
Investment 8 21 84 840
Agreement 10 22 88 880
Total 25 64 256 2,560
a. Field
Treatment Sessions Groups Participants Periods
Counterfactual 14 23 92 920
Investment 7 14 56 560
Agreement 6 13 52 520
Total 27 50 200 2,000
b. Laboratory
26
Table 6: Average outcomes by type of session, round and period: Field experiments
Period Treatment Round
Round 1 6.99 32.83 170.00
Round 2 7.24 34.26 170.00
Round 1 8.05 *† 38.19 *† 170.00
Round 2 5.61 *† 48.74 *† 170.00
Round 1 7.38 34.89 170.00
Round 2 7.15 34.09 170.00
Round 1 6.51 26.20 177.95
Round 2 6.95 28.52 178.95
Round 1 7.42 *† 28.25 182.19 *†
Round 2 5.46 *† 50.58 *† 172.43 *†
Round 1 7.26 *† 29.83 *† 179.50
Round 2 6.85 27.83 178.59
Round 1 6.58 22.88 184.00
Round 2 6.89 22.68 186.76
Round 1 7.18 18.40 *† 191.86 *†
Round 2 5.25 *† 51.54 *† 174.29 *†
Round 1 7.20 22.10 188.55 *†
Round 2 6.60 20.75 186.00
Round 1 6.60 16.92 190.33
Round 2 6.80 13.76 194.33
Round 1 6.90 7.75 *† 200.57 *†
Round 2 5.00 *† 50.12 *† 175.29 *†
Round 1 6.89 10.92 *† 197.36 *†
Round 2 6.40 10.88 192.41
Round 1 5.96 10.94 196.71
Round 2 6.46 5.40 201.52
Round 1 6.02 -5.00 *† 208.19 *†
Round 2 5.04 *† 51.96 *† 175.29 *†
Round 1 6.16 -2.08 *† 204.91 *†
Round 2 6.18 -19.08 198.00
Calculation of differences at 5% of error between CF and treatment for: * means (T-test)
and, † distributions (Mann-Whitney test)
Agreement
Counterfactual
Investment
Agreement
Agreement
Counterfactual
Investment
Agreement
Counterfactual
Investment
1
2
3
4
5
Counterfactual
Investment
Agreement
Counterfactual
Investment
Hours of
pumpingBenefit Well Depth
27
Table 7: Average outcomes by type of session, round and period: Laboratory experi-ments
Period Treatment Round
Round 1 8.27 39.29 170.00
Round 2 8.83 42.18 170.00
Round 1 7.96 37.80 170.00
Round 2 7.23 *† 49.88 *† 170.00
Round 1 7.75 36.60 170.00
Round 2 5.85 *† 27.23 *† 170.00
Round 1 8.47 32.00 183.09
Round 2 8.99 32.43 185.30
Round 1 8.32 33.07 181.86 †
Round 2 6.91 *† 48.30 *† 178.93 *†
Round 1 8.02 31.58 181.00 *†
Round 2 5.06 *† 21.31 *† 173.38 *†
Round 1 8.01 18.29 196.96
Round 2 7.32 12.50 201.26
Round 1 7.82 19.71 195.14 †
Round 2 6.45 † 42.84 *† 186.57 *†
Round 1 8.13 22.37 *† 193.08 *†
Round 2 5.87 *† 24.67 *† 173.62 *†
Round 1 4.49 3.23 209.00
Round 2 3.29 1.70 210.52
Round 1 5.30 4.82 206.43
Round 2 5.66 *† 37.64 *† 192.36 *†
Round 1 6.38 *† 6.42 *† 205.62 *†
Round 2 5.83 *† 23.56 *† 177.08 *†
Round 1 6.14 7.63 206.96
Round 2 7.35 13.08 203.70
Round 1 4.84 4.16 207.64
Round 2 5.98 *† 38.82 *† 195.00 *†
Round 1 4.54 † 2.31 *† 211.15 *†
Round 2 8.38 30.44 *† 180.38 *†
Calculation of differences at 5% of error between CF and treatment for: * means (T-test)
and, † distributions (Mann-Whitney test)
3
4
5
Agreement
Counterfactual
Investment
Agreement
Counterfactual
Investment
Agreement
Counterfactual
Investment
Agreement
Counterfactual
Investment
1
2
Agreement
Hours of
pumpingBenefit Well Depth
Counterfactual
Investment
28
Table 8: Average end-of-round outcomes by type treatment and round: Field experi-ments
Total Benefits 287.6 295.66 315.44 *† 274.47
Total hours
pumped35.57 34.88 26.36 *† 33.18
Final Well Depth 212.29 209.55 175.43 *† 202.73
Calculation of differences at 1% of error between Rounds for: * means (T-test) and, † distributions (Mann-Whitney test)
Total Benefits 287.6 *† 295.66 315.44 274.47
Total hours
pumped35.57 34.88 26.36 *† 33.18
Final Well Depth 212.29 209.55 175.43 *† 202.73
Calculation of differences at 1% of error between CF and treatment for: * means (T-test) and, † distributions (Mann-Whitney test)
a. Differences for each treatment between Round 1 and Round 2
b. Differences between CF and other treatments within Round 1 or Round 2
200.57
309.77
32.64
200.57
309.77
32.64
Investment Agreement Investment AgreementVariable
Round 2Round 1
Counterfactual Counterfactual
VariableRound 2 Round 2
Counterfactual Investment Agreement Counterfactual Investment Agreement
207.38
304.63
34.35
207.38
304.63
34.35
Table 9: Average end-of-round outcomes by type treatment and round: Lab experiments
Total Benefits 299.27 323.73 *† 327.21 *†
Total hours
pumped34.83 32.23 *† 30.98 *†
Final Well Depth 209.31 198.93 *† 193.92 *†
Calculation of differences at 1% of error between Rounds for: * means (T-test) and, † distributions (Mann-Whitney test)
Total Benefits 299.27 323.73 *† 327.21 *†
Total hours
pumped34.83 32.23 † 30.98 *†
Final Well Depth 209.31 198.93 193.92 *†
Calculation of differences at 1% of error between CF and treatment for: * means (T-test) and, † distributions (Mann-Whitney test)
211.52
300.45
35.38
211.52
300.45
35.38
Investment Agreement Investment AgreementVariable
Round 2Round 1
Counterfactual Counterfactual
Round 2
Counterfactual Investment Agreement Counterfactual Investment Agreement
a. Differences for each treatment between Round 1 and Round 2
b. Differences between CF and other treatments within Round 1 or Round 2
207
299.57
34.25
207
299.57
34.25
213.09
301.89
35.77
213.09
301.89
35.77
VariableRound 1
29
Table 10: Difference-in-Difference estimation of final outcomes in rounds 1 and 2: Fieldexperiments
(1) (2) (3) (4) (5)
Total pumping
hours
Total pumping
hours - FullTotal Benefits
Total Benefits -
FullFinal Well Depth
Treatment (Base = CF)
IN = 1 2.929** 3.292** -22.18* -21.19** 11.71**
AG = 1 2.244* 3.204** -14.11 -21.91** 8.974*
Round 2 1.702* 1.821* -5.143 -6.154 6.810
IN X Round 2 -10.92*** -11.36*** 32.99** 35.44*** -43.67***
AG X Round 2 -3.407** -3.858*** -16.05 6.413 -13.63**
Age -0.434** -0.455
Age2 0.00330* 0.00237
Female 0.829 2.859
Starting Year 3.748 47.02
Starting Year2 -0.000946 -0.0119
Head of household -2.644 2.92
Marital Status (Base = Single)
Married 0.262 -13
Cohabitant 1.462 24.68
Widow 1.483 3.803
Working time (Base = Only on farm)
Mostly on farm -0.554 4.729
Half on farm, half off-farm -2.741** -5.044
Mostly off-farm -4.827*** -15.14
Only off-farm -19.36*** -106.0***
Retired -1.462 -5.008
Education level (Base = Preeschool/None)
Primary school -1.745 -5.281
Secondary school -2.078 1.196
Technical school/Preparatory -1.659 8.714
Agricultural Technical school 1.692 21.13
University and Graduate school -3.586 5.804
Irrigated hectares 0.762*** 3.633***
Irrigated hecatares2 -0.00998*** -0.0416***
Land is ejido 0.761 17.62*
Source of water (Base = Dam)
Only water from well 1.246 7.183
Both water from well and dam 1.046 3.191
Primary crop alfalfa -0.0794 -4.915
Primary crop vine 3.520* 9.321
Area with drip irrigation -0.410*** -2.031***
Municipality (Base = Pabellón de Arteaga)
Aguascalientes -1.092 24.5
Asientos -0.678 -42.51***
Calvillo -4.246 10.3
Cosio -1.214 10.55
El Llano 0.383 -7.745
Rincón de Romos -3.922** -2.355
San Francisco 0.349 -45.68
San José de Gracia -4.809** 30.34***
Tepezalá -0.932 -8.894
Constant 32.64*** -3689 309.8*** -46349 200.6***
σu 5.882*** 4.373*** 25.95*** 9.117**
σe 6.411*** 6.380*** 74.32*** 48.67*** 14.93***
Observations 512 474 512 474 128
Number of caseid 256 237 256 237 64
Equation (5) estimated at the well (group) level
*** p<0.01, ** p<0.05, * p<0.1
Variables
Likelihood-ratio test performed for significancy of σu
30
Table 11: Difference-in-Difference estimation of final outcomes in rounds 1 and 2: Labexperiments
(1) (2) (3) (4) (5)
Total pumping
hours
Total pumping
hours - FullTotal Benefits
Total Benefits -
FullFinal Well Depth
Treatment (Base = CF)
IN -1.130 -1.130 -0.874 -2.971 -4.522
AG -0.554 -0.554 -1.176 -2.682 -2.214
Round 2 0.391 0.391 1.446 1.446 1.565
IN x Round 2 -2.409 -2.409 22.72*** 22.72*** -9.637
AG x Round 2 -4.237*** -4.237*** 26.50*** 26.50*** -16.95***
Age 0.142 -0.0317
Female 1.105 1.147
Major group (Base = Business )
Education/Psychology -1.041 -6.991
Energy/Envionment 3.372 14.00
Engineering -0.761 -11.35
"Hard" Sciences -1.451 -2.517
Human Devlopment/Medicine -0.397 -7.784
Information Sciences -1.952 -7.364
Liberal Arts -0.660 -8.363
Media -3.186* -13.77*
Natural Sciecnes 0.797 -7.820
CRT answer correct
Question 1 0.823 2.092
Question 2 -1.078 -2.030
Question 3 0.257 2.974
Constant 35.38*** 32.10*** 300.4*** 303.2*** 211.5***
σu 3.187*** 2.924*** 10.23*** 8.604*** 0
σe 6.135*** 6.135*** 24.73*** 24.73*** 13.01***
Observations 400 400 400 100 100
Number of caseid 200 200 200 50 50
Equation (3) is estimated at the well (group) level
*** p<0.01, ** p<0.05, * p<0.1
Variables
F test performed for significancy of σu
31
Table 12: Cox proportional hazard model of investment: Laboratory experiments
Variables Hazard ratio
Well depth 3.755**
Well depth sq. 0.996**
Age 6.540
Age sq. 0.957
Female 0.855
Major group (Base = Business )
Energy/Envionment 0.000
Engineering 0.559
"Hard" Sciences 1.323
Human Devlopment/Medicine 0.792
Information Sciences 3.483***
Liberal Arts 0.350
Media 1.598
Natural Sciecnes 0.000
CRT score (Base: Score = 3)
Score = 0 0.3440***
Score = 1 0.416**
Score = 2 0.520*
Observations 56
Robust standard errors in parentheses
*** p<0.01, ** p<0.05, * p<0.1
32
Table 13: Cox proportional hazard model of deviation from agreement: Field experi-ments
Variable Hazard rate
Well Depth 0.920*
Well Depth sq. 1.0002
Agrement (Base = 5)
Agreement = 6 2.953***
Agreement = 7 2.104
Agreement = 8 5.789***
Agreement = 9 1.542
Agreement = 10 1.365
Age 0.988
Education level (Base = Pre-school/None)
Primary school 0.372***
Secondary school 0.158***
Technical school/Preparatory 0.345**
Agricultural Technical school 1.401
University and Graduate school 0.207**
Irrigated hectares 0.899
Working time (Base = Only on farm)
Mostly on farm 2.915
Half on farm, half off-farm 1.234
Mostly off-farm 1.679
Retired 3.336*
Source of water (Base = Well)
Only water from dam 0.357*
Both water from well and dam 0.444
Land is ejido 0.224***
Primary crop alfalfa 0.532***
Municipality (Base = Pabellón de Arteaga)
Aguascalientes 1.005
Cosio 0.087***
El Llano 3.676*
San Francisco 3.142*
Tepezalá 1.183
Observations 83
Robust standard errors in parentheses
*** p<0.01, ** p<0.05, * p<0.1
33
10 Figures
Figure 1: Average Depth-to-water in Aguascalientes, 1965-2005
Source: Gobierno del Estado de Aguascalientes (2009)
Figure 2: Equilibrium Pumping hours for Myopic, Rational/strategic and Fully Coop-erative behaviors
34
Figure 3: Benefits of pumping for Myopic, Rational/strategic and Fully Cooperativebehaviors
Figure 4: Well depth for Myopic, Rational/strategic and Fully Cooperative behaviors
35
Figure 5: Simulated benefits for Myopic, Rational/strategic and Fully Cooperative be-haviors, and fixed arrangements
36
Figure 6: Average individual hours pumped by period and treatment
a. Field
b. Laboratory
37
Figure 7: Average individual benefits by period and treatment
a. Field
b. Laboratory
38
Figure 8: Average well depth by period and treatment
a. Field
b. Laboratory
39
Figure 9: Histogram of pumping hours from field experiment, by Round, Period andTreatment
a. Counterfactual
b. Investment
b. Agreement
40
Figure 10: Histogram of pumping hours from laboratory experiment, by Round, Periodand Treatment
a. Counterfactual
b. Investment
b. Agreement
41
Figure 11: Average individual total benefits by round and treatment
a. Field
b. Laboratory
42
Figure 12: Average individual total hours pumped by round and treatment
a. Field
b. Laboratory
43
Figure 13: Average final well depth by round and treatment
a. Field
b. Laboratory
44
Figure 14: Percentage of adoption of technology by period and experiment
Figure 15: Percentage of adoption of technology by period and round: Laboratoryexperiment
45
Figure 16: Histogram of agreed values of pumping hours in “AG” sessions
a. Field
b. Laboratory - Round 2
c. Laboratory - Round 3
46
Figure 17: Agreement deviation percentages in “AG” sessions
a. Field
b. Laboratory - Round 2
c. Laboratory - Round 3
47
References
Anderson, Jon, Stepehen V. Burks, Jeffrey Carpenter, Lorenz Gotte, Karsten Maurer, Daniele Nosenzo, Ruth Potter,Kim Rocha and Aldo Rustichini (2013), Self-selection and variations in the laboratory measurement of other-regardingpreferences across subject pools: evidence from college students and two adult samples, Experimental Economics,Vol. 16: 170-189.
Agaarwal, Rimjhim M. and Tulika A. Narayan (2004), Does inequality lead to greater efficiency in the use of localcommons? The role of strategic investment in capacity, Journal of Environmental Economics and Management, Vol.47: 163-182.
Bardhan, Pranab (2000), Irrigation and Cooperation: An Empirical Analysis of 48 Irrigation Communities in SouthIndia, Economic Development and Cultural Change, Vol 48 (4): 847-865.
Barham, Bradford L., Jean-Paul Chavas and Oliver T. Coomes (1998), Sunk Costs and the Natural Resource ExtractionSector: Analytical Models and Historical Examples of Hysteresis and Strategic Behavior in the Americas, LandEconomics, Vol. 74 (4): 429-448.
Belot, Michele, Luis Miller and Raymond Duch (2010), Who Should be Called to the Lab?: A comprehensive compar-ison of students and non-students in classic experimental games, Nuffield Centre for Experimental Social SciencesDiscussion Paper Series 2010-001.
Brown, Gardner Jr. and Robert Deacon (1972), Economic Optimization of Single-Cell Aquifer, Water Resources Re-search, Vol. 8(3): 557-564
Burt, Oscar R. (1964), Optimal Resource Use Over time with and Application to Ground Water, Management Science,Vol. 11(1) Series A: 80-93
Burt, Oscar R. (1967), Temporal Allocation of Groundwater, Water Resources Research, Vol. 3(1): 45-56
Burt, Oscar R. (1970), Groundwater Storage Control undr Institutional Restrictions, Water Resources Research, Vol.6(6): 1540-1548
Caldera Ortega, Alex Ricardo (2009), Gobernanza y sustenibilidad: Desarrollo institucional y procesos polıticos entorno al agua subterranea en Mxico. Los casos del valle de Leon y del valle de Aguascalientes, Ph.D. Dissertation inPolitical Sciences, Facultad Latinoamericana de Ciencias Sociales-FLACSO, Mexico City.
Cameron, A. Colin and Pravin K. Trivedi (2005), Microeconometrics: Methods and Applications, New York: CambridgeUniversity Press.
Cardenas, Juan Camilo, Luz Angela Rodriguez and Nancy Johnson (2011), Collective Action for Watershed Manage-ment: Field Experiments in Colombia and Kenya, Environment and Development Economics, Vol.16: 275-303.
Carey, Janis M. and David Zilberman (2002), A Model of Investment under Uncertainty: Modern Irrigation Technologyand Emerging Markets in Water, American Journal of Agricultural Economics, Vol. 84 (1): 171-183.
Carpenter, Jeffrey, Stephen Burks and Eric Verhoogen (2005), Workers: The Effects of Social Framing on Behavior inDistribution Games, Research in Experimental Economics, Vol 10:261-290.
Carpenter, Jeffrey, Cristina Connolly and Caitlin Knowles Myers (2008), Altruistic Behavior in a Representative DictatorExperiment, Experimental Economics, Vol. 11:282-298.
Comite Tecnico de Aguas Subterrneas/Ojocaliente-Aguascalientes-Encarnacion (COTAS-OAE) (2006), Escenarios delAgua 2015 y 2030 en el Acuıfero Interestatal Ojocaliente-Aguascalientes-Encarnacion: acciones para un desarrollocon sostenibilidad ambiental, dowloaded from http://www.cotas.org.mx/.
Coward, E.W. (1976), Principles in social organization in an indigenous irrigation system, Human organization, Vol.38(1): 28-36.
Dayton-Johnson, Jeff (2000), Determinants of collective action on the local commons: a model with evidence fromMexico, Journal of Development Economics, Vol. 62, pp.181-208.
Dosi, Cesare and Michele Moretto (2010), Environmental Innovation, War of Attrition and Investment Grants, Inter-national Game Theory Review, Vol. 12 (1): 37-59.
Frederick, Shane (2005), Cognitive Reflection and Decision Making, Journal of Economic Perspectives, Vol. 19(4),pp:25-42
48
Gardner, Roy, Elinor Ostrom and James M. Walker (1990), The Nature of Common-Pool Resource Problems, Rationalityand Society, Vol. 2(3):335-358.
Giordana, Gaston (2008), Wealthy people do better? Experimental Evidence on Endogenous Time Preference Het-erogeneity and the Effect of Wealth in Renewable Common-Pool Resources Exploitation, Document de Recherche,Laboratoire Montpellierain d’ Economie Theorique et Appliquee, No.2008-10.
Gisser, Micha and David A. Sanchez (1980), Comptetion Versus Optimal Control in Groundwater Pumping, WaterResources Research, Vol. 16(4):638-642.
Gobierno del Estado de Aguascalientes (2009), Sistema de Indicadores Ambientales del Estado de Aguascalientes, Aguas-calientes, Mexico: Gobierno del Estado de Aguascalientes and Instituto del Medio Ambiente.
Hackett, Steven, Edella Schlager and James Walker (1994), The Role of Communication in Resolving Common Dilem-mas: Experimental Evidence with Heterogenous Appropriators, Journal of Environmental Economics and Manage-ment, Vol. 27: 99-126.
Hardin, Garrett (1968), The Tragedy of the Commons, Science, 162: 1243-1248.
Instituto Nacional de Estadıstica y Geografıa (2009), Estados Unidos Mexicanos, Censo Agropecuario 2007, VII CensoAgrıcola, Ganadero y Forestal, Mexico D.F.: INEGI.
Janssen, Marco, John M. Anderies, J.C. Cardenas (2011), Head-enders as stationary bandits in asymmetric commons:Comparing irrigation experiments in the laboratory and the field, Ecological Economics, Vol. 70 (9): 1590-1598.
Koundouri, Phoebe (2004), Current Issues in the Economics of Groundwater Resource Management, Journal of Eco-nomic Surveys, Vol 18(5): 703-740
Koundouri, Phoebe (2004), Potential for groundwater management: Gisser-Sanchez effect reconsidered, Water ResourcesResearch, Vol. 40, W06S16, doi:10.1029/2003WR002164.
Madani, Kaveh and Ariel Dinar (2012), Non-cooperative institutions for sustainable common pool resource management:Applications to groundwater, Ecological Economics, Vol. 74: 34-45.
McGinnis, Michael D., ed. (2000), Polycentric Games and Institutions: Readings from the Workshop in Political Theoryand Policy Analysis, Ann Arbor, MI: The University of Michigan Press.
Moreno-Sanchez, Rocıo del Pilar and Jorge Higinio Maldonado (2010), Evaluating the role of co-management in im-proving governance of marine protected areas: An experimental approach in the Colombian Caribbean, EcologicalEconomics, Vol. 69: 2577-2567.
Moretto, Michele (2000), Irreversible investment with uncertainty and strategic behavior, Economic Modelling, Vol. 17:589-617.
Nieswiadomy, Michael (1985), The Demand for Irrigation Water in the High Plains of Texas, 1957-1980, AmericanJournal of Agricultural Economics, Vol. 67(3): 619-626
Negri, Donald H. (1989), The Common Property Aquifer as a Differential Game, Water Resources Research, Vol. 25(1): 9-15.
Ostrom, Elinor (1986), Issues of definition and theory: some conclusions and hypothesis, Proceedings of the conferenceon common property management, National Research Council, Washington D.C.: National Academy Press.
Ostrom, Elinor (1990), Governing the Commons: the evolution of institutions for collective action, New York: CambridgeUniversity Press.
Ostrom, Elinor (1992), Crafting institutions for self-governing irrigation systems, San Francisco, CA: Institute forContemporary Studies University Press.
Ostrom, Elinor and Roy Gardner (1993), Coping with Asymmetries in the Commons: Self-Governing Irrigation SystemsCan Work, The Journal of Economic Perspectives, Vol. 7(4): 93-112.
Ostrom, Elinor (2002), Chapter 24: Common-Pool Resources and Institutions: Toward a Revised Theory, in BruceGardner L. and Gordon C. Rausser (eds.), Handbook of Agricultural Economics, Vol. 2, Amsterdam: Elsevier ScienceB.V.
49
Provencher, Bill and Oscar Burt (1993), The Externalities Associated with the Common Property Exploitation ofGroundwater, Journal of Environmental Economics and Management, Vol. 24: 139-158.
Salcedo, Rodrigo (2014), Dynamic decision making in common-pool resource economic experiments: Behavioral hetero-geneity in the field and the lab, LACEEP Working paper
Suter, Jordan F., Joshua M. Duke, Kent D. Messer and Holly A. Michael (2012), Behavior in a Spatially ExplicitGroundwater Resource: Evidence from the Lab, American Journal of Agricultural Economics, Vol. 94(5):1094-1112.
Swallow, Brent, Nancy Johnson, Rurth Meinzen-Dick and Anna Knox (2006), The challenges of inclusive cross-scalecollective action in watersheds, Water International, Vol. 31(3): 361-376.
Trawick, Paul (2003), Against the Privatization of Water: An Indigenous Model for Improving Existing Laws andSuccessfully Governing the Commons, World Development, Vol. 31(6): 977-966.
Wang, Chenggang and Eduardo Segarra (2011), The Economics of Commonly Owned Groundwater When User DemandIs Perfectly Inelastic, Journal of Agricultural and Resource Economics, Vol 36(1): 95-120
Worthington, Virginia E., Oscar R. Burt and Richard L. Brustkern (1985), Optimal Management of a Confined Ground-water System, Journal of Environmental Econmics and Management, Vol. 12(3): 229-245
50