RESOURCE MANAGEMENT IN A DEVELOPING COUNTRY CONTEXT: IMPROVING
DECISIONS BY CONFRONTING
DIFFICULT TRADEOFFS IN COSTA RICA
By
Delanie Suzanne Kellon

A DISSERTATION
Submitted to
Michigan State University
in partial fulfillment of the requirements
for the degree of
DOCTOR OF PHILOSOPHY
Community, Agriculture, Recreation and Resource Studies
2011

ABSTRACT
RESOURCE MANAGEMENT IN A DEVELOPING COUNTRY CONTEXT: IMPROVING
DECISIONS BY CONFRONTING DIFFICULT TRADEOFFS
IN COSTA RICA
By
Delanie Suzanne Kellon
Though laudable and necessary, the international development field’s recent shift
toward more holistic approaches and methods presents a significant challenge to
governments and stakeholder groups throughout developing regions of the world. The
presence of critically important—but often conflicting—social, economic, and
environmental objectives in decision making for international development point to the
inevitability of some difficult tradeoffs; the need to give up something valued in order to
gain something else that is also valued but for different reasons. This dissertation
seeks to better understand, and gauge the effectiveness of, two approaches for
confronting these kinds of tradeoffs in an international development context. Presented
in three chapters, the dissertation begins by using insights from the decision sciences to
discuss ways to address common pitfalls in decision making about the environment.
The second and third chapters then report results from research that draws on these
concepts and explores two approaches for confronting environment-development
tradeoffs in Costa Rica: (1) a stated choice (SC) approach that relies upon conjoint
analysis and (2) a decision analytic method based on insights from multiattribute utility
theory. In addition to responding to policy oriented research questions, the second and
third papers also set the stage for future analysis of both approaches from a
methodological perspective.

ACKNOWLEDGMENTS

I am grateful to my academic advisor Dr. Joseph Arvai, my committee members Dr.
Robin Gregory, Dr. Robert Richardson, Dr. Louie Rivers, and Dr. Paul Thompson,
faculty and staff at EARTH University (especially Dr. José Zaglul, Dr. Daniel Sherrard,
and Dr. Ramón León), Graham Long, Carmen Daly, the interview team, and all the
research participants for their assistance and insights. This research was supported by
The Fulbright DDRA Program, and the National Science Foundation under award
number SES-0924210. Any opinions, findings, and conclusions or recommendations
expressed in this publication are mine and do not necessarily reflect the views of the
sponsors.

iii

TABLE OF CONTENTS

LIST OF TABLES

vi

INTRODUCTION

1

CHAPTER 1
FIVE PROPOSITIONS FOR IMPROVING DECISION MAKING ABOUT
THE ENVIRONMENT IN DEVELOPING COMMUNITIES: INSIGHTS
FROM THE DECISION SCIENCES
1. Introduction
2. Five propositions for improving decision making in developing
communities
2.1 Decision-making processes must account for the
constructive nature of preferences
2.2 The basis for development decisions must be realistic
and relevant
2.3 Expand deliberative processes to account for the expertise
of local communities
2.4 Rethink the goals of participatory decision-making
processes
2.5 Avoid decision paralysis
3. Conclusion
References
CHAPTER 2
A CHOICE EXPERIMENT TO ADDRESS ENVIRONMENTAL, HEALTH,
AND ECONOMIC DEVELOPMENT CONFLICTS ABOUT PINEAPPLE
PRODUCTION IN COSTA RICA
1. Introduction
2. Methods
2.1 Choice Experiment Design
2.2 Respondents
2.3 Analysis
3. Results
4. Discussion
4.1 Policy Implications
Appendices
References

iv

6
6
10
10
15
21
26
31
35
38

43
43
47
47
51
53
54
59
61
68
77

CHAPTER 3
STRUCTURING DECISION MAKING IN DEVELOPING COMMUNITIES:
INFORMING LAND USE POLICY IN COSTA RICA
1. Introduction
2. Methods
2.1 Context
2.2 Design
2.3 Respondents
2.4 Analysis
3. Results
4. Discussion
Appendices
References

78
78
82
82
84
88
90
90
93
102
112

CONCLUSION

116

v

LIST OF TABLES

Table 1. Attributes and levels used in the CE design (translated from Spanish). 68
Table 2. Example choice set (translated from Spanish).

68

Table 3. Summary sample characteristics.

69

Table 4. Short description of choice model covariates.

70

Table 5. Estimation results (Choice and complete models).

72

Table 6. Complete model by canton.

74

Table 7. Alternative pineapple production scenario.

75

Table 8. Alternatives developed for evaluation by stakeholders.
*Alternative 1 reflects the status quo.

102

Table 9. Worst and best ranges, derived from the options presented
in Table 8, used in the swing weighting procedure. Italicized figures
depict the ranks and weights elicited from a single respondent.

103

Table 10. Mean weights (wt.) and standard error (se) elicited for each
attribute by region. For the Tukey’s post-test, all significant differences are at
the p < 0.05 level where: 1 corresponds with Pococí ≠ Guácimo; 2 corresponds
with Guácimo ≠ Siquirres; and 3 corresponds with Pococí ≠ Siquirres.

104

Table11. Mean weights (wt.) and standard error (se) elicited for each
attribute by age. For the Tukey’s post-test, all significant differences are
at the p < 0.05 level where: 1 corresponds with 18-40 Years ≠ 41-60 Years;
2 corresponds with 41-60 Years ≠ >60 Years; and 3 corresponds with
18-40 Years ≠ >60 Years.

105

Table 12. Mean weights (wt.) and standard error (se) elicited for each
attribute as a function of education level (characterized by the number of
years of formal education). For the Tukey’s post-test, all significant
differences are at the p < 0.05 level where: 1 corresponds with
0-6 Years ≠ 7-11 Years; 2 corresponds with 7-11 Years ≠ >11 Years;
and 3 corresponds with 0-6 Years ≠ >11 Years.

106

Table 13. Mean weights (wt.) and standard error (se) elicited for each
attribute as a function of gender.

107

vi

Table 14. Mean weights (wt.) and standard error (se) by attribute relative
to the poverty line (PL) of ₡218,433 per month.

108

Table 15. Mean weights (wt.) and standard error (se) by attribute relative
to respondents’ relationship with the pineapple industry. Respondents
reporting “yes” either worked in—or had a close relative or family member
working in—the pineapple industry.

109

Table 16. Follow-Up survey results.

110

vii

INTRODUCTION

For more than half a century, research and practice in international development has
focused on improving the quality of life of people living in developing regions of the
world. More recently, development agencies and organizations have recognized the
need to also include priorities associated with environmental sustainability as part of
their mission. This expansion of international development to include environmental
objectives makes sense: in all regions of the world—developed and developing—human
health and quality of life are directly linked to the health of the environment and the
sustainable use of natural resources.
While entirely justified, this more holistic approach to international development
presents a significant challenge to governments and stakeholder groups throughout
developing regions of the world. On one side, communities must contend with rural
poverty due in part to un- or underemployment. To address this problem, there is
significant pressure in many countries—Costa Rica being but one example—to harvest
forests and develop large stretches of land to produce agricultural products for export.
On the other side is the recognition that socio-economic objectives must be balanced
with environmental considerations. Adding to this complexity, the development of rural
watersheds threatens many community water supplies, creating a conflict between local
residents, farmers, developers, and governmental agencies.
The presence of these critically important—but often conflicting—social,
economic, and environmental objectives in decision making for international
development point to the inevitability of some difficult tradeoffs; the need to give up

1

something valued in order to gain something else that is also valued but for different
reasons. This dissertation seeks to better understand, and gauge the effectiveness of,
two approaches for confronting these kinds of environment-development tradeoffs in an
international development context.
To start the discussion about methods, the dissertation begins by using insights
from the decision sciences to discuss ways to address common pitfalls in decision
making about the environment, then reports results from research that draws on these
insights and explores two approaches for confronting environment-development
tradeoffs in Costa Rica: (1) a stated choice (SC) approach that relies upon conjoint
analysis and (2) a decision analytic method based on insights from multiattribute utility
theory.
The first paper, Five propositions for improving decision making about the
1

environment in developing communities: Insights from the decision sciences , presents
the theoretical basis for the research. Decisions focused on managing natural
resources in developing areas present some of the most imposing challenges to policy
makers, scientists, and stakeholders alike. The response of policy makers and the
technical community in the face of these challenges has been significant. However, I
argue that the lack of attention devoted to supporting the underlying processes used to
make these complex decisions is cause for real concern. To this end, I present five
propositions based on work in the decision sciences that I believe stand out as essential
for improving decision-making processes in developing communities.

1

This chapter is currently in press in The Journal of Environmental Management.
2

The second paper, A choice experiment to address environmental, health, and
economic development conflicts about pineapple production in Costa Rica, discusses
the results of the stated choice (SC) treatment conducted in the Atlantic Region of
Costa Rica where decision makers and other stakeholders are faced with the need to
confront challenging tradeoffs regarding pineapple production and environmental
management.
A stratified random sample of 451 heads of household responded to a conjoint
survey in which they were asked to choose among alternative pineapple production
scenarios so that I could estimate weights associated with each of the presented
attributes. Consistent with previous work on SC methods, these weights served as
proxies for the tradeoffs across attributes (and associated resource management
objectives) that people were willing to make.
The primary objective of this research was to help inform policy decisions by
eliciting preferences for competing management scenarios from people living near
pineapple plantations. Specifically, I sought to determine: 1) whether residents, when
given a choice, would prefer current production practices or an alternative to the status
quo; 2) if residents prefer alternatives to the status quo, what are the attributes and
levels that most strongly influence their selection of an alternative?; and, 3) what
tradeoffs—if any—would residents be willing to accept across various aspects of
production?
The analysis suggests that most residents would prefer an alternative to the
status quo if given the choice; that three of the six pineapple production attributes
appear to most strongly influence community members’ selection of an alternative,

3

although there is variation depending on where they live; and that residents are, for the
most part, willing to make tradeoffs in terms of the kind and magnitude of improvements
in pineapple production scenarios that they would find acceptable in order to see even a
moderate reduction in negative environmental and human health impacts.
The third paper, Structuring decision making in developing communities:
Informing land use policy in Costa Rica, discusses the results of the deliberative
treatment carried out with 95 heads of household in the Limón Province of Costa Rica.
Just like the participants in the SC study presented in paper 2, these respondents were
also asked to evaluate alternative pineapple production scenarios, but they did so using
a deliberative analytic swing weighting method.
Decades of research in the decision sciences tell us that these kinds of
multiattribute decisions should be extremely challenging for decision makers. However,
recent studies of structured decision making (SDM)—and, indeed the research reported
in this paper—suggest that the quality of stakeholder input can be improved by ensuring
that people address a series of basic principles relating to the clarification of objectives
and their associated attributes, and directly confronting tradeoffs that arise during the
evaluation of management alternatives when these objectives conflict.
Therefore, the objectives of this research were twofold: First, I sought to help
inform policy decisions by eliciting land management preferences regarding the
pineapple industry from people living in communities surrounding plantations. Second, I
also studied the effectiveness of the SDM approach in a developing community context
with a specific focus on whether this method might hold promise as a decision aid to

4

practitioners and policy makers working in other international and community
development contexts.
The results suggest that although statistically significant differences appear to
exist across respondents in terms of their priorities, these differences are not strong
enough to impact the selection of preferred pineapple production scenarios. That is, the
data indicate that most community members would find the same production scenario
acceptable, even though they would all be required to make slightly different tradeoffs in
order to arrive at the this decision. The experience from this research indicates that the
SDM approach holds great potential for supporting complex decision-making processes
in Costa Rica and other developing community contexts.
In addition to answering the policy oriented research questions presented in the
second and third papers, another goal of my dissertation research is to better
understand and improve the quality of decision-making processes in international
development settings that increasingly must account for multiple stakeholders,
conflicting objectives, and tradeoffs across development and environmental objectives.
Both SC and deliberative tradeoff analysis methods have strengths and
weaknesses in developing community contexts, therefore, my next step will be to
analyze these pros and cons from a methodological perspective and make
recommendations for integrating more deliberation into conjoint surveys, and making
deliberative methods like the swing weighting exercise described here, a more efficient
data collection tool.

5

CHAPTER 1
FIVE PROPOSITIONS FOR IMPROVING DECISION MAKING ABOUT
THE ENVIRONMENT IN DEVELOPING COMMUNITIES: INSIGHTS
FROM THE DECISION SCIENCES

1. Introduction
Decisions focused on managing the natural environment, and the resources within it,
present some of the most imposing challenges to policy makers, scientists, and
stakeholders alike. Ecological systems, by their very nature, possess a high level of
structural complexity. In some cases, management interventions aimed at addressing
one aspect of a system will inadvertently affect many other components of the system
leading to unintended—and sometimes unwelcome—consequences. In other cases,
components of the system interact over large spatial or long temporal scales making it
difficult to fully understand, or even recognize, interconnected phenomena. Making
matters even more difficult, many ecological systems have endured repeated insults
that have chipped away at their resiliency; as a result, there is the very real risk that
certain management interventions may “break” some systems leading to irreparable
collapse. All of this leads to a high degree of uncertainty surrounding ecological
systems with respect to how they may—or may not—respond to management
interventions.
Adding to these challenges, decisions about the environment are not simply
decisions about the environment; they affect individuals, communities, business and
commercial interests, and in many cases, entire nations. As a result, these decisions
must contend with multiple, conflicting, and frequently poorly understood values that, in

6

turn, are linked to social, cultural, spiritual, economic, as well as other objectives or
concerns. Complicating matters further is the fact that these decisions must often be
made under severe time pressure, either because of impending environmental damage
(e.g., as in the case of an acute stressor such as contamination) or because of political
constraints faced by decision makers (e.g., the need to act under a given mandate).
2

These challenges are particularly significant in developing communities , where most
decisions about the environment directly influence the livelihood of one or more
stakeholder groups. Due to the precarious socio-economic status of people in these
areas, these direct—and even indirect—influences can have devastating effects on the
ability of resource-dependent individuals and communities to sustain themselves.
For example, one of the countries in which we work, Costa Rica, takes great
pride in being one of the 20 most biologically diverse countries in the world despite
covering just 0.03% of the earth’s surface. The country is also noted for its dedication
to natural resource conservation, which has resulted in the establishment of numerous
environmental laws and 160 continental protected areas. But, despite Costa Rica’s
environmental riches, its educated citizenry, and its proactive government, a series of
poor decisions regarding land use and resource development over the past several
decades have led to persistent environmental degradation in the form of significant
problems with erosion, water contamination, and the loss of biodiversity. In turn, these
problems have led to increased poverty and public health concerns in agriculture-

2

We use the term developing communities intentionally to encompass both developing
countries as well as communities within developed nations (e.g., the Inuit in the
Canadian Arctic) that, because of their local or cultural customs, or socio-economic
situation, are largely detached from their post-industrialized parent nations.
7

dependent rural communities, while simultaneously threatening overall ecosystem
health (Ballestero 2006; Sauma 2007).
We have encountered similar challenges when working in arctic communities. In
the Nunavut Territory of Canada for example, climate change has resulted in visible
physical changes that include increased melting rates of glaciers, sea ice, and
permafrost. These changes have led to dramatic declines in arctic vegetation,
indigenous wildlife, and—as a result—the health and safety of residents (as the stability
of traditional food sources and built infrastructure erodes). When combined with
pervasive poverty, extreme isolation, and a desire to maintain cultural traditions
developed over centuries, resource management decisions in the north have taken on
an aura of virtual impossibility.
The response of policy makers and the technical community working in these
developing areas—including ecologists, environmental scientists, economists,
sociologists, political scientists, and many others—has not been trivial. For example,
much has been written—and indeed, much work has been done—to address the need
for more comprehensive (in terms of how they are derived and understood) sciencebased inputs that better address key uncertainties and inform and improve the quality of
decision-making processes in developing communities (National Research Council
2007). Development economists have worked to understand, and in many cases to
help construct, conventional and contingent markets so they can better obtain the
values that affected stakeholders associate with environmental degradation and
protection (Whittington 1998). And, other social scientists have worked to bridge the

8

gap between these technical assessments and communities by promoting participatory
approaches as a means of improving environmental decision making (Ostrom 1996).
There is no doubt that these are necessary and important steps toward providing
both higher quality data inputs to decision makers and a more transparent, participatory
backdrop for decision making. However, our experiences as researchers and
facilitators of group planning and decision-making processes leave us concerned about
the lack of explicit attention devoted to supporting the decision-making process itself.
While some high-profile efforts have taken up this issue from a theoretical perspective
(EPA 2000; NRC 1996; NRC 2008), many important questions and obstacles persist.
For example, very little attention has been given to developing cross-disciplinary
methodologies that would help to focus technical analyses and their associated
outcomes on the stated needs of decision makers and stakeholders involved in
resource management consultations. Likewise, there has been only limited work on
how this information is used (or misused) by decision makers when designing
management alternatives or when informing tradeoffs when management objectives
conflict. And, importantly, there have been few comparative studies of competing
decision support approaches.
We view the barriers to achieving more effective processes for combining
analysis and deliberation through the lens of the decision sciences. From this
perspective, many of the problems typically encountered as part of unsuccessful
resource management processes (involving both expert and non-expert participants)
stem from the absence of a cohesive framework that brings together often disparate
technical and social threads during the process of decision making. Such a framework

9

helps diverse participants to clarify and express their issue-specific values and
concerns; have a voice—and, if necessary, a hand—in setting technical agendas aimed
at characterizing resource management problems and their potential solutions; be
involved meaningfully in the development of a recommended resource management
alternative (or alternatives); and think carefully about the costs and benefits of
competing alternatives. At the same time, such a framework helps decision makers to
account for and confront well-known cognitive limitations and judgmental biases that are
associated with highly complex and uncertain management contexts.
Using these concepts as a guide, this paper presents five propositions based on
work in the decision sciences that we believe stand out as essential for improving
decision-making processes in developing communities. These propositions, each one
building on its predecessor, are based on findings from research and practical work that
has taken place in a variety of developing communities and across a wide range of
resource management contexts.
2. Five propositions for improving decision making in developing communities
2.1 Decision-making processes must account for the constructive nature of preferences.
One of the assumptions widely held by social scientists and policy analysts is that
people possess a pool of pre-existing preferences that they consult during the process
of making judgments. Just as people are assumed to have preferences for things like
foods, drinks, pets, and political parties, they are assumed to also have preferences
about alternative strategies for economic development and environmental protection.
Indeed, survey research—whether based in sociology, political science, economics, and
related disciplines—would not be possible were it not for the assumption that these

10

kinds of preferences preexist and can be consulted by respondents during elicitation
processes.
It is true that, in a variety of contexts, such preexisting preferences can indeed be
indentified. For example, developmental psychologists have shown that people spend
their lives establishing preferences based on past experiences. Some of the resulting
and well-established preferences tend to be quite broad like preferring dogs as pets to
cats. Others are much more specific, like preferring greyhounds to beagles. Moreover,
these established preferences can quite easily be consulted during the processes of
judgment and decision making (Slovic & Lichtenstein 2006).
However, recent research in the decision sciences has demonstrated that there
are also many situations where existing preferences or preference orders needed to
inform required decisions are insufficient or absent altogether (Arvai et al. 2006b;
Gregory & Slovic 1997; Slovic & Lichtenstein 2006). Generally, these decision contexts
share one or more of the following three characteristics: First, the decision context may
be foreign with the implication that preexisting preferences simply cannot exist. In this
case, it is impossible for a decision maker to consult preexisting preferences for
decision problems that they are unfamiliar with or for alternatives that they have never
been exposed to before. Second, decision makers may be faced with the relatively
common situation in which the evaluation of competing alternatives causes two or more
preexisting preferences to conflict. In evaluations of resource management plans in
developing communities, for example, it is not uncommon for one alternative to be
particularly strong in terms of promoting economic development opportunities but weak
in the area of protecting the environment. If other alternatives under consideration

11

provide the opposite suite of benefits and costs, then decision makers who would
otherwise prefer alternatives that provide opportunities for economic development and
environmental protection will be required to confront a challenging and potentially
uncomfortable tradeoff. Third, people tend to have a difficult time translating qualitative
expressions of preference into numeric ones. While it may be the case, for example,
that some decision makers strongly prefer management alternatives that emphasize
environmental protection, how would they express these preferences on a Likert scale?
Alternatively, how would a relatively weak preference translate into decision makers’
willingness to pay for programs that promote environmental protection?
Under these very common conditions, people are unable to evaluate decision
problems and alternatives by simply drawing upon preexisting and stable preferences.
Instead, they must construct their preferences—and by extension, the decisions that
result from them—on the spot, largely in response to cues that are available to them
during the decision-making or elicitation process itself (Payne et al. 1993). From this
point of view, deliberative processes convened by researchers and development
practitioners—be they experimental or practical, or employed by individuals or groups—
have the de facto purpose of serving as architects of judgment and decision making
rather than as tools for simply revealing, as would an archaeologist, a person’s
preexisting preferences.
The implications of preference construction for decisions about the environment
in international development contexts—and elsewhere—are far reaching. But at the
most basic level, this view of constructed decision making calls into question the
generally accepted tenet in environmental management that better science (i.e., more

12

rigorous qualitative and quantitative findings from a wide spectrum of technical
disciplines) and broader public participation will yield higher quality decisions. We won’t
argue against the philosophy behind these guiding principles; indeed, better science
should be preferred to worse science just as broad participation should be preferred to
an absence of it. However, neither of these ingredients account for what happens
within the decision-making process itself. In other words, in addition to better
information and deliberation, more attention must be paid to what happens within the
process of decision making itself.
Facilitators (i.e., researchers, policy analysts, development practitioners, etc.) of
decision-making processes must, therefore, do a better job of accounting for how both
information about problems and deliberative processes are used—or misused—during
the construction of preferences. And, recognizing that decision makers rely heavily
upon contextual cues that are available to them during elicitation procedures as they
construct preferences, it is the responsibility of facilitators to provide the best possible
context—or structure—for decision making (Arvai et al. 2001; Gregory et al. 2001).
Essentially, we believe it is the role of those who facilitate decision-making processes to
help people construct the highest quality preferences—i.e., preferences that are well
informed, deliberated, and defensible—that are possible under given resource
constraints (e.g., time, information, cognitive or computational abilities).
It is worth noting that many advocates of broad-based participation worry that too
much structure will lead to biased input and unnecessarily constrain the breadth of
public involvement (e.g., see a review by Fischhoff 2005). While we agree, we would
also argue that more is needed than just an invitation for interested parties to

13

participate, respond to information, and deliberate. Such an approach, which is typical
of many participatory environmental management efforts, possesses substantial
shortcomings in terms of helping people to construct thoughtful and comprehensive
preferences in complex or unfamiliar contexts. For example, decision researchers have
long demonstrated that, in these kinds of loosely structured situations, both individuals
and groups grapple with a relatively predictable set of difficulties when making complex
decisions. Many relevant examples come from work on framing effects (Tversky &
Kahneman 1981), prospect theory (Kahneman & Tversky 1979), economic valuation
(Kahneman & Knetsch 1992), heuristics and biases (Tversky & Kahneman 1974), affect
(Wilson & Arvai 2006), and uncomfortable tradeoffs (Tetlock et al. 2000). One of the
fundamental conclusions from these lines of research is that people often end up
making decisions that—at best—only partially address the full range of their concerns
because they do not fully consider or comprehend information that is available to them
and, subsequently, fail to confront required tradeoffs when evaluating competing
alternatives.
Such findings suggest that along with the provision of information about a
problem coupled with opportunities for participation and deliberation, an overall
framework for decision making is needed to help structure relevant information so
participants can better understand often complex issues (Arvai & Gregory 2003). A
comprehensive decision making framework focuses on helping people to address one
3

or more of five basic needs, with each individual element supporting the next, as
required by a specific decision context. These are: (1) clearly defining the decision
3

We will discuss these further in subsequent sections.
14

problem that is to be the focus of analysis while taking into account the boundary
conditions under which the decision must be made; (2) identifying objectives that will
guide the decision-making process; (3) creating logical and internally consistent
alternatives that directly address these objectives; (4) establishing the predicted
consequences that are associated with alternative courses of action; and (5) confronting
inevitable tradeoffs when selecting among alternatives (Hammond et al. 1999; Keeney
1992).
Some decision contexts in developing communities—e.g., comprehensive
tourism planning efforts (e.g., see McDaniels & Trousdale 1999) and the design of
payment for environmental services (PES) programs (Engel et al. 2008; Pagiola
2008)—will likely require explicit attention to all five elements of this framework. Other
more tightly bounded (but not necessarily simpler) decision contexts, such as the
application of contingent valuation methods and social surveys will likely require only
certain component parts (e.g., clarifying objectives and addressing tradeoffs).
Regardless of the context, it is our view that without explicit knowledge of and attention
to these details during decision making, input from people—including interested and
affected stakeholders, policy makers, technical experts, etc.—may not accurately reflect
their carefully constructed preferences and judgments.
2.2 The basis for development decisions must be realistic and relevant.
Policy makers and researchers increasingly recognize the need to conduct development
work in a way that engages local decision makers and stakeholders in a more open and
participatory manner for both normative and instrumental reasons (Dietz & Stern 2008).
But despite this growing awareness—which has contributed to and is supported by a

15

rich interdisciplinary literature—most community development decisions continue to be
based on the perceptions and objectives of outside researchers, practitioners and
donors (typically foreigners from developed countries or nationals who are unaware or,
worse, disrespectful of the local context), instead of the needs of local stakeholders and
managers. This discrepancy, in turn, leads to decisions that are largely insensitive to
on-the-ground constraints and realities (e.g., regarding the specific nature of problems,
management objectives, alternatives, and desired tradeoffs). Often, the result is
development decisions that are short-lived or ineffective because they are unrealistic or
irrelevant to those for whom resources are being managed. In practice, these failures to
adequately address local realities and needs may transform well-intentioned efforts that
have real potential for effecting meaningful and positive change into projects that
ultimately fall well short of meeting this goal.
First, development decisions risk being unrealistic or irrelevant if decision makers
fail to appropriately define the scope of decision problems (or opportunities) or
recognize the boundary conditions under which they must be made, implemented, and
monitored. It’s one thing to approach development decisions with high ambitions about
protecting a given resource and the communities that rely upon it. However, to do so
without accounting for local legal, political, economic, institutional, ecological, or cultural
constraints represents a significant barrier to successful implementation, monitoring and
enforcement.
This is not to suggest that high aspirations during decision making for community
development should be blindly curtailed. Work by other decision analysts in developing
communities (Gregory & Keeney 1994; McDaniels & Trousdale 1999) has been explicit

16

about the need to think carefully about the scope of problems and openly question
apparent constraints in terms of how decision problems are defined or bounded.
Particular emphasis has been placed on challenging and expanding problem
statements that needlessly narrow the range of alternatives that managers and
stakeholders might consider. The same is true of expanding resource management
initiatives to account for both the full spectrum of stakeholders that should be involved in
development decisions and linked decisions, which are those that may influence parallel
or future decision-making processes. In our experience, a failure to include key
stakeholders or consider linked decisions, which might be folded into ongoing
development efforts with relative ease, often results in a cascade of other significant
problems that may appear only after time-limited development efforts have come to an
end.
At the same time, however, it is crucial that managers and local stakeholders
determine the appropriate scope and scale for problem statements that will be the focus
of decision-making efforts. Doing so means ensuring that the problem statement that is
to be the focus of development decisions is relevant to the people who will be affected
by it and, importantly, realistic given the resources that can be brought to bear on
addressing it. Development decisions must also be robust enough to survive and thrive
in light of political changes—e.g., changes in local, regional, or national leadership—that
tend to occur both quickly and often in many developing areas, especially at the
regional level. If these conditions cannot be met, managers should be prepared to
either allocate additional resources (e.g., time, money, expertise, etc.) to the decisionmaking process or simplify it by reducing its scope.

17

Alternatively, an overly narrow problem statement can just as easily render
technically sound resource management efforts unrealistic, and therefore, ineffective.
This often occurs when the management problem is developed in the absence of key
stakeholders. The resulting lack of diverse knowledge tends to allow the problem to be
oversimplified to the extent that resource degradation is attributed to one sector or
group of people. This overgeneralization of what is usually a complex issue makes it
difficult for stakeholders to address the problem in a holistic manner. In Costa Rica’s
Osa Peninsula, for example, resource managers dedicated to protecting biodiversity
originally viewed the local population’s resource and land use practices as the principle
threat to the area’s flora and fauna. As is typical in this kind of situation, the overly
narrow problem definition (i.e., deforestation, poaching) contributed to the exclusion of
local people from the decision-making process that was intended to respond to threats
to biodiversity. The absence of these legitimate stakeholders from resource
management efforts was subsequently seen as unproductive and ethically
unacceptable, and efforts have been made to improve this situation. In the meantime,
however, many opportunities to both make the problem statement more realistic and
address multiple linked decisions related to the issue of biodiversity protection (i.e., rural
economic development, land use planning, cultural preservation, local knowledge) have
been missed (Jimenez-Elizondo 2004).
Development decisions also risk being unrealistic or irrelevant if they do not
respond directly to stakeholders’ problem-specific concerns and related objectives.
Being clear about the objectives that will drive management efforts will help to identify
those who should be involved in the decision-making process. In most consultations,

18

for example, it will be important to include both local residents and technical experts
(who might be from industry, government, or the academy) and to include
representatives of a wide range of viewpoints. To the extent that key participants are
omitted, critical concerns and objectives are likely to also be ignored. The scale of the
initiative also matters. Selecting participants for a small local project may be relatively
easy. But, for a larger, more complex undertaking—such as one with national or
international implications—some limits will need to be placed on the set of participating
stakeholders (Gregory et al. 2001).
Thorough elicitations of management objectives from key experts and
stakeholders also helps to differentiate between means and ends. Ends objectives are
those that fundamentally matter in the context of the decision-making process. Means
objectives, by contrast, matter primarily because they influence one or more of the ends
objectives (Keeney 1992). These objectives can be visualized—e.g., using means-ends
networks, value trees, or influence diagrams—which can help to clarify the set of
concerns that are central to the majority of stakeholders. Even though the weights
placed on the various ends objectives—in other words, their relative importance to each
participant—may be quite different, the knowledge of these means-ends relationships
helps to focus and give direction to subsequent deliberations; this, in turn, can greatly
assist in the identification of a prioritized list of adaptation or mitigation actions (e.g.,
what to do first, what to do later, etc.).
Failing to accurately account for these objectives during decision making,
however well intentioned, may lead to significant problems for managers, development
professionals, and donor agencies for two main reasons. On the one hand, they stand

19

to lose credibility and local support because their work is judged as irrelevant in many
communities. On the other, resources may be wasted on addressing problems that
miss the mark in terms of what matters most to affected stakeholders.
During one of our projects in eastern Costa Rica, for example, we learned from
local stakeholders about a recent U.S. university-funded study that focused on
estimating the economic value (e.g., to inform the design of a PES initiative) of
protecting community access to drinking water sources. The objective that guided this
work—protecting community access to local water sources—had been identified by a
well-intentioned foreign research team without consulting local stakeholders. Only at
the end of the project did it become clear that while access was of concern to a very
small group of residents, it was largely irrelevant to the vast majority of people in the
community. Most people enjoyed easy access to water but were much more concerned
about its safety as a result of what they perceived as unacceptably high and
unregulated levels of pesticide runoff from adjacent pineapple plantations. The
research team in the field realized, too late, that they were not helping local
stakeholders to address their primary objective, which was to ensure clean and safe
drinking water. Community members, in turn, did not hesitate to express their
frustration that resources (including money, their time, and expertise) were being spent
on an issue that they considered to be of secondary importance. And importantly for
the local community, the proposed PES initiative—which could have been designed to
address the objective of safe drinking water—was abandoned.
In addition to being an inefficient use of scarce resources, these kinds of
breakdowns in decision making also demonstrate the practical difficulties that arise

20

when insufficient attention is paid to the knowledge, concerns, and objectives of local
stakeholders. Importantly for developing communities, these breakdowns in process
can lead to further harm to already vulnerable people, communities, and ecosystems.
2.3 Expand deliberative processes to account for the expertise of local communities.
In our experience, the problems described above occur to a large degree because
foreign researchers, practitioners, and donors working in developing communities rely
too heavily on their own beliefs, expertise, and past experiences when diagnosing—and
then addressing—resource management problems. This is not to belittle the
knowledge, ideas, or past work of these individuals and groups. But, while their
expertise can be valuable—and their resulting ideas and insights well intended—outside
experts frequently fall short of directly addressing the problem-specific concerns of local
people and communities. As we note above, one way to address these shortcomings is
to consult meaningfully with community stakeholders during the early stages of project
development. But beyond working with communities to fully define decision problems
and elicit objectives, recent work in the risk and decision sciences points to the
importance of expanding the definition of “expertise” in community development efforts.
Doing so levels the playing field between local people and those with specific technical
training, and accounts for the value and community-specific expertise of local
communities.
One way to achieve this goal is to engage community stakeholders in decision
making in a way that moves well beyond the consultative role advocated by earlier,
influential studies on deliberative democracy (Arnstein 1969; Renn et al. 1995) and
more recent work—which we outline above—on eliciting objectives that will guide

21

decisions. The US National Research Council (NRC) has written thoughtfully about this
issue, framing it as a matter of implementing what it terms an “analytic-deliberative”
process. The premise of the NRC’s analytic-deliberative framework is relatively
straightforward with community members and scientists (social, behavioral, physical,
natural, etc.) working together to jointly define decision problems and, importantly, set
the research agendas that will be used to develop the information base aimed at
addressing them (National Research Council 1996).
At the start of an analytic-deliberative process, preliminary data describing the
current state of an at-risk human-natural system informs an initial phase of inclusive
deliberations aimed at helping to define the overall decision problem (expressed in
terms of the likelihood or extent of harm to the system as well as its measured or
predicted consequences) to be assessed and managed. This first round of deliberation
then provides important and much-needed insight to the scientific community and
analysts about ways in which the ongoing assessment—including methods for data
collection and, to the extent it is possible, the attributes and measures used to
characterize incoming data—and its subsequent interpretation for decision making
ought to take place. This, in turn, provides the information on which to base these and
future decision-focused deliberations (National Research Council 1996).
One of the frequently ignored aspects of decision making is operationalizing the
objectives that will guide a given choice. It is of little help to a decision maker to
express an objective—such as improving the health of the environment or enhancing
the stability of communities—without having a very clear and appropriate sense of
exactly how to measure it. In order to complete this important step, decision makers

22

must identify the attributes of the objective that will be the focus of measurement (i.e.,
data collection). The process of establishing the attributes and related measures that
will be used to evaluate the performance of alternatives requires analysts, stakeholders,
and decision makers to consider what kinds of data exist or could be obtained, and
where information is lacking or highly contested. It is at this point that evaluations,
which should focus on information to support decisions, can—and often are—confused
with assessments of “expert knowledge”. This confusion typically results in the
designation of attributes and measures based more on what kind of data and inputs are
readily available, easy to obtain, or familiar rather than on the insights of stakeholders
and decision makers regarding what kinds of information could best help them to
evaluate how well proposed alternatives address stated objectives (Keeney & Gregory
2005).
As the complexity and uncertainty surrounding a risk to be managed increases,
so to does the degree to which the analytic-deliberative process becomes recursive. As
envisioned by the NRC, a typical analytic-deliberative process would proceed through
several successive rounds of deliberation and analytic work by stakeholders as part of a
purposeful march towards a management decision. Each round of analysis and
deliberation is meant to yield an improved understanding of both stakeholders’
objectives as they relate to a given decision problem and anticipated responses of a
system to alternative management options. In this sense, the analytic-deliberative
process is not simply a means for synthesizing the information obtained through a set of
unrelated risk assessments; it is an important shaper of a long-range risk assessment
and decision-making process (National Research Council 1996).

23

The analytic-deliberative process has been endorsed widely by researchers and
practitioners; the four common findings that underlie most endorsements are that: (1)
The results of the associated technical analyses will be more decision-relevant insofar
as they are framed in terms of measures and attributes that make the most sense to,
and are most desired by stakeholders and decision makers; this presumably makes it
easier for decision makers and interested and affected parties to recognize and respond
to changes within a system; (2) The process fosters greater trust in the overall risk
management process; this includes greater trust by stakeholders in the analytic
community and greater respect on the part of the analytic community for the level of
sophistication in thought displayed by stakeholders and decision makers; (3) Analyticdeliberative processes lead to a higher degree of learning about the social and technical
drivers of at-risk systems by all of the parties involved—stakeholder and analyst alike—
and by extension, work to reduce uncertainty (also see Section 2.5, below); and (4)
Analytic-deliberative processes lead to more defensible and thus, higher quality risk
management decisions insofar as they are specific to a well-defined problem,
responsive to the objectives and concerns of stakeholders, and informed by decisionrelevant science.
The net result is that, within this analytic-deliberative framework, decisionrelevant “expertise” is not defined by years of training or the highest degree held by an
individual. Instead, expertise is reflected in the ability of people—be they community
stakeholders or outside analysts—to contribute meaningfully to the setting of objectives
that will guide decisions as well as the attributes of these objectives that will help
decision makers to distinguish between alternative courses of action.

24

Importantly, this definition of expertise recognizes the importance and relevance
of local and traditional ecological knowledge, which refers to the collective and
indigenous wisdom about the local environment which is gained by people over
generations (Berkes 1999). Nearly all developing areas are home to indigenous or
long-standing populations in addition to more recently established local stakeholders
who bring with them their own traditional knowledge and experience. It is well
documented that these individuals and communities often hold important insights with
regard to understanding ecological processes and determining effective resource
management options based on their unique environmental and socio-economic
knowledge and experience (Wynne 1992). This acknowledgment of the importance and
relevance of traditional ecological knowledge, we believe, represents an encouraging
step forward in terms of improving natural resource management decisions that are
inextricably linked to long-term environmental resilience and the socio-economic wellbeing of vulnerable communities.
We acknowledge that, in many cases, analysts and development practitioners
may resist the idea of incorporating traditional ecological knowledge alongside inputs
from more mainstream, western science. Indeed, we have seen firsthand—e.g., in
Costa Rica, Ecuador, Tanzania, and Nunavut—many cases where decision makers and
program managers have been vocal in their concern about the credibility of traditional
knowledge. However, it has also been our experience that multi-stakeholder decisionmaking processes are flexible enough to account for the inclusion of objectives,
attributes, and measures that are based on insights from traditional knowledge holders.
Indeed, including information based on traditional knowledge often serves as an

25

important reality check for decision makers in that it helps to verify objectives used to
guide community development processes. If, for example, information provided by local
or traditional knowledge holders cannot be matched with a specific objective, then we
can be reasonably certain that some objectives—which might otherwise help to guide
the decision-making process—have yet to be considered. Moreover, it has been our
experience—as well as that of others (Berkes 1999; Kerr & Sanghi 1992)—that
traditional sources of knowledge often provide the best and sometimes only source of
information about historical environmental and cultural conditions, as well as the desired
balance between community development and environmental protection.
2.4 Rethink the goals of participatory decision-making processes.
It is difficult to find a researcher or practitioner who disagrees with the principle of
involving local stakeholders during decision making for international development – or
any other decision of consequence for that matter. Several texts have focused
generally on this aspect of international development (e.g., Chambers 1983; Fischer
2002). Specific research and practitioner-oriented projects have focused on an array of
community-based approaches with examples of these falling under the banners of
action research (McTagart 1997), participatory rural appraisal (Chambers 1994), comanagement (Baland 1996), and integrated watershed management (Rhoades 1998).
But despite a rich literature focusing on the importance of stakeholder-based decisionmaking approaches, many examples of practice often fall short of meeting expectations.
In our view, there are two related reasons behind these shortcomings.
First, many stakeholder engagement initiatives proceed without a clear and
decision-focused expression of the goals that participation is supposed to achieve. It is

26

frequently the case, for example, that stakeholder involvement in decision making is
undertaken because it is a procedural requirement by government, research, or donor
agencies. In many places, these stipulations are a well-intentioned response to
legitimate concerns about a historical lack of meaningful participation in important
policy-making processes. However, the guidance that typically accompanies these
requirements—engaging stakeholders early and often or paying attention to criteria
such as ensuring “fairness and competence” (Renn et al. 1995), for example—fails to
connect very good work that has been done in fields that often depend on stakeholder
engagement (e.g., rural development, sociology, anthropology, planning) with findings
from disciplines that focus exclusively on the process of decision making (e.g.,
behavioral decision research, behavioral economics, psychology).
This disconnect is apparent in the way participatory processes are typically
described in theoretical discussions (Grillo & Stirrat 1997; Li 2007), policy guidelines,
and case studies (Blauert & Dietz 2004; IIED 2008). Authors typically discuss
participation in terms of the way projects should be carried out (e.g., in terms of a
recommended set of guiding principals). While these discussions are valuable, they
rarely discuss how stakeholder engagement was or could have been structured to
provide the inputs that are required of internally consistent decision-making processes.
This lack of detailed guidance for decision makers and stakeholders about how to
implement decision-focused participatory processes inadvertently sends the message
that the biggest obstacles to successful implementation are (1) convincing decision
makers—either through encouragement or obligation—to utilize a participatory process,
whatever that might mean to them; and (2) getting resource managers, scientists,

27

decision makers and stakeholders to work together effectively (i.e., simply getting the
“right” stakeholders around the table is enough).
Second, many participatory processes over-emphasize conflict resolution and
consensus as necessary components of decision making. Such approaches assume
that conflict is bad for decision making, consensus is a prerequisite for resolving a
problem, and conflicts can be “resolved” and consensus is possible without coercion.
The rationale behind this viewpoint is understandable in that a focus on reducing conflict
and building consensus addresses long-standing inequalities (e.g., based on socioeconomic status, ethnicity, gender, etc.) that are pervasive between individuals and
groups. However, it has been our experience in the field that these assumptions often
exacerbate the very problems that conflict resolution and consensus processes are
thought to address (i.e., power inequalities; stakeholder marginalization).
We have been involved in several cases, most recently in Latin America and the
arctic, where a focus on consensus has shifted—in some cases subtly; openly in
others—key elements of group decision-making processes. For example, stakeholders,
technical experts, and consultants have been invited to take part in these processes
(while others have been excluded) on the basis of their ability to agree with one another
rather than on criteria relating to their possession of required expertise or expressions of
a diverse range of objectives. Likewise, problems and alternatives that were the focus
of analysis were chosen, and intractable issues ignored, based on their potential to
foster even short-term or shaky agreements. Methodologies for establishing the
predicted consequences of problems, or management plans for addressing them, had
been selected based on the extent to which they were easy to implement or document

28

rather than on their ability to answer fundamental questions or lend insight to impending
choices. And importantly, we have observed cases where minority views within groups
have been suppressed rather than explored, with conflict among group members
viewed as a problem to be overcome or avoided rather than as an opportunity for
providing additional clarity regarding the decisions at hand.
The previous discussion regarding the construction of preferences (Section 2.1),
elicitation of objectives (Section 2.2) and then establishing appropriate attributes and
measures for them (Section 2.3) begins to speak to our views about both the roles that
stakeholder participation should play and how participatory decision-making processes
should be structured. Our view mirrors the theoretical underpinnings of decision-aiding
approaches (Arvai et al. 2001; Gregory et al. 2001)—i.e., in behavioral decision
research and decision analysis—as compared to the basis for much of the current
thinking and practice in environmental management and consultation in disciplines such
as alternative dispute resolution and risk communication. In drawing on insights from
work on decision aiding, we also believe that the goals of stakeholder participation
should be expanded to include a formal exploration of conflicting objectives—which are
reflected in the tradeoffs that stakeholders are willing to make—when people are asked
to choose one alternative course of action over another.
Other researchers and practitioners share our views. Development economists,
for example, advocate the use of stated choice methods (Adamowicz et al. 1998) to
learn about the acceptability of certain tradeoffs from local community members and
other stakeholders. In most cases, paper-and-pencil surveys are used to present
respondents with a series of alternatives, which are composed of integrated (i.e.,

29

conjoined) combinations of different attributes (e.g., different levels of air quality, water
quality, and local employment). These combinations generally reflect actual or
projected variations in the attributes (e.g., different levels of air and water quality and
local employment opportunities). In the more sophisticated conjoint surveys, often
termed “choice experiments”, the represented attributes are specified by an
experimental design that estimates the separate and interacting effects of component
attributes. Multiple regression or similar analyses of decision makers’ choices among
the presented options are used to estimate the relative contributions of individual
attributes to the expressed preferences or other judgments for the conjoint alternatives.
Essentially, the choices made by people in response to survey questions are used to
reveal the tradeoffs that they are willing to make when comparing multiattribute
alternatives.
While these approaches represent a step in the right direction, they have been
criticized on the grounds that the resulting insights about tradeoffs reflect poorly
constructed judgments. In other words, information about implied tradeoffs may be
misleading because the judgmental processes that led to them were prone to
judgmental inconsistencies. As we point out in Section 2.1, this problem is likely to be
especially significant when the choice context is unfamiliar, as would be true for many
environmental policy options. With stated choice approaches, for example, questions
may be raised about the extent to which survey respondents are actually confronting
tradeoffs across all of the attributes and levels that are being presented to them in
choice experiments. If they are not—e.g., if they use a decision rule such as satisficing
(Simon 1955) and focus only on a small subset of attributes—the implied tradeoffs,

30

which assume that all of the available attributes are being considered, may not
accurately reflect what matters to decision makers. This in turn leads to spurious
conclusions for policy makers and managers.
To address this problem, decision scientists have developed alternative methods
for tradeoff analysis that can either be used to supplement conjoint surveys or on their
own when people are asked to evaluate a small set of discrete alternatives. All of these
approaches—swing weighting (Clemen 1996), even swaps (Hammond et al. 1998),
pricing out (Borcherding et al. 1991), etc.—require that respondents consider and then
weigh the relative importance of all attributes that differentiate the alternatives under
consideration. These weights are then used to help decision makers establish
preferences, or preference orders, across the options that are presented to them. A
detailed accounting of these methods is beyond the scope of this article. However, all
of these methods—including the conjoint methods deployed by economists—share in
the philosophy that the goals of participatory decision-making processes should be
expanded beyond engaging stakeholders in largely unstructured management
deliberations.
2.5 Avoid decision paralysis.
The amount of effort and expertise required to reconsider the manner in which complex
decision making approaches are structured is not trivial. Many development agencies,
such as The World Bank, have taken specific steps to build institutional capacity in the
area of decision support. But in the absence of the required effort and expertise, many
decision-making processes have nevertheless been undertaken where the participants
simply “muddle through” with varying degrees of success. In too many other cases,

31

however, important decisions have simply been put on hold, or abandoned altogether.
Frequently, policy makers put off making important decisions because of the high levels
of uncertainty and complexity that the required choices entail. They worry that they do
not know enough to come to a responsible judgment or that the decision context will
change in unpredictable ways, quickly outmoding a decision or rendering it
inappropriate.
In taking this kind of wait-and-see approach, decision makers effectively define a
responsible decision-making process as one that requires those who may be affected
by a choice to wait until it is possible to make a decision under greatly reduced, or even
better, zero uncertainty. To support this position, many decision makers invoke mantras
like “do no harm” or the relatively popular precautionary principle (Raffensperger &
Tickner 1999). But since all decisions about complex natural resource management
problems will include some degree of uncertainty, these approaches are often
unproductive and unrealistic because they do not account for the fact that in many
fragile areas—ecosystems and communities—failing to act can also cause harm.
Ultimately, what often goes unrealized in these cases is the fact that failing to
address a decision represents a decision in and of itself; i.e., people choose to avoid
making a given decision. For this reason, decision researchers have encouraged
people to evaluate discrete decision opportunities—to make a decision or to avoid
one—as alternative courses of action. In this sense, people should approach these
decisions in much the same way that we have discussed above: that is to identify
objectives to help guide the choice, make judgments about the consequences
associated with confronting (vs. delaying) a decision, and evaluating the risks and

32

benefits associated with choosing one option over the other. Supplementing these
principals, there should be an institutional commitment to long-term monitoring of both
decision-making processes and their resulting outcomes. By doing so, decision makers
and stakeholders will be able to proactively assess how problems, objectives,
alternatives, and tradeoff tolerances change as situations evolve. Likewise, they will be
better positioned to adapt deliberative processes to changing social, environmental, and
economic conditions.
Another useful way out of the trap of decision paralysis is to explicitly make
learning over time a fundamental objective of development decisions (Gregory et al.
2001; McDaniels & Gregory 2004). The concept of adaptive management was born
specifically out of the need to address this objective of learning from managed systems
over time (Arvai et al. 2006a; Walters 1986). The central argument of adaptive
management is that policy decisions are really research questions masquerading as
answers. By taking this view, the management of complex environmental and social
problems then can be regarded as a process of learning over time from policies
designed to reduce uncertainty and improve management’s ability to respond to
inevitable environmental, social, or economic surprises.
To operationalize this effort, adaptive management calls for the design and
monitoring of planned policy “experiments”, with replication and comparison of
management treatments at appropriate spatial and temporal scales. Rather than
making one-time decisions on the basis of the best existing knowledge, adaptive
management regards policy choices for complex environmental problems as part of an
overlapping sequence of iterative policy trials. At the same time, adaptive management

33

emphasizes monitoring and learning as the system changes, both in response to
external stimuli and in response to the manager’s actions (Walters 1986).
It is important to note that, while adaptive management proceeds based on the
philosophy of experimentation, the resulting trials need not proceed based on deliberate
manipulation by managers. While there have been several examples of what is termed
“active adaptive management”—e.g., within-season experimentation, complete with
control regimes and replication, in coastal fisheries (Walters & Holling 1990)— “passive
adaptive management” is also possible. Here, policy analysts monitor the results from a
series of spatially varied management experiments that are underway at the hands of
competing and overlapping sovereign political actors and institutions. In adaptive
management for climate change impacts, for example, policy analysts may track
different types of climate mitigation projects that have been undertaken in different
regions of the world (e.g., forestry projects in Latin America versus agricultural projects
in West Africa). In contrast to the active implementation of experimental treatments,
these spatially varied experimental probes are not the result of intentional manipulation.
Nevertheless, they represent varied policy treatments, the results from which can be
compared to test clearly formulated hypotheses about the behavior of complex systems
(Arvai et al. 2006a). Experimentation in this sense goes beyond management through
trial and error and casual observation; it is structured and theoretically driven, so that
new knowledge can be incorporated systematically into future treatments.
Overall, adaptive management presents an appealing framework for
strengthening the relationship between policy makers and stakeholders. While an
experimental focus is especially appealing to scientists, adaptive management reaches

34

well beyond the goal of simply enhancing technical understanding of natural systems
independent of human systems. The approach also recognizes that managed systems
present moving targets influenced largely by social drivers and, therefore, explicitly links
these human factors to management initiatives (Arvai et al. 2006a). The added appeal
of adaptive management, therefore, lies in its ability to help inform the judgments of
policy makers who must address complex problems with high levels of ecological,
social, and economic uncertainty.
3. Conclusion
While many examples of international development decisions have been widely
critiqued for their perceived failure to meet linked environmental, social and economic
goals (Agarwal 2001; Agrawal & Gibson 1999; Peet & Watts 2004), very little
meaningful guidance has traditionally been available to decision makers regarding how
best to address these concerns. The same is true of international and community
development efforts that are supposed to be “participatory”, but have been criticized for
being unnecessarily insulated from those who will be directly affected by decisions. In
the worst cases, a lack of effective collaboration between managers and stakeholders
has led to the further marginalization of vulnerable communities (Castro 2004;
Echeverría 2003; Jiménez 2005).
We have proposed that it will continue to be difficult for researchers, practitioners
and policy makers to address complex, multi-objective decisions without first
operationalizing decision-making processes in terms that go well beyond general
recommendations (e.g., involve all interested parties early in the process, build on
existing relationships, adopt a bottom-up management structure, etc.). Over

35

generalizing in this way has resulted in a failure to focus explicitly on several issues
central to higher quality decision making: recognizing and adapting to the constructive
nature of judgments; ensuring that the basis for decision making is both realistic and
relevant; expanding the definition of expertise to account for local and traditional
knowledge; adopting decision-focused stakeholder deliberation over public participation;
and proceeding with decision making in the face of complexity and uncertainty.
If we were to summarize this argument, it would be that decision-making ought to
be viewed as a process and not merely an occurrence. Just as improving
communication between managers and stakeholders involves more than bringing
people together and providing opportunities to hear opposing views, decision-making
processes require much more than the right combination of public participation, better
science, and lower uncertainty. Only when we accept that complex natural resource
management problems ought to be addressed through well-structured, deliberative
processes can we focus on developing decision support techniques to achieve more
defensible decisions.

36

REFERENCES

37

REFERENCES

Adamowicz, W., P. Boxall, M. Wilhams, and J. Louviere. 1998. Stated preference
approaches for measuring passive use values: Choice experiments and
contingent valuation. American Journal of Agricultural Economics 80:64-67.
Agarwal, B. 2001. Participatory Exclusions, Community Forestry, and Gender: An
Analysis for South Asia and a Conceptual Framework. World Development
29:1623-1648.
Agrawal, A., and C. C. Gibson. 1999. Enchantment and Disenchantment: The Role of
Community in Natural Resource Conservation. World Development 27:629-649.
Arnstein, S. R. 1969. A Ladder of Citizen Participation. Journal of the American Institute
of Planners:216-224.
Arvai, J. L., G. Bridge, N. Dolsak, R. Franzese, T. Koontz, A. Luginbuhl, P. Robbins, K.
Richards, K. Smith Korfmacher, B. Sohngen, J. Tansey, and A. Thompson.
2006a. Adaptive management of the global climate problem: Bridging the gap
between climate research and climate policy. Climatic Change 78:217-225.
Arvai, J. L., and R. Gregory. 2003. Testing alternative decision approaches for
identifying cleanup priorities at contaminated sites. Environmental Science &
Technology 37:1469-1476.
Arvai, J. L., R. Gregory, and T. McDaniels. 2001. Testing a structured decision
approach: Value-focused thinking for deliberative risk communication. Risk
Analysis 21:1065-1076.
Arvai, J. L., R. Gregory, D. Ohlson, B. A. Blackwell, and R. W. Gray. 2006b. Letdowns,
wake-up calls, and constructed preferences: People's responses to fuel and
wildfire risks. Journal of Forestry 104:173-181.
Baland, J. M., J.P. Platteau 1996. Halting Degradation of Natural Resources: Is there a
Role for Rural Communities? Oxford University Press, Oxford, UK.
Ballestero, M., and V. Reyes. 2006. Water Quality Management in Central America:
Case Study of Costa Rica in B. B. A.K. Biswas, C. Tortajada, and D.J.
Rodriguez, editor. Water Quality Management in the Americas. Springer, Berlin,
Germany.
Berkes, F. 1999. Sacred Ecology: Traditional Ecological Knowledge and Resource
Management. Taylor & Francis, London, UK.

38

Blauert, J., and K. Dietz. 2004. Of Dreams and Shadows: Seeking change forthe
institutionalisation of participation for natural resource management in M.
Pimbert, editor. Institutionalizing Participation. IIED, London.
Borcherding, K., T. Eppel, and D. von Winterfeldt. 1991. Comparison of weighting
judgments in multiattribute utility measurement. Management Science 37:16031619.
Castro, R., E. Monge, C. Rocha, H. Rodríguez. 2004. Gestión local y participativa del
recurso hídrico en Costa Rica. Centro de Derecho Ambiental y de los Recursos
Naturales (CEDARENA), San José, Costa Rica.
Chambers, R. 1983. Rural Development: Putting the Last First. Longman, London.
Chambers, R. 1994. Participatory Rural Appraisal (PRA): Analysis of Experience. World
Development 22:1253-1268.
Clemen, R. T. 1996. Making Hard Decisions: An Introduction to Decision Analysis.
PWS-Kent Publishing Co., Boston, MA.
Dietz, T., and P. Stern, editors. 2008. Public Participation in Environmental Assessment
and Decision Making. The National Academies Press, Washington DC.
Echeverría, J. 2003. Towards Integrated Water Resources Management: A case study
of the Lake Arenal Watershed. Pages 203-218 in G. W. P. C. America, editor.
Case Studies: Integrated Water Resource Management in Central America.
Masterlitho S.A., San José, Costa Rica.
Engel, S., S. Pagiola, and S. Wunder. 2008. Designing payments for environmental
services in theory and practice: An overview of the issues. Ecological Economics
65:663-674.
Environmental Protection Agency. 2000. Toward Integrated Environmental DecisionMaking. Science Advisory Board, Washington, DC.
Fischer, F. 2002. Citizens, Experts, and the Environment: The Politics of Local
Knowledge. Duke University Press, Durham, NC.
Fischhoff, B. 2005. Cognitive Processes in Stated Preference Methods in K. G. Maler,
and J. R. Vincent, editors. Handbook of Environmental Economics. Elsevier,
London, UK.
Gregory, R., J. L. Arvai, and T. McDaniels. 2001. Value-focused thinking for
environmental risk consultations. Research in Social Problems and Public Policy
9:249-275.
Gregory, R., and R. L. Keeney. 1994. Creating policy alternatives using stakeholder
values. Management Science 40:1035-1048.

39

Gregory, R., and P. Slovic. 1997. A constructive approach to environmental valuation.
Ecological Economics 21:175-181.
Grillo, R. D., and R. L. Stirrat, editors. 1997. Discourses of Development:
Anthropological Perspectives. Berg, Oxford.
Hammond, J., R. L. Keeney, and H. Raiffa 1999. Smart Choices: A Practical Guide to
Making Better Decisions. Harvard Business School Press, Cambridge, MA.
Hammond, J. S., R. L. Keeney, and H. Raiffa. 1998. Even swaps: A rational method for
making trade-offs. Harvard Business Review 76:137-138, 143-148, 150.
IIED. 2008. Towards a Global Forest Partnership: Consultation, assessment and
recommendations in I. I. f. E. a. Development, editor. IIED, London.
Jiménez, J., J. Calvo, F. Pizarro, E. González. 2005. Conceptualización de Caudal
Ambiental en Costa Rica: Determinación Inicial para el río Tempisque. Unión
Mundial para la Naturaleza (UICN), San José, Costa Rica.
Jimenez-Elizondo. 2004. The Politics of Participation in a Biodiversity Conservation
Project: Analysis of a Biological Corredor Project in Osa Peninsula, Costa Rica.
Department of Resource Development. Michigan State University, East Lansing.
Kahneman, D., and J. L. Knetsch. 1992. Valuing public goods: The purchase of moral
satisfaction. Journal of Environmental Economics and Management 22:57-70.
Kahneman, D., and A. Tversky. 1979. Prospect theory: An analysis of decision under
risk. Econometrica 47:263-291.
Keeney, R. L. 1992. Value-focused Thinking. A Path to Creative Decision Making.
Harvard University Press, Cambridge, MA.
Keeney, R. L., and R. Gregory. 2005. Selecting attributes to measure the achievement
of objectives. Operations Research 53:1-11.
Kerr, J., and N. K. Sanghi. 1992. Indigenous soil and water conservation in India's semiarid tropics. Sustainable Agriculture Program, International Institute of
Environment and Development, London, UK.
Li, T. M. 2007. The Will to Improve: Governability, Development, and the Practice of
Politics Duke University Press, Durham.
McDaniels, T., and R. Gregory. 2004. Learning as an objective within a structured risk
management decision process. Environmental Science & Technology 38:19211926.
McDaniels, T., and W. Trousdale. 1999. Value-focused thinking in a difficult context:
Planning tourism for Guimaras, Phillipines. Interfaces 29:58-70.

40

McTagart, R. 1997. Participatory Action Research: International Contexts and
Consequences. SUNY Press, New Yor, NY.
National Research Council 1996. Understanding Risk: Informing Decisions in a
Democratic Society. National Academy Press, Washington, DC.
National Research Council 2007. Strengthening Science-Based Decision Making in
Developing Countries. Agriculture and Water Management. National Academy
Press, Washington, DC.
National Research Council 2008. Public Participation in Environmental Assessment and
Decision Making. National Academies Press, Washington, DC.
Ostrom, E. 1996. Crossing the great divide: Coproduction, synergy, and development.
World Development 24:1073-1087.
Pagiola, S. 2008. Payments for environmental services in Costa Rica. Ecological
Economics 65:712-724.
Payne, J. W., J. Bettman, R., and E. J. Johnson 1993. The Adaptive Decision Maker.
Cambridge University Press, Cambridge, MA.
Peet, R., and M. Watts, editors. 2004. Liberation Ecologies: Environment, Development,
Social Movements. Routledge, London.
Raffensperger, C., and J. Tickner 1999. Protecting Public Health and the Environment:
Implementing the Precautionary Principle. Island Press, Washington, DC.
Renn, O., T. Webler, and P. Wiedemann 1995. Fairness and Competence in Citizen
Participation. Klewer Academic Publishers, Dordrecht, The Netherlands.
Rhoades, R. 1998. Participatory Watershed Research and Management: Where the
Shadow Falls. in S. A. a. R. L. Program, editor. Gatekeeper Series. IIED,
London.
Sauma, P. 2007. Pobreza, Desigualdad del Ingreso y Empleo: situación actual y
algunos escenarios prospectivos. CONARE y La Defensoría de los Habitantes,
San José, Costa Rica.
Simon, H. A. 1955. A behavioral model of rational choice. Quarterly Journal of
Economics 69:99-118.
Slovic, P., and S. Lichtenstein 2006. The Construction of Preference. Cambridge
University Press, Cambridge, UK.
Tetlock, P. E., O. V. Kristel, S. B. Elson, M. C. Green, and J. S. Lerner. 2000. The
psychology of the unthinkable: Taboo tradeoffs, forbidden base rates, and

41

heretical counterfactuals. Journal of Personality and Social Psychology 78:853870.
Tversky, A., and D. Kahneman. 1974. Judgment under uncertainty: Heuristics and
biases. Science 185:1124-1131.
Tversky, A., and D. Kahneman. 1981. The framing of decisions and the psychology of
choice. Science 211:453-458.
Walters, C., and C. S. Holling. 1990. Large-scale management experiments and
learning by doing. Ecology 71:2060-2068.
Walters, C. J. 1986. Adaptive Management of Renewable Resources. The Blackburn
Press, Caldwell, NJ.
Whittington, D. 1998. Administering contingent valuation surveys in developing
countries. World Development 26:21-30.
Wilson, R. S., and J. L. Arvai. 2006. When less is more: How affect influences
preferences when comparing low and high-risk options. Journal of Risk Research
9:165-178.
Wynne, B. 1992. Misunderstood misunderstandings. Public Understanding of Science
2:112-133.

42

CHAPTER 2
A CHOICE EXPERIMENT TO ADDRESS ENVIRONMENTAL, HEALTH,
AND ECONOMIC DEVELOPMENT CONFLICTS ABOUT PINEAPPLE
PRODUCTION IN COSTA RICA

1. Introduction
Costa Rica is a middle-income developing country located in Central America bordered
by Nicaragua to the north and Panama to the south. Despite its small land area
2

(approx. 51,100 km ), Costa Rica is one of the 20 most biologically diverse countries in
the world and forms an integral part of the Mesoamerican Biodiversity Hotspot. For this
reason, the country is also regarded for its dedication to natural resource conservation,
which has resulted in the establishment of numerous environmental laws and 160
continental protected areas (comprising 26% of Costa Rica’s total land area).
But as a developing country, Costa Rica is also focused on economic growth. In
addition to electric circuit manufacturing and tourism, which are its top two revenue
generators, Costa Rica also relies heavily on agricultural exports to support economic
development. Fertile land, regular rainfall, easy overland access to North and South
American markets, and direct maritime access to the European and Asian continents
have all greatly benefited the Costa Rican agricultural sector. While well known for
coffee and bananas, Costa Rica’s largest agricultural export is pineapple and its output
continues to increase. As of September 2010, pineapple exports—which saw a 18.5
percent increase in tons exported over 2009—account for more than $1 billion in sales
worldwide, with nearly $230 million worth of exports to the United States alone
(PROCOMER 2010).

43

Unlike coffee and—increasingly—banana production, which are predominantly in
the hands of domestic producers, pineapple production in Costa Rica is dominated by
foreign-owned multinational corporations such as Del Monte and Dole. These
corporations have invested heavily in pineapple production in Costa Rica; this includes
the maintenance of hundreds of large plantations, packing facilities, storage
warehouses, and transportation networks, as well as supporting infrastructure. In some
cases, these investments have resulted in ancillary benefits for Costa Rican
communities. Beyond providing a source of employment and income for tens of
thousands, some pineapple producers also build and maintain infrastructure that
members of communities use. Examples of the services provided by the pineapple
producers include the provision of transportation infrastructure (road, bridges, and
associated maintenance). Some of the larger pineapple producers also make packing,
storage, and export infrastructure available to smaller scale operations, which allows
them to get their products to market.
While clearly a boon to economic development in Costa Rica, the dramatic
intensification of large-scale pineapple production over the last 10 years has also come
at a significant price. In particular, to meet export demands and comply with
phytosanitary requirements, the once low impact crop now requires significant amounts
of pesticide and herbicide, which has resulted in a buildup of agrochemicals in
groundwater near some large pineapple plantations. Not surprisingly, many residents of
communities adjacent to plantations are concerned about the impacts of agrochemicals
on their air and water. At the same time, a lack of adherence to appropriate soil
conservation techniques by some producers has resulted in severe erosion problems.

44

There is concern among agronomists that if unchecked, the degree of soil degradation
could cause a situation so critical that affected lands cannot be cultivated for decades.
Additionally, improper post-harvest management of the crown and leaves by some
producers has created a pest problem—predominantly in the form of Stomoxys
Calcitrans—an aggressive biting fly that, if not controlled, can traumatize cattle and
wreak havoc on beef and milk production (SEPSA 2010).
Given the competing costs and benefits of pineapple production, many Costa
Ricans are beginning to call for more careful regulation of the industry. Policies being
discussed range from more stringent regulation and monitoring of the existing pineapple
industry to significant limits on the scale of production that would be allowed in the
country. Some Costa Ricans have even begun to call for a ban on pineapple production
in certain, sensitive areas. Given the range of options available, and the likelihood that
decision makers will have to confront challenging tradeoffs, the objective of our
research was to help inform policy decisions by eliciting preferences for competing
management plans from people living near pineapple plantations.
To meet this objective, we used a stated preference survey—namely a choice
experiment (CE)—where respondents are asked to make choices among pairs of
pineapple management scenarios. Choice experiments have been used widely in
developed country contexts. Traditionally, application of the CE method has focused on
eliciting preferences for consumer goods (e.g., Green & Srinivasan 1978; Schupp et al.
2003). But, more recently, CEs have been applied to explore preferences and estimate
non-market values in a variety of environmental management contexts (e.g., Boxall et
al. 1996; Farber & Griner 2000). However, the use of CEs in developing country

45

contexts is only slowly gaining traction (for a detailed review, see Bennett & Birol 2010).
For this reason, a secondary objective of the research reported here was to explore the
effectiveness of the CE method—mainly in terms of participant comprehension of, and
satisfaction with, the approach—in a developing country context.
Briefly, CEs are used to explore the role of both financial and non-market
variables in people’s preferences. They provide the added perspective of welfare
economics, which places both environmental and social considerations alongside more
commonly considered financial values. CEs are based on the argument that any good,
service, policy, or program is best described and evaluated in terms of its component
attributes and the levels that these attributes take (Lancaster 1966). For this reason,
CEs rely upon survey instruments to present respondents with a series of alternatives,
characterized as multidimensional composites or scenarios, that present integrated
(conjoined) combinations of different attributes (e.g., the effects of alternative policies on
contamination levels, employment, costs to consumers, etc.). These combinations
generally reflect actual or projected variations in the attributes (e.g., different levels of
water quality and local employment opportunities). In the more sophisticated CEs, the
represented attributes are specified by an experimental design that estimates the
separate and interacting effects of component attributes (Louviere 1988). Multiple
regression or similar analyses of decision makers’ choices among the presented
scenarios are used to estimate the relative contributions of individual attributes to the
expressed preferences. In other words, the choices made by people in response to
survey questions are used to reveal both preference orders across alternatives and the

46

tradeoffs across attributes that they are willing to make when comparing these
alternatives.
2. Methods
2.1 Choice Experiment Design
The survey instrument used in this research was developed after a series of in-depth
discussions between the authors, pineapple producers, technical experts, and
community stakeholders. Through these discussions, the researchers identified
appropriate attributes, associated measures, and attribute levels that characterize the
range of regulatory scenarios that may be available to policy makers. The resulting
instrument, which was prepared in Spanish, was also extensively pretested in the field
with individuals randomly selected from communities in the cantons where the CE was
carried out (see Section 2.2, below) to ensure that the attributes and attribute levels
made sense to respondents.
The final CE design consisted of six attributes that varied on between two and
five attribute levels (Table 1). The attributes that, during the design phase, were found
to be most relevant to survey respondents in the Atlantic Region included the following:
(1) The maximum plantation size that would be allowable. This value ranged from
small (less than 50 ha) operations used by independent national producers, some
of whom are organic producers, to the current large (greater than 250 ha)
monocultures owned by nationals and foreign multinationals;
(2) Whether or not external buffer zones, comprised of trees and tall scrub, would be
required between plantations and neighboring communities. Buffer zones serve
two primary purposes, namely blocking unsightly views of the plantations from

47

neighboring communities and major roadways, and—importantly—creating a
physical barrier between communities and plantations that make it difficult for
aerosolized agrochemicals to pass;
(3) Whether or not appropriate soil conservation techniques (which would depend on
factors such as the farm’s soil characteristics and slope), to prevent erosion and to
ensure the long-term viability of agricultural lands currently used for pineapple,
would be required;
(4) The annual frequency with which highly leachable pesticides and herbicides, such
as bromacil (i.e., agrochemicals that may pass through the soil into groundwater),
may be applied on plantations. This value ranged from zero to four applications
per year;
(5) The annual frequency, with which random compliance checks by governmental
regulatory agencies would be conducted on pineapple plantations, regardless of
their size. Monitors would help to ensure that agrochemicals were being applied at
the proper frequency, that appropriate soil conservation techniques were being
used, and that appropriate buffer zones were implemented. This value ranged
from zero to three or more times per year; and
(6) The additional cost, in Costa Rican colones (₡), that would be borne
(hypothetically) by households that would be willing to pay a fee in exchange for
improved compliance monitoring of pineapple production practices (i.e., regarding
buffer zones, soil conservation and agrochemicals), and/or to help pay
infrastructure costs if pineapple plantations were to shrink in size. Essentially, the

48

cost attribute allows us to test the degree to which the other five attributes are
viewed by households as amenities.
The rationale behind the increase in cost stems from the fact that large
pineapple producers are financially able to help provide infrastructure used by the
communities. If the scale of production was to shrink, producers’ revenues would
decrease as would their ability (and motivation) to provide ancillary community
services. As a result, communities would be required to contribute to covering the
costs of these services, which they do not wish to lose, through an increase in the
local fee structure. At the same time, an increased desire at the community level
for compliance monitoring would result in an increase in agency costs, which also
would be partially passed on to citizens. However, not all households in the areas
where we conduced this research pay property taxes, so we had to devise a
method for them to contribute to the costs associated with increased environmental
monitoring and smaller pineapple plantations. Since all households pay fees to
Grupo ICE, the national electricity and telecommunications (television, as well as
wired and wireless phones and internet) provider; bills from Grupo ICE were
selected as the payment vehicle. After consulting with stakeholders and
regulators, we established a means of including a “community service contribution”
as a monthly addition to households’ Grupo ICE bills. This monthly fee ranged
from between ₡0 (the status quo) and ₡2,452.
An orthogonal, main effects experimental design comprised of these attributes
and associated levels was constructed using the PASW (formerly SPSS) Conjoint
module. A total of 25 unique alternatives were created, in addition to a “status quo”

49

alternative (which was labeled as such). Respondents were asked to compare a
randomly selected alternative with the status quo alternative and then select their
preferred option (see Table 2 for an example choice set); thus, a total of 25 choice sets
(i.e., profiles) were possible. Each respondent received a total of five randomly selected
choice sets, which were presented to them one at a time, on laminated cards.
Accompanying the CE was a short survey that was administered orally to all
respondents. The purpose of the survey was to collect baseline demographic
information about respondents, including information about whether they—or a member
of their family—work on a pineapple plantation, whether they are a farmer, and whether
they are a member of any local associations or committees. However, because CEs
are used relatively infrequently in developing country contexts, we also used the survey
to collect data about respondents’ level of satisfaction with the method. All responses
were recorded using 5-point Likert scales with neutral midpoints. The questions we
posed asked for respondents’ (1) level of satisfaction with their responses across the
five choice sets (1=very unsatisfied, 5=very satisfied); (2) the level of ease with which
the CE could be completed (1=very difficult, 5=very easy); and (3) the degree to which
respondents’ choices reflected their concerns associated with pineapple production
(1=very poorly, 5=very well). Using similar scales, we also asked for respondents’ selfreported knowledge level about pineapple production and their views on the importance
of eliciting preferences about management options from the community members.
The CE and accompanying survey were administered orally, and in Spanish.
Fieldwork was carried out by the first author and an experienced local research
assistant, together with university students recruited from the National University of

50

Costa Rica’s Sarapiquí Campus. At the end of each day, completed instruments were
checked by the first author and the research assistant to ensure that the survey
procedures were being followed correctly and without deviation over the course of the
study. The complete instrument was administered in a manner consonant with human
subjects requirements and took, on average, 30 minutes for respondents to complete.
2.2. Respondents
A bulk of Costa Rica’s pineapple production occurs in the Limón Province of the
country’s Atlantic Region. Recently, this area has been the site of disagreement, as
well as outright conflict, between pineapple producers, the government, and local
communities. In an attempt to begin to resolve some of the conflict, a Presidential
Commission was formed in 2008 and charged with developing a better understanding of
the concerns of both community stakeholders and pineapple producers, and with
making recommendations to the government about possible ways to address these
concerns. EARTH University was asked to serve as a neutral mediator during the
commission’s work and one of the authors (León) was named as the commission’s
facilitator. For this reason, our research focused on the Limón Province and our results
will be shared with members of the Presidential Commission, as well as with
government representatives, pineapple producers and interested members of local
communities. However, we stress that our research was neither requested, funded, nor
sanctioned by the Presidential Commission, nor any of its members.
Within the Limón Province, we selected three cantons (Guácimo, Siquirres and
Pococí) as our main sampling sectors because most of the pineapple production in the
Atlantic Region is concentrated there. A stratified random sample of 451 households

51

was drawn from these cantons. Rather than sampling on a village-by-village basis, we
worked with the National Institute of Statistics and Census (INEC) based in the capital,
San José. In addition to our desire for a legitimately drawn stratified random sample,
we wanted to mirror the sampling procedures utilized by the Government of Costa Rica
(when conducting the census and other surveys of national interest) so as to add
credibility to our research. We also wanted to avoid oversampling respondents, both
from certain high-conflict areas where strongly negative feelings about pineapple
production prevail, as well as from areas so removed from any production—much less
conflict—that respondents might have no opinion at all about the issue.
During the draw of respondents, INEC took into account each canton’s total
population and its urban, semi-urban, and rural distribution in order to produce a
representative group of sampling segments for each canton. Each of these segments
was then rendered on cadastral maps. A random sample of 10 households was then
randomly selected for interviews from each segment. Researchers visited a total of 650
houses (569 of which were eligible), collected 472 surveys, and ended up with a total of
451 completed surveys for a response rate of 0.79.
Interviews were conducted only with respondents who were responsible for
making financial decisions, either jointly or alone, on behalf of the household. Seventythree percent of these respondents were female, and the average age of respondents
was 43. Fifty two percent of respondents were originally from the Limón Province, and
of that group, the average length of residency is 43 years. Respondents not originally
from Limón have lived in their community for an average of 24 years and in their canton
for an average of 32 years. The majority of participants, 71 percent, were married or in

52

“free union” with their partner (47 percent were married and 24 percent in “free union”).
The majority of respondents (66 percent) reported a monthly income of less than
₡218,433 (the poverty line), with 26 percent earning that income through farming. An
average of 41 percent of respondents reported that they themselves (7 percent) or a
family member (34 percent) currently work (or had previously) in some aspect of
pineapple production. On average, respondents had completed about six years of
education, and their spouses (including “free unions”) had also completed about six
years, while 23 percent of the respondents (and 21 percent of spouses) either did not
complete six years of elementary school or had no elementary education at all.
Eighteen percent of respondents self-identified as active members of a community,
canton, or provincial committee or organization such as a school committee, community
development association, or rural aqueduct association. (See Table 3 for summary
sample characteristics.)
2.3 Analysis
Statistical analysis of the CE data was carried out using a binary random effects panel
probit model in STATA/SE 10.0. Respondents’ choices of an alternative over the status
quo were regressed against the six choice attributes outlined in Section 2.1. Other
explanatory variables, including standard demographic data (i.e., age, gender,
education, income) and context specific information (e.g., whether the respondent, or
anyone in his/her family, has worked on a pineapple plantation), were also tested, as
well as numerous interaction terms. The complete model was interacted with the
canton of study given that population demographics and the geographical distribution
and intensity of pineapple production vary across the three cantons. Dummy coding

53

was used for categorical variables and continuous variables (i.e., cost) were not
recoded.
Numerous models were estimated, but we will mainly report on the results of: 1)
the choice model (choice variables only); 2) the complete model, which includes
demographic variables and significant interaction terms; and 3) the complete model for
each of the three cantons. There is a short description of choice model covariates in
Table 4, and the estimation results for these models are presented in Table 5 and Table
6.
3. Results
Parameter estimates of the variables that are statistically significant in the two full
sample models are consistently higher in the complete model with the exception of
Monitoring_3, which stayed the same (see Table 4). The following discussion focuses
on the results of the complete model regressed by canton (see Table 5).
As expected, Cost was highly significant at the 1 percent level with a negative
linear effect on respondents’ choice of an alternative to the status quo. It is worth noting
that this result was consistent in all models except when the complete model was
regressed for Siquirres. In this canton, Cost became less significant, with a slightly
lower coefficient. This indicates that the negative linear effect of the Cost variable may
not be quite as strong in Siquirres as it is in Guácimo and Pococí. Therefore,
respondents from Siquirres may be slightly less influenced against choosing an
alternative to the status quo as the Cost level increases. This is not surprising given the
fact that this canton has been the location of some of the most contentious conflicts
between communities and farms, mostly with respect to concerns over water

54

contamination from the application of agrochemicals. Siquirres is also the canton with
the highest percentage of its area dedicated to pineapple production. Taking into
account both of these factors, respondents in this canton might be more willing to make
a financial contribution in exchange for improved pineapple production practices,
especially (as we discuss below) improvements in agrochemical application.
Agrochemicals_2 was highly significant at the 1 percent level and its positive
coefficient indicates that the inclusion of this variable in an alternative to the status quo
had a significant positive impact on the respondents’ selection of an alternative. The
fact that a moderate level of agrochemical application had a greater impact than lower
levels (i.e., level 0 or level 1) on the selection of an alternative to the status quo may
speak to the fact that respondents were realistic—as opposed to idealistic—when
evaluating this variable. This is understandable given that residents in the study area
are familiar with production practices (not limited to pineapple production) and most
people depend in some way on farming for their livelihood. Therefore, while they
generally oppose high (i.e., status quo) application levels, which many see as
unnecessary and avoidable, they tend to believe that moderate application is more
realistic and necessary than none or very low levels.
When analyzed by canton, Agrochemicals_2 was highly significant at the 1
percent level in Siquirres, and at the 5 percent level in Guácimo, but not at all in Pococí.
It is important to point out that the other Agrochemicals attribute levels
(Agrochemicals_0 and Agrochemicals_1) were not at all significant in any cantons. This
makes sense given respondents’ knowledge of agricultural practices. Additionally, it is
not surprising that respondents in Siquirres were so strongly influenced by the presence

55

of Agrochemicals_2 in the alternatives they evaluated considering that the most
contentious conflicts over water contamination due to pesticides has taken place in that
canton. To our knowledge, at the time of this data collection there had not been a
conflict of this intensity in Pococí, which could at least partially explain why respondents
from this canton were not strongly influenced by this attribute when deciding whether or
not to select an alternative to the status quo.
Monitoring_3 was highly significant at the 1 percent level with a positive impact
on selecting an alternative over the status quo. While Monitoring_1 is marginally
significant, it is to be expected that Monitoring_3 would exert a stronger influence on the
respondents’ decisions to select an alternative than did Monitoring_1. When analyzed
by canton, however, Monitoring_3 is only significant (and at the 5 percent level) in
Siquirres. (It is barely significant at the 10 percent level in Guácimo.) Despite not being
extremely significant in these two cantons, its positive high coefficients reveal that it
does influence the selection of an alternative over the status quo. Interestingly, in
Pococí, it is not at all significant. It seems possible that this attribute can be a least
partially explained by the degree of environmental-production conflict experienced in
Siquirres and lacking in Pococí. Another explanation could be that some respondents
doubt that increased monitoring would actually be successful and were therefore not as
influenced by this variable when deciding whether to select an alternative to the status
quo.
Soil Conservation was significant at the 5 percent level with a positive coefficient
in the complete model, so it appears that the requirement of soil conservation
techniques in alternatives to the status quo does positively influence the respondents’

56

decision to opt for the alternative. When analyzed by canton, however, we see that Soil
Conservation is only significant in Guácimo. That said, it is important to note that where
it is significant, it exerts a fair amount of influence. In Guácimo, Soil Conservation has a
slightly lower positive coefficient than does Agrochemicals_2 and Farm Size_150, and a
slightly higher one than Monitoring_3. It is worth noting that respondents in this canton
appear to have been influenced to an almost equal degree by all six of the choice
attributes instead of anchoring on just two or three of them, as appears to be the case of
respondents in Siquirres and Pococí (who in addition to Cost, anchored strongly on
Agroghemicals_2 and Monitoring; and Farm Size and Buffer Zone, respectively).
It is important to point out that four variables that were not significant in the
Choice Model do become significant in the Complete Model. These variables are: Farm
Size_Small, Farm Size_Medium, Monitoring_1, and Buffer Zone. That said, only Farm
Size_Medium is significant at the 1 percent level, while the other three are marginally
significant at the 10 percent level. All have positive coefficients except for Buffer Zone.
Given that the interaction term, Farm Size x Buffer Zone, is significant (at the 1 percent
level) and has a positive coefficient, we can interpret that Farm Size_Medium (and to a
small degree, on Farm Size_Small) has a strengthening effect on Buffer Zone regarding
the impact that this variable has on respondents’ selection of an alternative to the status
quo.
Among the numerous demographic variables that were tested to further improve
the model fit, significant effects emerged only when the following variables were
included: Piña Worker; Age; and Low Income. In terms of impact on selecting an
alternative to the status quo, Piña Worker has a positive impact, while both Low Income

57

and Age have negative coefficients. In the complete model all are significant at the 5
percent level, however, when run by canton, only Piña Worker was significant, and only
in Pococí. No significant effect was found (and therefore also not presented in Table 5)
for the other demographic variables, quadratic effects and interactions that were tested.
Finally, it is worth noting that when the Complete Model is analyzed by canton,
some important changes in variable significance levels and coefficients occur. While
five variables were no longer significant when the model is regressed for Pococí alone,
the variables that remain significant are more so, and their coefficients are significantly
higher. Especially notable is Farm Size_Small, which becomes significant at the 1
percent level (previously at the 10 percent level) with a coefficient of .944. Farm
Size_Medium retains at the same significance level, but its coefficient also increases.
Buffer Zone moves from a significance level of 10 percent to 5 percent, with an
important increase in its coefficient. Regarding the demographic variables, Piña Worker
stays at the same significance level (5 percent), but its impact is stronger. When the
model is regressed in Guácimo, six of the variables are no longer significant, and while
the significance levels of the variables that remain actually decrease, their coefficients
increase. This is true of Farm Size_Medium, Agrochemicals_2 and Monitoring_3. Soil
Conservation and Cost retained the same significance levels, and their coefficients also
increased. In Siquirres, nine of the significant variables became non significant. Of the
four that remained significant, Agrochemicals_2 and Monitoring_3 stayed at the same
level, Monitoring_1 increased from the 10 percent to the 5 percent level, and all of their
coefficients increased. However, only in Siquirres did Cost decrease in significance
from the 1 percent level to the 5 percent level (no change in the coefficient). To

58

summarize, respondents in the three cantons are impacted differently by the choice
attributes. In Guácimo, significant attributes are ranked as follows: 1) Cost; 2)
Agrochemicals_2; 3) Farm Size_150; 4) Soil Conservation; and 5) Monitoring_3. In
Siquirres, the order is: 1) Cost; 2) Agrochemicals_2; 3) Monitoring_3; and 4)
Monitoring_1. Lastly, in Pococí: 1) Cost; 2) Farm Size_25 and Farm Size_150; and 3)
Buffer Zone.
Regarding respondents’ opinion of the elicitation method itself, our data show
that, on average, respondents were: satisfied with their responses across the five
choice sets ( x = 3.6, sd = 0.8); neutral about the level of ease with which the CE could
be completed ( x = 3.4, sd = 0.9); and felt that their choices did a good job reflecting
their concerns associated with pineapple production ( x = 3.7, sd = 0.9). Respondents
self-report a moderate level of knowledge about pineapple production ( x = 2.6, sd =
0.9) and feel it is important that community members be consulted about their
preferences regarding pineapple production practices ( x = 4.3, sd = 0.9).
4. Discussion
This paper presented the results of a choice experiment conducted in the Atlantic
Region of Costa Rica where decision makers and other stakeholders are faced with the
need to confront challenging tradeoffs regarding pineapple production and
environmental management. The primary objective of our research was to help inform
policy decisions by eliciting preferences for competing management plans from people
living near pineapple plantations. Specifically, we wanted to learn about: 1) whether
residents, when given a choice, would prefer current production practices or an
alternative to the status quo; 2) if residents prefer alternatives to the status quo, what

!

59

!

are the attributes and levels that most strongly influence their selection of an
alternative?; and finally, 3) what tradeoffs—if any—would residents be willing to accept
across various aspects of production?
Regarding the first question, this choice experiment indicates that residents
would prefer an alternative to the status quo. Overall, the status quo option was the
least popular of the 26 production scenarios, being selected only 28 percent of the time.
When respondents had to choose between the status quo option and the most popular
alternative, the status quo was selected just 10 percent of the time. Even when it was
compared to the least popular option, it was selected only 52 percent of the time.
However, based on our sample, most community members do not want to halt
pineapple production entirely, as has been reported in the media. Rather, it appears
that many residents of the major pineapple producing cantons of the Atlantic Region
see pineapple production as both beneficial and problematic and would prefer the
industry to use more environmentally sound production practices while continuing to be
competitive in the market, and therefore, continuing to provide jobs and contribute to the
agricultural economy of the region. This may sound like an unrealistic goal, and it would
be if people were not willing to make tradeoffs, but our data show that this is not the
case. Second, despite important and policy-relevant variations in the preferences of
residents from the three cantons, our data show that the attributes (and levels) that
most influenced respondents’ selection of an alternative over the status quo were: Cost
(the lower the better); Agrochemicals (2); Monitoring (3), and Soil Conservation (Yes).
Based on the variables making up our complete model, we can say that, on average,

60

households would be willing to pay approximately ₡190 if the following alternative
production scenario were to be implemented:
Finally, our data show that respondents were willing to make tradeoffs with
respect to the attributes they appear to care most about in order to see an overall
improvement in pineapple production practices. The clearest indicator of this is the fact
that alternatives to the status quo were selected far more often (72 percent of the time)
than the status quo itself even though a cost of at least 105 colones was included as
part of the alternative (versus no cost in the status quo). Another important example is
that most respondents appear to have not been as influenced by the inclusion of a
buffer zone unless an alternative also included a large farm. If the alternative did not
include a large farm, respondents were far more influenced by the level of agrochemical
application, the frequency of monitoring and the requirement of soil conservation
techniques. However, if a large farm was part of the alternative, having a buffer zone
became more important than some of these otherwise most influential attributes and
levels. Lastly, even in the canton most impacted by agrochemical contamination,
respondents were not extremely influenced by the “best” level of agrochemical
application (i.e., none). To the contrary, most of these respondents—as well as those in
the other two cantons—appear to be willing to accept a moderate level of application
(i.e., 2 times/year), presumably in exchange for paying a lower cost or seeing a gain in
another area, such as monitoring or soil conservation.
4.1 Policy Implications
This research was conceived with the joint goal of testing tradeoff analysis methods as
well as contributing useful information about community members’ preferences to the

61

current policy debate about pineapple production in the Atlantic Region. To this end,
there are three main points that can be made about how these findings may be useful to
the ongoing policy discussion. All three of them contradict what has typically been
reported by the national and international press, which tend to portray the situation in
terms of stakeholders representing extreme positions and entrenched ideals with no
willingness to collaborate.
First, it appears that many community members in the study area are not
opposed to pineapple production. They simply feel that it can, and should, be done in a
way that better protects environmental and human health. Even among those who are
staunchly opposed to large-scale monoculture in general, or to conventional pineapple
production specifically, many recognize that the plantations provide much needed jobs,
although they point out that labor conditions need to be improved (e.g., training for field
workers on safe agrochemical handling and application; provision of proper safety
equipment to field workers). Similarly, many community members do not think that
most pineapple producers are in extreme incompliance with environmental regulations,
but they tend to believe that the industry as a whole has not been held as transparent
and accountable as it should be, and therefore has been slow to respond to
environmental and health concerns.
Second, many community members are willing to make tradeoffs about how to
address their concerns about pineapple production. A misperception exists that people
who say anything negative about pineapple production, or even simply question a
farm’s practices, are environmental or labor activists who want to shut a farm—if not the
entire industry—down. Our data reveal that most residents in the study area do have

62

significant concerns about the industry’s environmental and human health impact, but
that when given the opportunity to consider pineapple production in a more nuanced
way than just whether it is “good” or “bad”, they are willing to make tradeoffs. That is,
they are willing to identify the aspects of production that concern them the most and
consider the possibility of not having all of their concerns addressed immediately in
exchange for seeing some real progress made with respect to one or two aspects at a
time.
Finally, our research suggests that there is reason for the stakeholders involved
in this conflict—who have tried formally on two occasions to address concerns about the
region’s pineapple production by working together on a multi-stakeholder commission—
to be optimistic about their ability to make meaningful, yet incremental, progress toward
addressing some of the most critical aspects of this extremely complex issue. While
they might not be able to address the most contentious issues first, our research
suggests that community members may be open to making improvements in a less
contentious area (i.e., buffer zones) first. Ideally, that experience would help to build the
trust and dialogue needed to take on the more complex aspects of pineapple production
(i.e., agrochemicals).
Any stakeholder representatives who are truly willing to address concerns about
pineapple production from a holistic perspective (i.e., production, environmental
management, human health, worker safety and labor conditions)—and other phases of
our research confirm that this willingness does exist—should feel confident that many
community members from throughout the region are both knowledgeable (despite being
overly modest in self-assessments of their knowledge) enough about agricultural

63

production and natural resources and, more importantly, willing to work with producers
and government representatives to discuss and address concerns (of all stakeholders)
in a strategic way. Essentially, instead of insisting that all of their concerns be resolved
immediately and simultaneously, our research suggests that many community members
would be willing to support efforts to make incremental improvements as long as they
feel that their concerns and suggestions are being taken seriously. As stakeholders at
the regional level have suggested, this kind of approach will probably be more effective
if carried out at the regional level. Our research supports this idea since it reveals that
while community members in the three different cantons are most likely concerned
about the same things in general, they rank them differently in terms of importance
based on the context specific realities of each canton.
In the interest of future work of this kind, it is worth reporting that implementing
this choice experiment involved a few challenges. First, the fact that stakeholder groups
in the study area had recently been—or were currently—engaged in varying degrees of
conflict regarding pineapple production (with respect to concerns about production
practices, environmental management, human health and labor conditions) could have
introduced a bias in choice behavior favoring alternatives to the status quo. Similarly,
over sampling in communities that do not feel impacted in any way by pineapple
production could have biased respondents toward choosing the status quo. With this in
mind, researchers worked with Costa Rica’s National Statistics Institute (INEC) to
develop a truly random sampling procedure that would be viewed as legitimate by
government officials, pineapple producers, community groups and other stakeholders,
and take into account the demographic characteristics and urban-rural population

64

distributions of the three cantons, in addition to helping researchers avoid over sampling
in high- or no-conflict areas.
Second, there is a feeling among some rural residents in Costa Rica that
governmental institutions do not provide reliable services in their communities. Since
the payment vehicle for the Cost attribute is linked to a public institution (Grupo ICE),
this feeling could have biased some people against being willing to contribute to help
fund infrastructure projects, improved production practices, and environmental
monitoring, despite the fact that we explained that ICE would not have anything to do
with administering the funds and was simply an example of a potential form of collecting
their contribution in a convenient manner. Along these lines, some people became so
upset by the idea of paying anything when they view pineapple producers as wealthy
companies and government officials as either unable (i.e., due to resource limitations
and/or a central government bias toward supporting pineapple production at any cost)
or not willing to do their job, that it was sometimes difficult for them to stop anchoring on
the cost attribute and thoroughly evaluate the others. In short, trying to help
respondents think in terms of a hypothetical, yet realistic, situation was sometimes
challenging.
The third challenge was logistical. Due to the fact that the research area was
large (sq. kilometers) and the segments chosen by INEC in order to ensure an unbiased
random sample were at a very small scale, a great deal of work was required by the
research team in order to simply orient the segments on larger scale maps of the
cantons and then locate the areas in the field. This was further complicated by the fact
that a number of the segments covered areas that were considered too dangerous for

65

the research team to conduct interviews due to increased violence (i.e., armed robbery,
carjackings, other gang-related activity) in these areas. While these activities are very
localized and easily avoidable with accurate information and a good knowledge of the
local geography and road network, the safety of the interviewers was the researcher’s
main concern during data collection. Therefore, ensuring interviewer safety while also
being respectful of participants and their time, and meeting data collection goals
required considerably more logistical research, planning, and on-the-fly adjustments
than is normally required when doing research in Costa Rica, which is, overall, an
extremely peaceful and safe country.

66

APPENDICES

67

APPENDICES

Table 1. Attributes and levels used in the CE design (translated from Spanish).
Attribute
Maximum plantation size
Requirement that external
buffer zones be established
Requirement that soil
conservation techniques be
used
Allowed number of
leachable agrochemical
applications/year
Number of random
compliance checks/year
Monthly household fee on
Grupo ICE bill

Attribute Levels
Small
<50 ha

Medium
50-250 ha

Large
>250 ha

No

Yes

No

Yes

0

1

2

0

1

₡0

₡105

₡463

4
3 or more

₡1,837 ₡2,452

Table 2. Example choice set (translated from Spanish).
Alternative A
(status quo)
Large
>250 ha

Attribute
Maximum plantation size
Requirement that external buffer
zones be established
Allowed number of leachable
agrochemical applications/year
Requirement that soil
conservation techniques be
used
Number of random compliance
checks/year
Monthly household fee on Grupo
ICE bill

68

Alternative B
Medium
50-250 ha

No

Yes

4

1

No

Yes

0

3

₡0

₡1,837

Table 3. Summary sample characteristics.
Full
Sample

Guácimo

Siquirres

Pococí

Percent women (%)

73

70

77

73

Average age

43

42

44

42

26

27

28

23

71

68

73

71

Less than
₡218,433

Less than
₡218,433

Less than
₡218,433

Less than
₡218,433

26

36

21

22

41

50

41

35

6

6

6

6

6

6

6

6

18

18

22

15

Average years of
residency in canton
Percent married or
in “free union” (%)
Average income
(colones/household
/month)
Percent farmer (%)
Percent
working/having
worked in pineapple
production (%)
Respondent’s
average education
level (years)
Spouse’s average
education level
(years)
Percent involved in
community
organization (%)

69

Table 4. Short description of choice model covariates.
Variable

Description

Nature

Min

Max

Dummy

0

1

Dummy

0

1

Omitted
Dummy

0

1

Dummy

0

1

Dummy

0

1

Dummy

0

1

Omitted
Dummy

0

1

Omitted
Dummy

0

1

Dummy

0

1

Dummy

0

1

Dummy

0

1

Dummy

0

1

Continuous

₡0

₡2,4
52

Dummy

0

1

18

98

0

98

Attribute main
effects
Farm Size_Small
Farm Size_Medium
Farm Size_Medium
Agrochemicals_0
Agrochemicals_1
Agrochemicals_2
Agrochemicals_4
Monitoring_0
Monitoring_1

Monitoring_3+

Soil Conservation
Buffer Zone
Cost

Farm Size of
<50 ha
Farm Size of
50-250 ha
Farm Size of
>250 ha
Allowed number of
applications/ year = 0
Allowed number of
applications/ year = 1
Allowed number of
applications/ year = 2
Allowed number of
applications/
year = 4
Number of random
compliance
checks/year = 0
Number of random
compliance
checks/year = 1
Number of random
compliance
checks/year = 3 or
more
Soil conservation
techniques are
required
External buffer zones
are required
Monthly household
contribution on ICE bill
(Colones)

Other covariates
Gender
Age
Years in Community

Male = 1
Respondent’s age
Years respondent has
lived in the community
70

Continuous
Continuous

Table 4 (cont’d)
Years in Canton
Education_
Elementary
Education_High
School

Education_College+

Farmer
Plantation Worker

Low Income

Married/Free Union

Community Group

Years respondent has
lived in the canton
Respondent has
completed 0-6 years of
school
Respondent has
completed 7-11 years
of school
Respondent has
started or completed
licentiate degree
and/or graduate
studies
Respondent is a full- or
part-time farmer
Respondent or a family
member has worked or
is working for a
pineapple farm
Respondent’s monthly
household income falls
below the poverty line
(₡218,433)
Respondent’s civil
status is: married or
free union
Respondent is a
member of a local or
regional association or
organization

Continuous

0

98

Omitted
Dummy

0

1

Dummy

0

1

Dummy

0

1

Dummy

0

1

Dummy

0

1

Dummy

0

1

Dummy

0

1

Dummy

0

1

Continuous

0

300

Continuous

0

6

Interaction effects
Farm Size x Buffer
Zone
Agrochemicals x
Monitoring x Soil
Conservation

Interaction between
Farm Size and Buffer
Zone
Interaction between
Agrochemicals,
Monitoring, and Soil
Conservation

71

Table 5. Estimation results (Choice and complete models).
Choice Model

Complete Model

Coefficient

St.
error

Coefficient

St.
error

Farm Size_Small

-.026

.106

.364

.216*

Farm Size_
Medium

.039

.086

.367

.136***

Agrochemicals_0

.022

.101

.043

.141

Agrochemicals_1

.062

.118

.217

.150

Agrochemicals_2

.224

.117*

.511

.148***

Monitoring_1

.138

.103

.201

.121*

Monitoring_3+

.259

.105**

.332

.138**

Soil Conservation

.202

.077***

.256

.114**

Buffer Zone

.093

.077

-.390

.227*

Cost

-.000

.000***

-.000

.000***

Gender

.228

.188

Age

-.012

.006**

Years in
Community

-.004

.007

Years in Canton

.005

.007

.078

.201

.312

.290

Farmer

.152

.188

Plantation
Worker

.390

.164**

Variable
Attribute main
effects

Other covariates

Education_
High School
Education_
College+

72

Table 5 (cont’d)
Low Income

-.356

.181**

Community
Group

-.168

.214

Interaction effects
Farm Size x
.003
.001**
Buffer Zone
Agrochemicals x
Monitoring x Soil
-.065
.038*
Conservation
Note: *** 1% significance, ** 5 % significance, and * 10% significance

73

Table 6. Complete model by canton.
Guácimo

Siquirres

Pococí

Coefficient

St.
Error

Coefficient

St.
error

Coefficient

St.
error

.394

.401

-.266

.409

.944

.358***

.620

.279**

-.027

.246

.585

.221***

.112

.278

.186

.276

-.033

.216

.182

.314

.197

.276

.329

.234

.655

.313**

.824

.268***

.189

.234

Monitoring_1

.331

.252

.486

.231**

-.052

.184

Monitoring_3+

.484

.289*

.687

.269**

-.024

.208

Soil
Conservation

.564

.238**

.273

.211

.073

.178

Buffer Zone

-.478

.424

.110

.429

-.809

.371**

Cost

-.001

.000***

-.000

.000**

-.000

.000***

Age

-.018

.013

-.009

.011

-.013

.010

Plantation
Worker

.384

.340

.066

.300

.676

.275**

Low Income

-.139

.403

-.480

.325

-.401

.298

Variable
Attribute main
effects
Farm
Size_Small
Farm
Size_Medium
Agrochemicals_0
Agrochemicals_1
Agrochemicals_2

Other
covariates

Interaction
effects
Farm Size x
.004
.002*
.001
.002
.004
Buffer Zone
Agrochemicals
x Monitoring x
-.181
.078**
-.088
.071
.033
Soil
Conservation
Note: *** 1% significance, ** 5 % significance, and * 10% significance

74

.002**
.061

Table 7. Alternative pineapple production scenario.

Large
>250 ha
No

No

4

2

No

Yes

0

3

₡0

Attribute

Preferred
Attribute
Levels
Medium
50-250 ha

₡190

Status Quo

Maximum plantation size
Requirement that external buffer
zones be established
Allowed number of leachable
agrochemical applications/year
Requirement that soil conservation techniques be used
Number of random compliance
checks per year
Monthly household contribution
on Grupo ICE bill

75

REFERENCES

76

REFERENCES

Bennett, J., and E. Birol, editors. 2010. Choice Experiments in Developing Countries:
Implementation, Challenges, and Policy Implications, Cheltenham, UK.
Boxall, P., W. Adamowicz, M. Williams, J. Swait, and J. Louviere. 1996. Comparison of
stated preference approaches to the measurement of environmental values.
Ecological Economics 18:243-253.
Farber, S., and B. Griner. 2000. Using conjoint analysis to value ecosystem change.
Environmental Science & Technology 34:1407-1412.
Green, P. E., and V. Srinivasan. 1978. Conjoint analysis in consumer research: Issues
and outlook Journal of Consumer Research 5:102-123.
Lancaster, K. 1966. A new approach to consumer theory. Journal of POlitical
Economics 74:217-231.
Louviere, J. J. 1988. Analysing decision making: Metric conjoint analysis. Sage
University Papers Series in Quantitative Applications in the Social Sciences.
Sage Publications, Newbury Park, CA.
Schupp, A., J. Gillespie, W. Prinyawiwatkul, and C. E. O'Neil. 2003. Consumerpreferred attributes of a fresh ground beef and turkey product: A conjoint
analysis. Journal of Food Distribution Research 34:47-52.

77

CHAPTER 3
STRUCTURING DECISION MAKING IN DEVELOPING COMMUNITIES: INFORMING
LAND USE POLICY IN COSTA RICA

1. Introduction
For more than half a century, research and practice in international development has
focused on improving quality of life for people living in developing regions of the world.
Much of this work encompasses projects aimed at poverty reduction, encouraging
democratic governance, facilitating private sector development, enhancing human
health, providing sustainable infrastructure for basic education, and encouraging social
justice and equality. The federal governments of almost every industrialized nation
sponsor agencies and programs aimed at meeting these goals; the United States
Agency for International Development, Canada’s International Development Agency,
the Swedish International Development Cooperation Agency, and the UK Department
for International Development are high-profile examples. Several other countries and
organizations participate in international development activities through representation
in the United Nations (UN) or with donor agencies such as the World Bank.
Recently, these organizations have recognized the need to incorporate
objectives associated with community participation and environmental sustainability—
together with other, more traditional development objectives such as economic
development, education, and human health—in decisions about international and
community development. From a development perspective, the expansion of the
international development mission to include environmental objectives makes sense: in
all regions of the world—developed and developing—human health and quality of life

78

are directly linked to the health of the environment and the sustainable use of natural
resources. And given that one of the goals of international development initiatives is to
empower local communities through democratic processes, it also makes sense to
involve community stakeholders in these decisions.
Successfully eliciting community input about environmental policies is not easy.
Decades of research in the decision sciences have shown that getting input from people
about these kinds of multiattribute decisions is extraordinarily difficult. Research dating
back to the seminal studies of bounded rationality (Simon 1955) points out that people
do not readily evaluate all of the available alternatives in a given decision in terms of the
pros and cons associated with all of their associated attributes (e.g., economic impacts,
human and environmental health, etc.). Instead, people focus only on a very small
subset of the attributes—usually those that are the most salient or easiest to evaluate—
and ignore others; this lexicographic process implies that choices made typically do not
accurately reflect their full range of values and concerns. Other, related problems also
persist. For example, people find it difficult to identify and characterize the objectives
that matter to them and which, logically, should guide choices among the alternatives
that are under consideration (Bond et al. 2008). And in the cases where people do
attempt to balance multiple objectives when choosing among alternatives, they have a
hard time balancing the pros and cons of competing alternatives when these objectives
conflict (Keeney & Raiffa 1993; Tetlock 2000).
Complicating matters further are findings relating to the widespread use of a
panoply of simplifying heuristics that reduce the amount of time and effort required by
people to make a choice but introduce systematic biases. The seminal work in this area

79

has identified a series of judgmental heuristics that, when applied by decision makers,
violate basic tenets of rationality; the availability and representativeness heuristics along
with the faming effects associated with prospect theory (Kahneman & Tversky 1979;
Tversky & Kahneman 1973; Tversky & Kahneman 1981) are commonly cited examples.
Research in this area has continued, resulting in a profusion of scholarly articles (e.g.,
see Hsee 1996; Rivers & Arvai 2007; Slovic et al. 2002) and popular books (e.g., see
Ariely 2008; Fishbein & Cappella 2006; Fishbein & Yzer 2003) that point out common
flaws—and their consequences—in human judgment and decision making.
But while studies characterizing the cognitive processes that lead to poor
decisions continue to proliferate, research focused on how to improve the quality of
input from people lags far behind. The bulk of this work in the area of international and
community development has been focused on two needs: improving decisions by
providing better information to decision makers (Ajzen 2001; Millennium Ecosystem
Assessment 2005), and creating more opportunities for stakeholders to become
involved in decision-making processes (Chambers 1994; Ostrom 1996; Rhoades 1998).
It’s difficult to argue against either of these positions. Indeed, providing a solid
foundation of supporting information to inform the judgments of stakeholders and
decision makers should be preferred to the provision of an inferior one. Likewise, policy
makers and analysts should make it possible for stakeholders to participate in
consequential decision-making processes. But, equally important is providing
empirically derived guidance about how to help effectively combine these needs with
insights from studies about how people make decisions and, importantly, how to make
these decision better (Ajzen & Fishbein 1980).

80

The question is, how? Recent studies of “structured decision making” suggest
that in addition to the provision of information and stakeholder participation in decisionmaking processes, the quality of stakeholder input can be improved by ensuring that
people address four basic principles (Arvai & Gregory 2003; Gregory et al. 2001): (1)
thoroughly exploring and then defining what matters to affected stakeholders in the form
of clearly articulated and agreed-upon objectives; (2) creating a set of attractive and,
importantly, feasible management alternatives; (3) employing the best available
technical information to characterize the impacts of these alternatives in terms of the
agreed upon-objectives; and (4) directly confronting the value tradeoffs that inevitably
arise when objectives conflict.
While these principles build upon the concepts of multiple criteria decision
analysis (Clemen 1996; Hammond et al. 1999; Keeney 1992), it’s important to stress
that SDM is not simply “decision analysis lite”. Rather, SDM—with its foundation in
multiattribute utility theory—adds insights from good practice in analytic-deliberative
processes (NRC 1996) while also reflecting findings from behavioral decision research.
For example, being explicit about first exploring a full range of objectives and then
narrowing the list based on those areas that are most relevant (vs. most salient) to the
decision at hand helps to avoid problems associated with bounded rationality. Likewise,
being clear about the impacts of the alternatives under consideration in terms of how
they meet, or do not meet, stated objectives makes the direction of improvement from
the status quo unambiguous; this, in turn, helps to negate problems associated with
framing effects (Gregory & Slovic 1997). And, from the standpoint of both developing
the information base for decision making as well as facilitating community involvement,

81

SDM processes are recognized for integrating disparate technical knowledge and
stakeholder values while also democratizing decision-making processes.
Structured decision making (SDM) approaches have been applied to a range of
problems in North America with positive results (Arvai & Gregory 2003; Barham 2002;
Failing et al. 2007; Wilson & McDaniels 2007). There have been few applications—
successful or otherwise—of SDM, however, in developing country contexts. Among the
reasons for this are the limited time that stakeholders have to devote to these kinds of
multi-party initiatives, the distances that people must travel in order to take part, the
general lack of technical facilities where people can interact with computer-based
decision support tools, and the language and cultural barriers that often exists between
predominantly western SDM facilitators and local stakeholders. With the need for—and
obstacles to—getting meaningful input from people about community development
initiatives squarely in mind, the goal of the research reported here was to develop and
test a SDM framework that could be used to elicit input from community members about
land use decisions in developing communities of rural Costa Rica.
2. Methods
2.1 Context
Costa Rica is a middle-income developing country located in Central America bordered
by Nicaragua to the north and Panama to the south. Despite its small land area
(approx. 51,100 km2), Costa Rica is one of the 20 most biologically diverse countries in
the world and forms an integral part of the Mesoamerican Biodiversity Hotspot. But as a
developing country, Costa Rica is also keenly interested in economic growth. In
addition to electric circuit manufacturing and tourism, which are its top two revenue

82

generators, Costa Rica also relies heavily on agricultural exports to support economic
development. Costa Rica’s largest agricultural export is pineapple and its output
continues to increase.
Pineapple production in Costa Rica is dominated by foreign-owned multinational
corporations. These corporations have invested heavily the crop, including hundreds of
large plantations, packing and storage facilities, and transportation networks. (In many
cases, these corporations also make packing, storage, and export infrastructure
available to smaller scale operations so that they may get their products to market.) In
addition, they build and maintain other infrastructure that people in rural communities
use, such as roads and bridges. They also provide employment for tens of thousands
of people in rural Costa Rica.
Yet large-scale pineapple production in Costa Rica comes at a cost. In particular,
to meet export demands and comply with phytosanitary requirements, the once low
impact crop now requires significant amounts of pesticide and herbicide, which has
resulted in a buildup of agrochemicals in groundwater near some large pineapple
plantations. Not surprisingly, many residents of communities adjacent to plantations are
concerned about the impacts of agrochemicals on their air and water. At the same time,
a lack of adherence to appropriate soil conservation techniques by some producers has
resulted in severe erosion problems. There is concern among agronomists that if
unchecked, the degree of soil degradation could cause a situation so severe that
affected lands cannot be cultivated for decades. Additionally, improper post-harvest
management of the crown and leaves by some producers has created a pest problem—
predominantly in the form of Stomoxys Calcitrans—an aggressive biting fly that, if not

83

controlled, can wreak havoc on cattle production, both in terms of beef and milk. As a
result, farmers and other residents who live on lands adjacent to the pineapple farms,
and who fail to benefit directly from these multinational operations, are carrying an
unequal share of the costs.
In light of the partial accounting and uneven distribution of the benefits and costs
of pineapple production, many Costa Ricans are beginning to call for more careful
regulation of the industry. Policies being discussed range from more stringent
regulation and monitoring of the existing pineapple industry to significant limits on the
scale of production that would be allowed in the country. Some Costa Ricans have
even begun to call for an outright ban on pineapple production in certain ecologically
sensitive areas. For these reasons, the objectives of the research reported here were
twofold: First, and foremost, we aimed to help inform policy decisions by eliciting
pineapple production preferences from people living in communities surrounding
plantations. Second, we sought to study the effectiveness of the SDM approach in a
developing community context, to test whether the methods might hold promise as a
decision aid to government officials in Costa Rica and other developing countries that
are dealing with similar issues.
2.2 Design
Our research unfolded in two phases: (1) A series of individual meetings and smallgroup workshops conducted with technical experts (agronomists, soil scientists, etc.),
regional officials from the Costa Rican Ministry of Agriculture and Ministry of
Environment and Energy, pineapple producers, and community representatives
identified through contacts with community development associations and municipal

84

government offices; and (2) one-on-one SDM sessions with residents of communities
located near pineapple plantations.
Workshops: These workshops and meetings addressed the first three elements
of the SDM approach (i.e., identifying the problem; eliciting objectives; and creating
4

alternatives). A first round of workshops and consultations , which typically lasted
between 2 and 6 hours, focused on eliciting from different stakeholders and expert
groups a list of guiding objectives that participants felt should be addressed by any
future management initiatives. These objectives included ensuring the economic
viability of pineapple as an export crop, protecting human health and safety, maintaining
environmental health, ensuring that land remains cultivable if or when pineapple
production in the area ceases, creating realistic land-use policies that could be
sustained though enforcement, and maintaining access to critical infrastructure by
members of the public. All consultations and workshops were held at convenient area
locations and discussions were conducted in Spanish.
Facilitated discussions in these workshops also focused on identifying the
components of different pineapple production scenarios that would help to meet these
guiding objectives. These included: (1) reducing in the number of leachable herbicide
and pesticide applications per year as a means of protecting human and environmental
health; (2) limiting the allowable size of plantations as a means of protecting
environmental health; (3) requiring buffer zones to both prevent the airborne spread of
agrochemicals (to protect human health) and to provide a visual barrier (an aesthetic
improvement); (4) requiring that soil conservation techniques be employed by producers
4

Initial meetings and workshops took place between March, 2008 and October, 2009.
85

as a means of ensuring long-term cultivability; (5) mandating monitoring and compliance
checks to ensure that agreed-upon policies were being enforced; and (6) generating
5

revenue, in Costa Rican colones (₡) , to ensure that local infrastructure was protected
by the government if production capacity in the area was reduced, and also to help fund
more intensive environmental monitoring. The rationale behind the increase in cost
stems from the fact that large pineapple producers are helping to provide infrastructure
used by the communities. If the scale of production was to shrink, producers’ revenues
would decrease as would their ability (and motivation) to provide ancillary community
services. As a result, community members would be asked by the government to help
cover the costs of these services, which they do not wish to lose, through a monthly
contribution. Since few households in the areas where we conduced this research pay
property taxes, we had to devise a method to pass the costs associated with smaller
pineapple plantations on to local residents. All households pay fees to Grupo ICE, the
national electricity and telecommunications (television, as well as wired and wireless
phones and internet) and electricity provider; as a result, monthly bills from Grupo ICE
were selected as the revenue collection method.
6

A final round of 1-day workshops was conducted with experts and managers;
these workshops were aimed at assembling a series of five realistic management
alternatives, including the status quo (Alternative 1) as well as the expected impacts (or
effort levels) associated with these alternatives on the objectives identified in our first
round of workshops (Table 8).

5
6

During the period this research was conducted, $1 USD ≈ ₡ 523.
The second round of workshops took place in November 2009.
86

SDM Sessions: The purpose of these sessions, which lasted 30 minutes on
average, was to ask community members to evaluate the five alternatives and then rank
them from most to least acceptable. To inform the evaluation process, we developed a
software tool that ran on a laptop computer carried in the field by the research team.
The software interface depicted the five management alternatives and included a values
weighting module, which was used by respondents to prioritize objectives and rank
alternatives. The software also computed the overall subjective utility, via a linear utility
model, of each alternative based on the weights that each respondent assigned to the
attributes.
The values weighting model made use of a swing weighting (SW) procedure (von
Winterfeldt & Edwards 1986) which presented respondents with two hypothetical
alternatives: one comprised the worst possible consequences associated with all of the
alternatives under consideration and another comprised of the best possible
consequences. Respondents were then asked to identify which of the attribute pairs
they would most want to swing from the worst to the best in order to, in their minds,
make the largest improvement to the system. Respondents repeated this procedure for
all of the attributes in the set. Once all of the attributes were ranked, respondents were
then asked to assign 100 points to highest ranking attribute with the others assigned a
percentage of this weight (Table 9). A weight of zero was allowed when respondents
judged an attribute to be irrelevant (Baron 2000; Clemen 1996).
After completing the swing weighting procedure, respondents were shown the
rank order of alternatives that best reflected their stated priorities. If they desired,
respondents were also given the opportunity to alter their priorities (i.e., re-weight)

87

which, depending upon the extent to which they changed their weights, could change
the rank order of alternatives. Only the final weights, and their corresponding ranking of
alternatives, were used in our analysis. After showing respondents the preference order
implied by their weights, we also asked them to select their preferred management
alternative.
When respondents were finished with the elicitation process, they were asked to
complete a short follow-up survey. The survey, administered orally and in Spanish,
consisted of 6 questions. These questions, with responses recorded on 5-point Likert
scales, asked for respondents to provide ratings of: (1) their level of satisfaction with
their resulting choices, specifically the rank order of alternatives (where 1=very
unsatisfied and 5=very satisfied); (2) the level of difficulty associated with the decisionmaking process (where 1=very difficult and 5=very easy); (3) how well their ranking of
alternatives reflected what mattered most to them in the context of pineapple production
(where 1=very poorly and 5=very well); (4) the degree to which the decision-making
process seemed realistic in light of their knowledge about agricultural production in the
region (where 1=very unrealistic and 5=very realistic); (5) the extent to which the
decision making process seemed biased in the direction of specific alternatives (where
1=extremely biased and 5=not at all biased); and (6) how certain they were in their
ability to consider tradeoffs when making their choices (where 1=not at all certain and
5=very certain).
2.3 Respondents
We selected three cantons (Pococí, Guácimo, and Siquirres) within the Limón Province
in the Atlantic region of the country as our study site because most of the pineapple

88

production in the Atlantic region is concentrated there. A stratified random sample of 95
households was drawn from these cantons. Rather than sampling on an ad-hoc villageby-village basis, we worked with the National Institute of Statistics and Census (INEC)
based in the capital, San José to draw a stratified random sample of respondents (and
because we wanted maximize the credibility of our research by following the sampling
procedures utilized by the Government of Costa Rica when conducting the national
census and other surveys of national interest). We also wanted to avoid oversampling
respondents, both from certain high-conflict areas where strongly negative feelings
about pineapple production prevail, as well as from areas so removed from any
production—much less conflict—that respondents might have no opinion at all about the
issue. During the draw of respondents, INEC took into account each canton’s total
population and its urban, semi-urban, and rural distribution in order to produce a
representative group of sampling segments for each canton. Each of these segments
were then rendered on cadastral maps which. A sample of three households was then
randomly selected for interviews from each segment. Researchers visited a total of 287
houses (152 of which were eligible), collected 121 surveys, and ended up with a total of
95 completed surveys for an adjusted response rate of 0.63.
The SDM sessions described in this paper were conducted only with
respondents who were responsible for making financial decisions, either jointly or alone,
on behalf of the household. Thirty-eight percent of these respondents were male, and
the average age of respondents was 44 years. Forty-one percent of respondents were
originally from the Limón Province, and of that group, the average length of residency is
40 years. Respondents not originally from Limón have lived in their community for an

89

average of 26 years and in their canton for an average of 31 years. The majority of
participants, 59 percent, were married or in “free union” with their partner (41 percent
were married and 18 percent in “free union”). The average monthly income of
respondents was below the Costa Rican poverty line—<₡218,433—with 18 percent
earning that income through farming. An average of 43 percent of respondents
reported that they themselves (27 percent) or a family member (16 percent) currently
work (or had recently worked) in some aspect of pineapple production. The majority of
respondents (59 percent) had only a primary school (i.e., completion of grades 1
through 6) education and 12 percent of respondents self-identified as active members of
a community, canton, or provincial committee or organization (such as a local
environmental committee or the Rural Aqueduct Association).
2.4 Analysis
Statistical analysis of the SW data was carried out using general linear models in PASW
18. Means weights assigned to attributes by respondents were compared using
analyses of variance (ANOVA) coupled with Tukey’s post-tests (with appropriate
Bonferroni adjustments included), and when necessary, independent sample t-tests.
3. Results
Based on the results from a separate stated choice survey (Kellon et al. 2011), we
undertook several comparisons (see Tables 10 through 15) of respondents’ swing
weights (i.e., priorities) that we believed would yield significant differences across
attributes. A multivariate test of differences between groups using the Wilks’ Lambda
criterion was statistically significant for all three 3-level comparisons; i.e., respondents’
region (canton) of residence (p = 0.001), age (p = 0.015), and level of education (p =

90

7

.001). According to the analyses of variance that followed, respondents’ priorities
regarding the presence of buffer zones (p = 0.004), compliance checks (p = 0.001), and
cost (p = 0.011) were significantly different across the three study regions (Table 10).
Differences across respondents’ priorities as a function of age (Table 11) were
significant for the requirement that soil conservation practices be undertaken by
plantation owners (p = 0.045). In terms of respondents’ level of education (Table 12),
respondents’ priorities regarding the presence of buffer zones (p = 0.001), the
requirement that soil conservation practices be undertaken (p = 0.026), and cost (p =
0.039) all differed significantly.
Independent sample t-tests were performed for three additional groups:
respondents’ gender and relationship to the poverty line, and whether the respondent or
a close family member works on a pineapple plantation (Tables 13 through 15). By way
of our comparisons concerning gender (Table 13), women placed a higher average
weight on limiting both the frequency of pesticide applications (p = 0.027) and the size
of pineapple plantations (p = 0.011) while men were more concerned about cost (p =
0.001).
When we considered respondents’ level of income (Table 14), respondents who
earn at a level that is below the poverty line tend to place greater emphasis on limiting
the frequency of pesticide applications (p = 0.038). Those above the poverty line, by
contrast, placed a higher average weight on limiting plantation size (0.019), ensuring

7

Within-attribute differences, determined using a Tukey’s post-test, are shown in
Tables 10 through 12. All post-test results shown are significant by a margin of at most
p < 0.05.
91

that buffer zones be established between villages and plantations (p = 0.006), and
enforcing a larger number of random compliance checks (p = 0.006).
Those respondents who work on pineapple plantations, or have a close family
member that works on one (Table 15), placed a higher average weight on the
requirement that soil conservation be undertaken (p = 0.001) and cost (0.039). Those
respondents who neither work on a plantation, nor have a close relative who does,
placed a higher average weight on limiting the number of allowed pesticide applications
(p = 0.001) and the requirement the buffer zones be established (p = 0.026).
These results are not unimportant in that they are revealing of some key
between-subject differences. However, of greater interest to us as decision scientists
was the meaning of the average weights across groups in terms of what they might
mean for respondents’ preferences for different management alternatives. When
respondents’ mean weights were analyzed using a stepwise linear utility model, we
found that differences in average weights of fewer than 20 points across individual
attributes had virtually no impact on the overall preference order.
Specifically, an analysis of mean swing weights showed that Alternative 2 always
performed best in terms of overall utility, followed by Alternatives 3 and 5, which were
nearly indistinguishable. These alternatives were followed by Alternative 4 and
Alternative 1, respectively. After completion of the weighting task, 66 percent of
respondents told us they would select Alternative 2 if given the choice with 26 percent
and 8 percent opting instead for Alternatives 5 and 3, respectively. Alternatives 4 and 1
were never selected by respondents when given the opportunity to choose.

92

4. Discussion
The primary objective of our research was to help inform policy decisions by eliciting
judgments from stakeholders about different management options regarding the
pineapple industry in Costa Rica. From the standpoint of our workshop participants, the
SDM approach was viewed as adding much-needed precision and civility to a
consultative process which, to date, had been fraught with inconsistency and conflict. In
the workshops with experts and stakeholders, participants were able put aside longstanding conflicts and, instead, focused on a discussion of critical objectives and
realistic management alternatives. Perhaps the best measure of success for these
workshops lies in the reaction the objectives, attributes, and alternatives that were
developed received from participants in the individual SDM elicitation sessions.
Generally speaking, participants in these sessions felt that the information presented to
them for analysis did not seem biased (Table 16, Question 5; x = 4.9, sd = 0.2);
moreover, the decisions they were asked to make using this information felt very
realistic to them (Table 16, Question 4; x = 4.3, sd = 0.7).
Still, with the individual SDM elicitation sessions, we did not observe any major
differences in terms of the individual utility of the five management alternatives (Table 8)
across any of the group-level comparisons we undertook (Tables 10-15). Specifically,
across all of the group-level comparisons, neither Alternative 1 (i.e., the status quo) nor
Alternative 4 (the organic option) were at all competitive in terms of their overall utility.
Alternative 2, by contrast, performed the best across all of the groups we studied. This
option mandates a moderate reduction in the allowable size of pineapple plantations,
reduces by more than half the maximum number of agrochemical applications allowed

93

on pineapple plantations, and requires that both soil conservation practices be used and
buffer zones be established between plantations and any neighboring communities. To
ensure that these changes were implemented, Alternative 2 also calls for a moderate
increase in the number of random compliance checks by government monitors that
would take place each year. The overall cost to local households of implementing
Alternative 2 is ₡463 per month.
Respondents were largely indifferent between Alternatives 3 and 5. Since these
alternatives were substantially different from one another, we feel these results are
worthy of attention. Whereas the utility of Alternative 2 was driven largely by a focus on
reducing the number of pesticide applications and ensuring the implementation of soil
conservation practices as well as the establishment of buffer zones, Alternative 3
tended to be favored by respondents who attached higher weights to a drastic reduction
in the allowable plantation size. The utility of Alternative 5, by contrast, was linked most
strongly to respondents’ desire to increase the frequency of monitoring and/or minimize
the cost of management, or increase farm size. In terms of the remaining options, there
was a universal dislike of the status quo (the only option with no cost); even
respondents who were most concerned about the cost of management agreed that a
revision to the current management structure for pineapple in Costa Rica was needed.
At the same time, however, there was near-universal dislike of the most costly option,
Alternative 4.
In terms of our second research objective, which focused on the effectiveness of
the SDM approach in a developing community context, we observed several clear
benefits. During our discussions with respondents during the elicitation procedure,

94

people agreed that the SDM framework helped them to take what many in our study
area viewed as an intractable problem and organize it into a series of interrelated, yet
cognitively manageable, steps; e.g., thinking about objectives, connecting these
objectives to sensible attributes, reviewing the available management alternatives in a
user-friendly format, and confronting tradeoffs across the full range of their concerns. At
the same time, people were pleased that they were given the opportunity to offer their
own opinions, in a constructive fashion, about the management of an important issue in
their communities. In the past, people had been asked for their opinions through
interviews and surveys for example. However, many respondents reported that our
research was the first time that they had the opportunity to reflect on their preferences in
a more in-depth and constructive manner.
Our follow-up survey results (Table 16) seem to bear this out. For example,
respondents felt strongly that the SDM approach helped them to confront tradeoffs by
balancing the pros and cons associated with all of the alternatives (Question 6; x = 4.5,
sd = 0.5) and, therefore, make choices that reflected what mattered most to them
(Question 3; x = 4.3, sd = 0.5). It is our view that, in large part, respondents’ high level
of satisfaction with their decisions (Question 1; x = 4.2, sd = 0.8) can be attributed to
this aspect of the SDM method.
However, it is worth noting that the SDM method did not necessarily make the
decision easier for respondents (Question 2; x = 3.4, sd = 0.9). A common
misconception about SDM is that it simplifies complex decisions. In fact, there is
nothing simple about a complex problem that requires people to confront challenging
tradeoffs. What SDM does is provide people with a systematic method for clarifying the

95

multiple, related aspects of a decision and then balancing the pros and cons of
alternatives through a deliberative and iterative process. We acknowledge that this
process still can be challenging for decision makers. The benefit, however, is in helping
to ensure that the resulting decision is internally consistent.
It is also worth noting that other, seemingly more straightforward and easy,
approaches for eliciting these kinds of preferences exist. Stated choice methods
(Adamowicz et al. 1998; Boxall et al. 1996) such as conjoint analysis, for example, have
been widely used in Costa Rica (e.g., see Alpizar & Carlsson 2003). These methods
use survey instruments to present respondents (local community members, other
stakeholders) with a series of alternatives, characterized as multidimensional
composites or scenarios, that present integrated (i.e., conjoined) combinations of
different attributes (e.g., different levels of air quality, water quality, and local
employment). Combinations generally reflect actual or projected variations in the
attributes (e.g., different levels of air and water quality and local employment
opportunities). In the more sophisticated conjoint surveys, often termed “choice
experiments”, the represented attributes are specified by an experimental design that
estimates the separate and interacting effects of component attributes (Louviere 1988).
Multiple regression or similar analyses of decision makers’ choices among the
presented options are used to estimate the relative contributions of individual attributes
to their expressed preferences for the conjoint alternatives. In other words, the choices
made by people in response to survey questions are used to reveal the tradeoffs that
they are willing to make when comparing multiattribute alternatives. Stated choice
methods/choice experiments are being used with greater frequency in developing

96

countries to inform the design of PES programs. Specifically, conjoint surveys and
choice experiments are used to identify the environment-development tradeoffs that
people seem willing to make so that appropriate payment vehicles and amounts can be
designed and built into PES contracts.
However, as decision scientists, were believe that there are significant
challenges associated with stated choice approaches stemming from how judgments
requiring tradeoffs across benefits and costs are constructed. A wealth of research in
psychology and behavioral economics demonstrates that people are not the ideally
rational maximizers of multiattribute utility that many policy makers assume they are
(Ariely 2008; Kahneman et al. 1982; Tversky & Kahneman 1974). Instead, people
construct their judgments in response to contextual and experiential cues that are
available to them during the decision making process (Arvai et al. 2006; Payne et al.
1992; Slovic 1995; Slovic & Lichtenstein 2006). As a result, information about implied
tradeoffs may be misleading because the judgmental processes that led to them were
prone to judgmental inconsistencies. This problem is likely to be more significant when
the choice context is unfamiliar, which has shown to be true for many environmental
policy options (Gregory et al. 1993).
With stated choice approaches, for example, questions may be raised about the
extent to which survey respondents are actually confronting tradeoffs across all of the
attributes and levels that are being presented to them in choice experiments. This
problem, which has been termed “process validity” in the literature on choice
experiments (Carson et al. 1994; Mitchell & Carson 1995), may bedevil even the most
carefully designed process. For example, conjoint surveys assume that respondents

97

are considering the full set of described attributes when choosing among options; the
expectations are that some of these attributes will be more important than others and
choices between the presented options will provide important insights about the
tradeoffs that people are willing to make (Adamowicz et al. 1998; Green & Srinivasan
1978; Louviere 1988).
One problem here is that a given attribute weight of zero, or near zero, may not
mean that a given attribute is not important (as the zero value implies). Instead, such
an attribute may be quite important to a decision maker but may nevertheless largely be
ignored during the construction of preference. A common reason for this, which has
been studied in other contexts (Arvai & Gregory 2003; Arvai et al. 2006; Arvai et al.
2007; Gregory 2003; Gregory et al. 2001; Lichtenstein et al. 2007), is tradeoff avoidance
associated with constitutive incommensurability (Baron & Spranca 1997; Tetlock 2000;
Tetlock et al. 2000). When confronted with tradeoffs that make decision makers feel as
though they must subvert some morally significant values in favor of others, many
people respond by anchoring on a single important attribute while blocking out the
others. This occurs not because the blocked attributes are not important—quite the
opposite. These attributes are important but they are discounted because they are
difficult or uncomfortable to balance against other attributes.
The swing weighting procedure in a SDM approach counters this problem by
making the need to evaluate alternatives on an attribute-by-attribute basis explicit.
Swing weighting has a built-in advantage over other approaches—like stated choice
methods—in that it is very sensitive to the range of values that an attribute takes on. In
other words, respondents are required to consider the range of possible outcomes

98

across all of the attributes and then make a determination about which of these
ranges—and by extension, the associated attribute—is most in need of attention;
respondents then repeat this procedure for each attribute in the set (see Table 9). As a
result, the risk that respondents will make decisions based solely on a single attribute—
which we believe to be a significant shortcoming of stated choice approaches—is
reduced considerably (Clemen 1996; von Winterfeldt & Edwards 1986). At the same
time, SDM does not oversimplify the decision by unrealistically dissecting a complex
problem into its component parts (which tends to falsely portray them as unrelated).
Rather, after helping decision makers evaluate each attribute and their levels, SDM’s
holistic approach and explicit focus on deliberative tradeoff analysis helps decision
makers weigh the pros and cons of the consequences of different combinations of
attribute levels.
To conclude, decisions focused on managing environmental systems present
some of the most imposing challenges to policy makers, scientists, and stakeholders
alike. Adding to these challenges, decisions about the environment are not simply
decisions about the environment; they affect individuals, communities, as well as
business and commercial interests. As a result, these decisions must contend with
multiple, conflicting, and often poorly understood values that, in turn, are linked to
social, cultural, spiritual, economic, as well as other objectives and concerns.
Complicating matters further is the fact that these decisions must often be made under
severe time pressure, either because of impending environmental damage or because
of political constraints faced by decision makers. These challenges are made all the
more pressing and significant in developing communities, where most decisions about

99

the environment directly influence the livelihood of one or more stakeholder groups.
Due to the precarious socio-economic status of people in these areas, these direct—
and even indirect—influences can have devastating effects on the ability of resourcedependent individuals and communities to sustain themselves (Kellon & Arvai 2011).
Together, these challenges motivated our research. In the end, we believe our findings
speak volumes about the value of applying SDM approaches in an international
development context.

100

APPENDICES

101

APPENDICES

Table 8. Alternatives developed for evaluation by stakeholders. *Alternative 1 reflects
the status quo.
Objective

Attribute

*Alt. 1

Alt. 2

Alt. 3

Alt. 4

Alt. 5

Human &
Env. Health

Allowed number of
leachable
agrochemical
applications/year

4

1

1

0

2

Env. Health

Maximum plantation
Year
size

>250
ha

50250
ha

<50 ha

<50 ha

>250
ha

Human
Health

Requirement that
external buffer zones
be established

No

Yes

No

No

Yes

Long-Term
Cultivability

Requirement that soil
conservation
techniques be used

No

Yes

Yes

No

No

Enforcement

Number of random
compliance
checks/year

0

1

1

>3

>3

Infrastructure &
Enforcement

Monthly household
fee on Grupo ICE bill

₡0

₡463

₡1,837

₡2,452

₡105

102

Table 9. Worst and best ranges, derived from the options presented in Table 8, used in
the swing weighting procedure. Italicized figures depict the ranks and weights elicited
from a single respondent.
Objective
Human &
Env. Health
Env. Health
Human
Health
Long-Term
Cultivability
Enforcement

Attribute
Allowed number of
leachable
agrochemical
applications/year

Worst

Best

4

0

1

100

Maximum plantation
size
Year

>250
ha

<50
ha

5

65

No

Yes

2

90

No

Yes

3

78

0

>3

6

60

₡2,452

₡0

4

70

Requirement that
external buffer zones
be established
Requirement that soil
conservation
techniques be used
Number of random
compliance
checks/year

Infrastructure
Monthly household fee
&
on Grupo ICE bill
Enforcement

103

Rank Weight

Table 10. Mean weights (wt.) and standard error (se) elicited for each attribute by
region. For the Tukey’s post-test, all significant differences are at the p < 0.05 level
where: 1 corresponds with Pococí ≠ Guácimo; 2 corresponds with Guácimo ≠ Siquirres;
and 3 corresponds with Pococí ≠ Siquirres.
Attribute
Allowed number of
leachable
agrochemical
applications/year
Maximum plantation
size
Year
Requirement that
external buffer
zones be
established
Requirement that
soil conservation
techniques be used
Number of random
compliance checks
per year
Monthly household
fee on Grupo ICE
bill

Pococí
wt.
se

Guácimo
wt.
se

Siquirres
wt.
se

pANOVA

PostTest

5.2

0.203

NA

69.0 23.3 76.5 19.4 80.70 19.7

0.172

NA

89.5

7.1

81.5 17.8 84.30 10.9

0.004

1,3

93.1

5.3

87.4 14.8

87.6

13.3

0.057

NA

94.8

8.4

85.6 10.9

80.2

17.3

0.001

1,3

92.5 10.5 77.7 23.6

86.1

19.0

0.011

1

92.3

8.1

n=31

92.3 13.3

n=31

104

95.5

n=33

Table 11. Mean weights (wt.) and standard error (se) elicited for each attribute by age.
For the Tukey’s post-test, all significant differences are at the p < 0.05 level where: 1
corresponds with 18-40 Years ≠ 41-60 Years; 2 corresponds with 41-60 Years ≠ >60
Years; and 3 corresponds with 18-40 Years ≠ >60 Years.
Attribute

18-40
Years
wt.
se

41-60
Years
wt.
se

>60 Years
wt.
se

pANOVA

PostTest

Allowed number of
leachable
agrochemical
applications/year

94.0

87.8 15.4

0.345

NA

Maximum plantation
size
Year

74.3 20.6 81.4 18.9 66.3 25.0

0.111

NA

Requirement that
external buffer zones
be established

82.1 14.6 86.3

90.9 13.0

0.067

NA

Requirement that soil
conservation
techniques be used

92.9 10.1 86.4 12.5 85.6 14.1

0.045

NA

Number of random
compliance checks
per year

88.7 10.2 84.0 16.2 86.9 18.4

0.263

NA

Monthly household
fee on Grupo ICE bill

83.6 20.9 85.6 18.6 90.0 16.3

0.686

NA

7.7

n=45

95.1

7.0

9.3

n=34

105

n=16

Table 12. Mean weights (wt.) and standard error (se) elicited for each attribute as a
function of education level (characterized by the number of years of formal education).
For the Tukey’s post-test, all significant differences are at the p < 0.05 level where: 1
corresponds with 0-6 Years ≠ 7-11 Years; 2 corresponds with 7-11 Years ≠ >11 Years;
and 3 corresponds with 0-6 Years ≠ >11 Years.
7-11
Years
wt.
Se

Attribute

0-6 Years
wt.
se

Allowed number
of leachable
agrochemical
applications/year

92.1 10.8 96.8

Maximum
plantation size
Year
Requirement that
external buffer
zones be
established
Requirement that
soil conservation
techniques be
used
Number of
random
compliance
checks per year
Monthly
household fee on
Grupo ICE bill

>11 Years
wt.
se

pANOVA

PostTest

8.8

0.065

NA

73.1 21.7 81.6 18.2 72.0 24.4

0.351

NA

87.9

0.001

2,3

4.2

0.010

3

86.5 15.8 87.5 10.9 86.0 12.7

0.824

NA

89.9 16.6 81.8 20.0 70.0 23.1

0.008

3

9.6

91.0

84.8 11.7 70.0 21.6

86.5 14.0 92.4

n=57

5.1

7.1

n=28

106

96.9

n=10

Table 13. Mean weights (wt.) and standard error (se) elicited for each attribute as a
function of gender.
Attribute

Women
wt.
se

Men
wt.
se

p

Allowed number of
leachable
agrochemical
applications/year

95.6

6.0

89.7

12.5

0.027

Maximum plantation
size
year

80.2

20.7

67.8

19.9

0.011

Requirement that
external buffer zones
be established

83.7

13.7

87.4

11.4

0.244

Requirement that soil
conservation
techniques be used

90.2

12.4

88.0

11.6

0.324

84.1

12.4

91.0

15.8

0.096

80.8

20.7

93.1

13.9

0.001

Number of random
compliance checks
per year
Monthly household
fee on Grupo ICE bill

N=59

107

N=36

Table 14. Mean weights (wt.) and standard error (se) by attribute relative to the poverty
line (PL) of ₡218,433 per month.
Attribute

< Poverty
wt.Line se

> Poverty
wt.Line se

P

Allowed number of
leachable agrochemical
applications/year

94.5

6.6

91.9

12.0

0.038

Maximum plantation size

69.7

20.9

82.6

21.4

0.019

Year
Requirement that
external buffer zones be
established

81.2

15.7

88.9

7.5

0.006

Requirement that soil
conservation techniques
be used

88.0

11.9

90.7

12.4

0.226

Number of random
compliance checks per
year

83.3

16.3

90.9

10.1

0.006

Monthly household fee
on Grupo ICE bill

82.9

20.2

89.7

16.6

0.149

n=49

108

n=43

Table 15. Mean weights (wt.) and standard error (se) by attribute relative to
respondents’ relationship with the pineapple industry. Respondents reporting “yes”
either worked in—or had a close relative or family member working in—the pineapple
industry.
Attribute

Family: No
wt.
se

Family: Yes
wt.
se

Allowed number of
leachable agrochemical
applications/year

97.0

4.8

88.5

11.7

0.001

Maximum plantation size

77.1

20.6

73.4

22.0

0.283

Requirement that external
Year
buffer zones be
established

87.7

9.0

81.5

16.3

0.026

Requirement that soil
conservation techniques
be used

85.2

12.7

94.9

8.7

0.001

Number of random
compliance checks per
year

86.5

14.8

87.0

13.3

0.864

Monthly household fee on
Grupo ICE bill

81.3

19.6

90.9

17.7

0.039

n=54

109

n=41

P

Table 16. Follow-Up survey results.
Follow-Up Survey Questions

x

sd

1. Satisfaction with choices

4.2

0.8

2. Difficulty with decision-making
process

3.4

0.9

3. Choices reflect what matters

4.3

0.5

4. Choices seemed realistic

4.3

0.7

5. Information seemed unbiased

4.9

0.2

6. Ability to balance pros and
cons

4.5

0.5

110

REFERENCES

111

REFERENCES

Adamowicz, W., P. Boxall, M. Wilhams, and J. Louviere. 1998. Stated preference
approaches for measuring passive use values: Choice experiments and
contingent valuation. American Journal of Agricultural Economics 80:64-67.
Ajzen, I. 2001. Nature and Operation of Attitudes. Annual Review of Psychology 52:2758.
Ajzen, I., and M. Fishbein 1980. Understanding Attitudes and Predicting Social
Behavior. Prentice-Hall, Inc., Upper Saddle River.
Alpizar, F., and F. Carlsson. 2003. Policy implications and analysis of the determinants
of travel mode choice: An application of choice experiments to metropolitan
Costa Rica. . Environment and Development Economics 8:603-619.
Ariely, D. 2008. Predictably Irrational: The Hidden Forces That Shape Our Decisions.
Harper Collins, New York, NY.
Arvai, J. L., and R. Gregory. 2003. Testing alternative decision approaches for
identifying cleanup priorities at contaminated sites. Environmental Science &
Technology 37:1469-1476.
Arvai, J. L., R. Gregory, D. Ohlson, B. A. Blackwell, and R. W. Gray. 2006. Letdowns,
wake-up calls, and constructed preferences: People's responses to fuel and
wildfire risks. Journal of Forestry 104:173-181.
Arvai, J. L., R. Gregory, and M. Zaksek. 2007. Improving wildfire risk management
decisions in W. Martin, I. Martin, and C. Raish, editors. Wildfire and Fuels
Management: Human Perceptions and Management Implications. RFF Press,
Washington, DC.
Barham, E. 2002. Towards a theory of values-based labelling. Agriculture and Human
Values 19:349-360.
Baron, J. 2000. Measuring value tradeoffs: problems and some solutions in E. U.
Weber, J. Baron, and G. Loomes, editors. Conflict and Tradeoffs in Decision
Making: Essays in Honor of Jane Beattie. Cambridge University Press, New
York, NY.
Baron, J., and M. Spranca. 1997. Protected values. Organizational Behavior and
Human Decision Processes 70:1-16.
Bond, S., K. Carlson, and R. L. Keeney. 2008. Generating objectives: Can decision
makers articulate what they want? Management Science 54:56-70

112

Boxall, P., W. Adamowicz, M. Williams, J. Swait, and J. Louviere. 1996. Comparison of
stated preference approaches to the measurement of environmental values.
Ecological Economics 18:243-253.
Carson, R. T., J. J. Louviere, D. A. Anderson, P. Arabie, D. S. Bunch, D. A. Hensher, R.
M. Johnson, W. F. Kuhfeld, D. Steinberg, J. Swait, H. Timmermans, and J. B.
Wiley. 1994. Experimental analysis of choice. Marketing Letters 5:351-367.
Chambers, R. 1994. Participatory Rural Appraisal (PRA): Analysis of Experience. World
Development 22:1253-1268.
Clemen, R. T. 1996. Making Hard Decisions: An Introduction to Decision Analysis.
PWS-Kent Publishing Co., Boston, MA.
Failing, L., R. Gregory, and M. Harstone. 2007. Integrating science and local knowledge
in environmental risk management: A decision-focused approach. Ecological
Economics 64:47-60.
Fishbein, M., and J. N. Cappella. 2006. The role of theory in developing effective health
communications. Journal of Communication 56:S1-S17.
Fishbein, M., and M. C. Yzer. 2003. Using theory to design effective health behavior
interventions. Communication Theory 13:164-183.
Green, P. E., and V. Srinivasan. 1978. Conjoint analysis in consumer research: Issues
and outlook Journal of Consumer Research 5:102-123.
Gregory, R. 2003. Incorporating value tradeoffs into community-based environmental
risk decisions. Environmental Values 11:461-488.
Gregory, R., J. L. Arvai, and T. McDaniels. 2001. Value-focused thinking for
environmental risk consultations. Research in Social Problems and Public Policy
9:249-275.
Gregory, R., S. Lichtenstein, and P. Slovic. 1993. Valuing environmental resources: A
constructive approach. Journal of Risk and Uncertainty 7:177-197.
Gregory, R., and P. Slovic. 1997. A constructive approach to environmental valuation.
Ecological Economics 21:175-181.
Hammond, J., R. L. Keeney, and H. Raiffa 1999. Smart Choices: A Practical Guide to
Making Better Decisions. Harvard Business School Press, Cambridge, MA.
Hsee, C. K. 1996. The evaluability hypothesis: An explanation for preference reversals
between joint and separate evaluations of alternatives. Organizational Behavior
and Human Decision Processes 67:247-257.

113

Kahneman, D., P. Slovic, and A. Tversky 1982. Judgment Under Uncertainty: Heuristics
and Biases. Cambridge University Press, Cambridge, UK.
Kahneman, D., and A. Tversky. 1979. Prospect theory: An analysis of decision under
risk. Econometrica 47:263-291.
Keeney, R. L. 1992. Value-focused Thinking. A Path to Creative Decision Making.
Harvard University Press, Cambridge, MA.
Keeney, R. L., and H. Raiffa 1993. Decisions With Multiple Objectives: Preferences and
Value Tradeoffs. Cambridge University Press, Cambridge, UK.
Kellon, D. S., and J. Arvai. 2011. Five propositions for improving decision making about
the environment in developing communities: Insights from the decision sciences.
Journal of Environmental Management In Press.
Lichtenstein, S., R. Gregory, and J. Irwin. 2007. What's bad is easy: Taboo values,
affect, and cognition. Judgment and Decision Making 2:169-188.
Louviere, J. J. 1988. Analysing decision making: Metric conjoint analysis. Sage
University Papers Series in Quantitative Applications in the Social Sciences.
Sage Publications, Newbury Park, CA.
Millennium Ecosystem Assessment 2005. Ecosystems and Human Well-being:
Synthesis. Island Press, Washington, DC.
Mitchell, R., and R. Carson. 1995. Current issues in the design, administration, and
analysis of contingent valuation surveys in P. Johansson, B. Kristrom, and K.
Maler, editors. Current issues in environmental economics. Manchester
University Press, New York, NY.
National Research Council 1996. Understanding Risk: Informing Decisions in a
Democratic Society. National Academy Press, Washington, DC.
Ostrom, E. 1996. Crossing the great divide: Coproduction, synergy, and development.
World Development 24:1073-1087.
Payne, J. W., J. R. Bettman, and E. J. Johnson. 1992. Behavioral decision research: A
constructive processing perspective. Annual Review of Psychology 43:87-132.
Rhoades, R. 1998. Participatory Watershed Research and Management: Where the
Shadow Falls. in S. A. a. R. L. Program, editor. Gatekeeper Series. IIED,
London.
Rivers, L., and J. L. Arvai. 2007. Win some, lose some: The effect of chronic losses on
decision making under risk. Journal of Risk Research 10:1085-1099.

114

Simon, H. A. 1955. A behavioral model of rational choice. Quarterly Journal of
Economics 69:99-118.
Slovic, P. 1995. The construction of preference. American Psychologist 50:364-371.
Slovic, P., M. L. Finucane, E. Peters, and D. G. MacGregor. 2002. The affect heuristic.
Pages 397-420 in T. Gilovich, D. Griffin, and D. Kahneman, editors. Intuitive
Judgment: Heuristics and Biases. Cambridge University Press, Cambridge, UK.
Slovic, P., and S. Lichtenstein 2006. The Construction of Preference. Cambridge
University Press, Cambridge, UK.
Tetlock, P. E. 2000. Coping with trade-offs: Psychological constraints and political
implications in A. Lupia, S. L. Popkin, and M. D. McCubbins, editors. Elements of
Reason: Cognition, Choice, and the Bounds of Rationality. Cambridge University
Press, Cambridge, UK.
Tetlock, P. E., O. V. Kristel, S. B. Elson, M. C. Green, and J. S. Lerner. 2000. The
psychology of the unthinkable: Taboo tradeoffs, forbidden base rates, and
heretical counterfactuals. Journal of Personality and Social Psychology 78:853870.
Tversky, A., and D. Kahneman. 1973. Availability: a heuristic for judging frequency and
probability. Cognitive Psychology 5:207-232.
Tversky, A., and D. Kahneman. 1974. Judgment under uncertainty: Heuristics and
biases. Science 185:1124-1131.
Tversky, A., and D. Kahneman. 1981. The framing of decisions and the psychology of
choice. Science 211:453-458.
von Winterfeldt, D., and W. Edwards 1986. Decision Analysis and Behavioral Research.
Cambridge University Press, Cambridge, UK.
Wilson, C., and T. McDaniels. 2007. Structured decsion-making to link climate change
and sustainable development. Climate Policy 7:353-370.

115

CONCLUSION

The objective of this research was to better understand, and gauge the effectiveness of,
two approaches for confronting difficult environment-development tradeoffs. A
comparison of the findings from the stated choice and SDM treatments reveals
strengths and weaknesses of both approaches with respect to their practical
implications, as well as the appropriateness of the methods in similar decision-making
contexts.
The conjoint method is, on one hand, an extremely efficient data collection
method that lends itself to powerful statistical analysis. On the other hand, the method
assumes rational, utility maximizing actors; the options generated by orthogonal design
are not entirely realistic; respondents are asked to compare sets of conjoined attributes
without first evaluating each of the attributes and their associated levels, thereby
increasing the risk of anchoring on highly affective attributes; and respondents cannot
see the consequences of their true weights, which means they cannot make informed
tradeoffs.
By contrast, the swing weighting method acknowledges the constructed nature of
respondents’ preferences; presents respondents with realistic options (actual production
scenarios); gives respondents the opportunity to deliberate about their choices, consider
tradeoffs, and change their decisions; and allows respondents to evaluate both
individual attributes and the consequences of combining different levels of attributes,
which helps to inform tradeoffs. The weakness of this method is that interviewers
require more training since it is not a simple survey, therefore making it a potentially

116

less efficient data collection method. Efficiency could be further compromised if
respondents are not familiar with assigning weights, as the interviews would tend to
take longer to conduct.
Results from respondents’ own evaluations revealed that the people who went
through the swing weighting process reported a higher degree (statistically significant)
of satisfaction with their decision-making process than those who responded using the
conjoint protocol. Specifically, the swing weighting respondents were generally more
satisfied with their choices; the choices they made better reflected what matters most to
them; the options they chose seemed more realistic; the information they considered
when making their choices seemed unbiased; and it was easier for them to balance the
pros and cons of each of the options. Interestingly, there was no difference between
treatment groups regarding the perceived difficulty of making choices.
These findings show that confronting difficult tradeoffs is, by nature, difficult—
regardless of the method used. That said, SDM appeared to help respondents improve
their decision-making process by better understanding their own preferences and
preference orders; making more informed and deliberated tradeoffs; enjoying the
process more (by not being annoyed by options that seemed biased or unrealistic and
by not being confused since they understood their own decision criteria); and being
more satisfied with their choices.
The practical implications of findings from both treatments show that while both
methods produced useful findings, the SDM approach revealed what appears to be a
more holistic—and therefore, more policy relevant—understanding of residents’
preferences regarding the pineapple production practices they were asked to consider.

117

Essentially, respondents from the conjoint treatment appear to be less willing to make
tradeoffs across the different attributes and levels than those from the SDM treatment,
and importantly, it is not clear why conjoint respondents were willing (or unwilling) to
make tradeoffs. Conversely, the SDM results reveal a greater willingness on behalf of
respondents to make tradeoffs across attributes and levels, and due to the way
respondents worked through the swing weighting method (ranking and weighting
attributes first, evaluating the production alternative their weights produced, and
repeating the process until they were satisfied with their choice), they were able to
understand and express their tradeoff analysis process. The SDM treatment’s more
nuanced findings regarding community members’ main concerns and their willingness
to make tradeoffs when addressing these concerns (and their reasons for doing so) will
be critically important to the stakeholders working to improve the environmental
sustainability of pineapple production in the Atlantic Region of Costa Rica.
In the end, both methods produced valuable results, and it is clear that for certain
kinds of decision processes, a stated choice approach would be entirely appropriate.
Basically, decisions that are not complex or highly affective and for which respondents’
preferences and preferences orders already exist typically do not require much
deliberation and a stated choice method would be very useful. However, this research
suggests that for decisions similar in nature or complexity to the one presented here, a
stated choice approach would tend to oversimplify the decision problem. It would also
generalize the inherent tradeoffs by not giving respondents the opportunity to
adequately construct their preferences, consider the impact of these preferences, and
further evaluate their willingness to make tradeoffs before making their final choice. For

118

these reasons, this research indicates that a SDM approach such as swing weighting
has great potential for helping stakeholders address difficult environment-development
decision problems. To be sure, more research is needed in order to evaluate the
performance of swing weighting and other SDM methods in similar decision contexts, as
well as to improve these methods by finding more efficient and context-appropriate
ways to use them as part of data collection in the field.

119