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