Ecological Economics 103 (2014) 44–55

Contents lists available at ScienceDirect

Ecological Economics

j ourna l homepage: www.e lsev ie r .com/ locate /eco lecon

Payments for ecosystem services and landowner interest:Informing program design trade-offs in Western Panama

Esther Alice Duke a,1, Joshua H. Goldstein a, Tara L. Teel a, Ryan Finchum a, Heidi Huber-Stearns a, Jorge Pitty b,Gladys Beatriz Rodríguez P. c, Samuel Rodríguez c, Luis Olmedo Sánchez b

a Human Dimensions of Natural Resources Department, Colorado State University, 1480 Campus Delivery, Fort Collins, CO 80523-1480, United Statesb La Fundación para el Desarrollo Integral del Corregimiento de Cerro Punta, Cerro Punta, Chiriquí, Panamác La Fundación Vida, Salud, Ambiente y Paz, Boquete, Chiriquí, Panamá

E-mail address: Esther.Duke@ColoState.edu (E.A. Duke1 Tel.: +1 970 491 2197; fax: +1 970 491 2255.

http://dx.doi.org/10.1016/j.ecolecon.2014.04.0130921-8009/© 2014 Elsevier B.V. All rights reserved.

a b s t r a c t

a r t i c l e i n f o

Article history:Received 17 March 2013Received in revised form 20 December 2013Accepted 16 April 2014Available online xxxx

Keywords:ConservationEligibilityIncentivesPrivate landsSocioeconomic status

Experience with payments for ecosystem services (PES) highlights the effects of program design on landownerparticipation, impacting the program's ability to achieve environmental and, where applicable, social objectives.We conducted an exploratory study inwestern Panama at the initial stage of PES consideration to identify poten-tial landowner interest in PES and factors that would affect landowner interest and eligibility. We report the re-sults from a household survey of 344 farmers and ranchers (92% response rate). Eighty percent of therespondents expressed interest in PES participation. Respondents' stated interest was significantly related tofarm size, income, age, land tenure, and previous involvement in conservation. We also found that alternativespecifications for landowner eligibility requirements, targeting criteria, and other parameters could greatly affectlandowners' ability to participate, most strongly for respondents lower in socioeconomic status. We provide aframework for exploring potential landowner interest in PES at the very first stage of program exploration,from which program design can be strategically advanced with realistic PES scenarios to explore efficient pay-ment levels and projected environmental benefits. Our findings highlight the importance of making explicittrade-offs that result from alternative PES design choices in affecting landowners' interest and eligibility toparticipate.

© 2014 Elsevier B.V. All rights reserved.

1. Introduction

Payment for ecosystem services (PES) is an institutional mechanismbeing deployed globally to internalize the economic value of ecosystemservices into societal decisions (Tacconi, 2012). PES is generally consid-ered first and foremost a tool to achieve environmental objectives. Inpractice, however, many programs in developed and developing coun-tries incorporate implicit or explicit social objectives related to liveli-hoods, poverty alleviation, and/or rural development (Li et al., 2011;Turpie et al., 2008). Including social objectives generally leads to effi-ciency loss in achieving environmental objectives given budget limita-tions, yet this trade-off may be acceptable to policymakers and add tolong-term program viability (Milder et al., 2010; Muñoz-Piña et al.,2008). For example, national-scale government programs in China,Mexico, Ecuador, and Vietnam (among others) have mechanisms inplace specifically targeting the poor (de Koning et al., 2011; Wunder,2008).

To achieve its objectives, PES needs to effectively engage supply anddemand-side participants, and generally also intermediaries who

).

facilitate program administration and transactions. The supply-side re-lates to participation by landowners and land managers, often agricul-tural producers. In designing a PES program to enable supply-sideparticipation (once environmental and social, where applicable, objec-tives are defined), decision-makers must make choices about programstructural factors such as landowner eligibility requirements (e.g., min-imum enrolled parcel size), land targeting criteria (e.g., steeply-slopedlands), and payment levels. These specifications in turn expand orlimit the potential pool of interested landowners who are actually ableto participate. A particular area of focus for research has been on howprogram structure affects landowners' ability to participate at differentsocioeconomic levels, with special concern for adverse impacts on thepoor. This has led to discussion of pro-poor PES programs that also ad-dress poverty concerns (Milder et al., 2010). For example, Pagiolaet al. (2005) describe these programs as ones that maximize their po-tential positive impacts and minimize their potential negative impactson the poor by (1) keeping transaction costs low, (2) providing targetedassistance to encourage participation, (3) avoiding implementation ofprograms in areas with conflicts over land tenure ((land rights, owner-ship, and/or possession of a title to the land), (4) providing access tocredit if needed, and (5) ensuring the social context is well understood.If PES programs do not have eligibility guidelines designed with the

45E.A. Duke et al. / Ecological Economics 103 (2014) 44–55

poorest households in mind, then payments to those who are eligiblecould exacerbate social inequalities (Milder et al., 2010; Wunder,2008). Indeed, experience to-date demonstrates that PES impacts onpoverty prevention, alleviation, livelihoods, and landowner participa-tion more broadly, depend upon site-specific characteristics and pro-gram structure, highlighting the need to consider these supply-sidefactors at the program design stage (Engel et al., 2008; Pagiola et al.,2005).

Recognizing this need, we worked with stakeholders in westernPanama to collect information to guide the initial stage of regional PESexploration. At the time of this study Panama has not implemented anational PES program, yet PES development is being explored at region-al and national scales (Lichtenfeld, 2007). Our objective was to conductan exploratory study to gather information about the supply-side of aPES program in western Panama that would help decision-makers un-derstand how landowner characteristics and program structure wouldaffect landowners' interest and ability to participate in a future PESscheme. We chose this objective in collaboration with local stake-holders, who expressed that one of their first program explorationgoals was to investigate PES supply-side factors.

To achieve our objective, we developed and applied a frameworkthat addresses four questions: (1) Are landowners familiar with thePES approach? (2) Do landowners support PES development, and ifyes,which type of program focus (e.g., forest conservation, reforestation,agroforestry) would be most attractive? (3) Which landowner charac-teristics affect stated interest to participate in a future PES program?And (4) how would decisions about specific program design elements(e.g., landowner eligibility criteria) affect the ability of landownersfrom different socioeconomic classes to participate in a PES scheme?By answering these questions through a detailed case study in westernPanama, we aimed to inform PES development in the study region, aswell as contribute to the broader knowledge base that policymakersglobally can use to determine the appropriateness of, and guide the de-sign of, PES as a tool to achieve environmental and social objectives.

2. Methods

2.1. Study Area

Our project was conducted in the Chiriquí province of westernPanama, specifically in the buffer zone of La Amistad Bi-nationalUNESCO Biosphere Reserve and World Heritage Site (El ParqueInternacional LaAmistad; PILA). PILA is an important focus for conserva-tion efforts in the region, as it contains one of the region's largest ex-panses of primary forest, with approximately one million hectaresunder protection harboring an estimated 4% of Earth's species (Clarket al., 2006). Agriculture is a major economic activity in this region,with products going to regional and national markets. The predominantcrops include coffee, potatoes, plantains, and other vegetables.Most op-erations contain a mix of crops and livestock.

Conservation efforts in the region are threatened by challenges relat-ed to management of PILA (e.g., funding limitations, agricultural en-croachment) and changes in the protected area buffer zone includingagricultural intensification (e.g., increasing pesticide use) and landslidesfrom farming on steeply-sloped lands (The Nature Conservancy, 2007).This once remote area is now in danger of losing its UNESCO BiosphereReserve designation due to concerns over the impacts of these problems(UNESCO, 2008). Recognizing these challenges, government agenciesand non-governmental organizations (NGOs) are investigating PES asa strategy for addressing these concerns.

We focused on two specific sites within the PILA buffer zone wherePES is being considered: (1) the “Boquete site” encompassing the dis-trict of Boquete including the town of Alto Boquete and the surroundingcountryside, and (2) the “Renacimiento site” encompassing the districtsof Renacimiento and Bugaba including the town of Cerro Punta, thesmall neighboring community of Guadalupe, and communities along

the road to the Costa Rican border (Fig. 1). The two sites are similar intheir agricultural profile and conservation challenges. One difference isthat the Boquete site has been experiencing an influx of expatriatesand retirees, leading to rising land prices and pressure to sell agricultur-al lands to developers (The Nature Conservancy, 2007).

2.2. Community-Based Participatory Research

To develop a researchprogram to assist local stakeholders, we used acommunity-based participatory research approach (Altman, 1995). Welaunched our project inMay 2009 by facilitating a two-day collaborativeworkshop on ecosystem services in the town of Guadalupe (district ofBugaba, Panama). The workshop involved 21 participants fromPanama and Costa Rica including protected area managers and otherconservation practitioners, local farmers, and other stakeholders whoall shared a common interest in effective management of PILA and thepark's buffer zone. Information gathered through the workshop includ-ed a map showing qualitatively the supply and demand of ecosystemservices for communities in the study region, a prioritized list of themost important ecosystem services produced in the region, and a listof opportunities and challenges for using PES to support farmers andconservation efforts in the region. This information was used to guidedevelopment of the PES survey described below.

Throughout the project, local partners served as “boundary-span-ning” agents (Reid et al., 2009) helping to connect the researchers,local actors, and policymakers and to reduce concerns about distrustof extractive research. Our primary partner in the Boquete site was thenon-profit Fundación Vida, Salud, Ambiente y Paz (FUNDAVISAP),which provides leadership in organizing local groups and the local gov-ernment to have an active, participatory, proactive role in the land-useplanning process. Our primary partner in the Renacimiento site wasthe non-profit Fundación para el Desarrollo Integral del Corregimientode Cerro Punta (FUNDICCEP), whose main objective is to promote sus-tainable development for communities in the PILA buffer zone.

2.3. Sampling

Our study population consisted of landownerswith farms or ranchesin the two sites described above. We compiled a master producer listfrom multiple crop-specific lists provided by Panama's Ministry ofAgricultural Development (Ministerio de Desarrollo Agropecuario dePanamá; MIDA), local farming cooperatives, and the Agricultural Devel-opment Bank (Banco de Desarrollo Agropecuario; BDA). The lists avail-able were similar but not identical for the two study sites.

For the Boquete site, we obtained lists of onion and potato farmersfrom 2007 and 2009, coffee producers from 2009, and local coffee coop-erative members from 2009. After removing duplicates, our combinedlist for this site included 747 farmers. For the Renacimiento site, we ob-tained lists of onion and potato farmers from 2007 and 2009, farms test-ed for coffee plant disease from 2008, and participants in MIDA farmerassistance programs from 2009. After removing duplicates, our com-bined list for this site included 910 farmers. According to local officials,our combined master list accounted for almost all farmers in our studysites. We randomly selected a representative sample of 500 producersfrom our master list, including 250 from each site. A total of 374 pro-ducers from this samplewere successfully located across both sites. Pro-ducers are not required to register or update contact information, andsome included on the original list had moved, passed away, or soldtheir farms.

2.4. Survey Design and Data Collection

A collaborative research team of academic, government, and NGOpartners designed the survey instrument to gather information aboutlandowners' potential interest and ability to participate in a regionalPES program that could be developed in the future (Appendix A).

Fig. 1.Map of the western Panama study region highlighting the study sites (dark grey) and protected areas (stripes).

46 E.A. Duke et al. / Ecological Economics 103 (2014) 44–55

Survey development was informed by findings from published studieson interest/willingness to participate in hypothetical conservation in-centive programs (notably Kramer and Jenkins, 2009), along with nu-merous studies on actual PES program participation and informationgathered from our May 2009 workshop.

The survey collected information on five topics: (1) descriptive in-formation about the respondents' farm or ranch and their managementof it, (2) land uses and types of crops on the property, (3) prior partici-pation in conservation programs, (4) potential interest in participatingin alternative PES scenarios with relative ranking of preferences acrossscenarios, and (5) respondents' demographic information and house-hold characteristics.

For survey topic #4, we included three PES scenarios based upongeneral land-use types that are appropriate for the study region andwhich are incorporated in neighboring Costa Rica's PES program:(1) “forest conservation”: conservation of existing forest; cultivationof crops and livestock grazing would be prohibited; (2) “reforestation”:prescribed planting of a variety of native trees, with a total minimumdensity of 70 trees per hectare; cultivation of crops and livestock grazingwould be prohibited; and (3) “agroforestry”: growing trees in combina-tion with managing land for crops or livestock production; the land-owner would be required to plant native trees on the property (iftrees were not already present), but the landowner would still beallowed to grow crops or raise livestock. Respondents who expressedinterest in a given scenario were then asked a series of questionsabout how much land they would enroll if the program was offered intheir community, how they would prefer the program to be structured,what type of program administrator they would prefer to work with,and factors related to potential eligibility barriers.

Eligibility requirements for landowner participation are a commonfeature of PES programs (Wunder, 2008). While such requirementsmay be desirable for certain reasons (e.g., to lower transaction coststhat can arise from working with a large number of small landowners;aggregating larger parcels of greater conservation value), they also prac-tically reduce the pool of eligible landowners. Of particular concern can

be disproportionately negative impacts on the eligibility to participateof poorer households (Hope et al., 2005; Pagiola et al., 2005). We didnot include eligibility requirements related to a parcel's conservationvalue, since our research focused on understanding programdesign fac-torsmost likely to impact programaccess by landowners across levels ofsocioeconomic status. Land quality, ecosystem service additionality, andpotential efficiency loss associated with the inclusion of social variablesare all considerations thatmust beweighed carefullywhenmaking pro-gram design decisions, though these are beyond the scope of thisanalysis.

Our exploratory study aimed to collect information about land-owners' overarching preferences that could then be used in futurework to present more targeted PES scenarios at more advanced pro-gram design stages. Accordingly, our three PES scenarios were purpo-sively general and descriptive. We did not include payment amountsnor specify the identity of buyers, since these were not plausibly defin-able at this initial stage of research. Respondents' stated interest in PESparticipation, as measured here, was based upon their reaction to thePES approach and the general land-use scenarios presented to them,and it did not yet account for a financial calculation or considerationof quantitative environmental benefits. In the Discussion section, weevaluate the inferences that can be drawn from our survey, as well asits limitations and future research needs.

We worked with our local partner organizations to train localPanamanian interviewers to administer and help refine the PES survey.These local partners provided input to refine survey questions, and eachinterviewer received training on survey administering techniques in-cluding conductingmock interviews prior to data collection. The surveywas administered orally in Spanish by interviewers visiting eachlandowner's house and reading the questions to the head of the house-hold and recording his/her responses. In this way, we avoided responsebias due to varying levels of literacy.When phone numbers or email ad-dresseswere available, theheadof thehouseholdwas contactedprior tothe visit in order to schedule the interview at a convenient time. The in-terviewers made up to three attempts by phone or in person to contact

47E.A. Duke et al. / Ecological Economics 103 (2014) 44–55

each household before determining that the landowner could not be lo-cated. For both study sites, the sample was handled by farm instead ofby individual. This means that if a new household (different from thatlisted as the contact on the master list) had taken over the farm, thathousehold was offered the opportunity to contribute to the study.

2.5. Data Analysis

Statistical analyses of the original quantitative survey data and thequantitized (originally qualitative) survey data were performed usingPASW-SPSS 18.0 (Predictive Analytics SoftWare — Statistical Packagefor the Social Sciences). We set p≤ 0.05 as theminimum cut-off for de-termining statistical significance. Our effect size indices are based onminimal, typical, and substantial measures as defined in Vaske, 2008.

2.5.1. Measurement of Demographic Variations, Preferences, and theLandowner Context

Crosstabulation procedures measuring the strength of associationbetween variables were used to test relationships between dichoto-mous and/or categorical (e.g., gender) demographic measures, alongwith t-tests when the independent variable was dichotomous and thedependent variable was continuous (e.g., income). These analyseswere also used to identify variations in survey responses by locationand associations between specific populations and location. Respondentinterest across PES scenarios was measured in this way as well.

We explored interest and involvement in conservation and sustain-able agriculture efforts, which previous studies report as being asignificant predictor of future conservation activities, such as PES partic-ipation (Jolejole et al., 2009; Kramer and Jenkins, 2009; Sorice andConner, 2010). Specifically, we asked about landowners' current or pre-vious participation in this type of program.

We also asked respondents to identify which general type of organi-zation (federal government, local government, private company, NGO,or no preference) they would prefer to work with as the PES programadministrator.

2.5.2. Factors Affecting Landowner Interest in PESModeling influences on stated interest to participate for our dichot-

omous dependent variables (the forest conservation, reforestation, andagroforestry PES scenarios) were conducted using logistic regression, ageneralized linear model that extends linear regressions to includenon-continuous outcomes using logit link functions and a maximumlikelihood estimator (Aldrich and Nelson, 1984). We used a logistic re-gression model in this way to regress respondents' replies to questionsabout willingness to enroll for each of the three PES scenarios against aseries of likely explanatory variables to uncover factors that were hy-pothesized to influencewillingness to participate. Based upon a system-atic literature review of 56 publications identifying factors influencinglikelihood of landowner participation in PES, we selected the followingexplanatory variables to include in our model to predict respondents'stated interest to participate: household income, farm income, age,past conservation program participation, land tenure (defined asowning and possessing a title to the land), and farm size. Please notethat because our survey targeted landowners, most of the land tenuredifferences were due to possessing or not possessing a title. We aimedto only survey those who owned land; however, our lists did end upcontaining a small number of land renters (2%). For these analyses, weused a simultaneous logistic regression model, meaning that all vari-ables were entered into the model at the same time.

Participation in the reforestation and agroforestry scenarios waslikely possible for all respondents. For the forest conservation scenario,only respondents with existing forested land would be able to partici-pate. Accordingly, interest in participating in the forest conservationscenario was recoded into a dichotomous variable by combining the“Not possible, I do not have forest on my land” responses with thoseindicating an unwillingness to participate. For the purposes of this

analysis, we focused on uncovering factors influencing interest in par-ticipating without breaking down responses by the reason for their in-terest or lack of interest in participating (e.g., socioeconomicconstraint, natural capital constraint).

2.5.3. Exploring Eligibility and Creating A Socioeconomic Status IndexOne-dimensional measures of socioeconomic status (e.g., income,

occupation, education) are generally found to be inadequate forrepresenting a household's livelihood or well-being condition, particu-larly in developing countries (Wagle, 2008). Several studies haveresponded to this issue by creating composite indicators of socioeco-nomic status that integrate multiple factors (Cooke, 2005).

Based upon a literature review, we identified the factors listed inTable 1 as being relevant to constructing a multi-factor index of socio-economic status for our study region. Variables were measured ondifferent scales (e.g. levels 1–3 for cooking fuel and 14 levels for educa-tion). Therefore, we combined variables into an additive index usingstandardized Z-scores to create scores representing each respondent'ssocioeconomic status score. The resulting Z-scores were grouped intothree equal interval categories for low,medium, and high socioeconom-ic status. These groupswere divided using natural breakpoints closest tothirdswithin the Z-score results (0–0.362, 0.363–0.674, 0.675–1). Thesegroupings enabled us to examine how socioeconomic status related torespondents' stated interest and potential eligibility to participate inPES, as described further below.

2.5.4. Eligibility Variable Creation and Eligibility AnalysesWedefined eligibility requirements basedupon three representative

requirements used in other PES programs: (1) owning the land (so notrenting and not necessarily having a land title), (2) having a land title,and (3) owning a minimum of 5 ha (as a proxy for ability to enroll atleast 5 ha). We chose 5 ha as the cut-off, as this resulted in the smallfarm size (b5 ha) in our study region being designated as ineligible.While other cut-offs could have been chosen, the importance of thisanalysis is in providing a first approximation of the relationship be-tween farm size, socioeconomic status, and landowner eligibility forour study region. We also considered the land targeting criteria ofsteeply-sloped lands, given that other programs have included this(e.g., China's Grain-for-Green program; Liu et al., 2008) and soil erosionfrom agricultural expansion on steeply-sloped lands is an importantconcern in our study region.

We used the eligibility requirements to perform analyses that classi-fied landowners as eligible or ineligible for PES participation, examininglandowners as an entire group and then subdivided into the high,medium, and low socioeconomic status groups. The test statistic of chisquared was used for analyzing the relationship between eligibilityand socioeconomic status.

2.5.5. Potential Landowner Participation Curves and Eligibility AnalysisTo examine trade-offs between eligibility requirements and poten-

tial landowner participation (including effects related to socioeconomicstatus), we created what we call potential landowner participationcurves by graphing the percentage of respondents interested and ableto enroll a given number of hectares under increasingly restrictive com-binations of eligibility requirements and the steeply-sloped landcriteria.

3. Results

We received responses from 344 people, resulting in a 92% responserate.

3.1. Demographic Variations, Preferences, and the Landowner Context

Respondents were predominately male (78%) and an average of56 years old (SD = 15). Both the mean and median levels of education

Table 1Indicators used in composite measures of household socioeconomic status based upon the systematic literature review.

Source Indicator

Education Housing Income Possessions Source of water Toilet/septicsystem

Profession Gas orelectric

Electricity # of occupants

Cinner et al. (2008) √ √ √ √ √Cinner et al. (2007) √ √ √ √ √ √ √Cooke (2005) √ √ √McClanhan et al. (2008) √ √ √ √ √ √ √Rai et al. (2008) √ √ √Sahn and Stifel (2000) √ √ √ √ √Sayer et al. (2007) √ √ √ √

48 E.A. Duke et al. / Ecological Economics 103 (2014) 44–55

were 9 years of school (SD = 4). Mean household income was $500–599 per month, with a median of $400–499 per month and a mode of$200–299 per month. Seven percent of households reported incomeless than $100 per month (Table 2).

The vast majority of farms (82%) were 30 ha or smaller, and 45%were less than 5 ha. Of the 98% of respondents who owned their land,81% reported that they held a legal title. Respondents in the Boquetesite reported significantly more secure land tenure and more farmingof steeply-sloped lands with effect size (or the strength of relationshipbetween variables) ranging from minimal to typical (Table 3). Almosttwo times as many respondents in the Boquete site reported farmingon steep slopes (26%) compared to respondents from the Renacimientosite (14%).

Coffee was reported by the greatest number of respondents as thehighest income-producing land use (54%), and 24% of respondentswho reported that they plan to change their current land usewere plan-ning to increase land allocated to coffee production. The primary moti-vation reported for this change was increasing income (41% amongthose reporting a reason). However, 8% were motivated to changetheir current land use for conservation reasons, including responseslike “for conservation, to protect the environment, land is more suitedto newuse”. Among these 25% had plans to reforest and 50% self-report-ed intent to change to increase tourism.

Thirty-one percent of respondents reported participation in conser-vation and sustainable agriculture programs, including 16% who heldland in a private reserve, and 11% who had land enrolled in Conserva-tion International's Conservation Coffee certification program. Respon-dents in the Boquete site were 12% more likely to have enrolled theirland in a program, with a greater average number of hectares in theseprograms (χ2 = 7.82, df = 2, p b 0.05, Cramer's V = 0.151; note:responses were recoded into three categories measured as none,

Table 2Head of household characteristics reported for the entire sample (combined) as well as brocommunities on a given variable are identified by p value, and the systemic measure offers(minimal, typical, or substantial).

Variable Combineda Boquetea Renacim

Age 56 60 52GenderMale 78% 73% 83%Female 22% 27% 17%

Income 6.03c 6.46c 5.56c

EducationNo formal 1% 1% 1%Low (1st–5th) 15% 15% 15%Medium (6th–12th) 61% 54% 70%High (N12th) 23% 30% 15%

a Percents are listed for all variables except for age and income for which the mean is listedb The test statistic is chi squared and the systemic measure is Cramer's V unless indicated

(minimal ≥ 0.1 typical ≥ 0.3, substantial ≥ 0.5).c A household income level of 5 represents $500–599 per month and a level of 6 represents

≤10 ha, N10 ha). For the remaining 69% of respondents who reportednever having participated in these types of programs, themost commonreasons cited were not having access to them, not receiving an invita-tion to participate, and lack of familiarity with the programs. Forty-two percent of the respondents reported having worked with at leastone organization in the region. The most commonly-reported organiza-tions included MIDA (32%), the National Environmental Authority(Autoridad Nacional del Ambiente, ANAM) (10%), and a land-titlingprogram known as the National Land Administration Program(Programa Nacional de Administración de Tierras, PRONAT) (8.5%).

3.2. Framework Questions to Provide Supply-side Information for PESProgram Design

3.2.1. Are Landowners Familiar With the PES Approach?When asked about their familiarity with the PES approach, a major-

ity of respondents (66%) self-reported having “no knowledge”, whileonly 1% indicated having “high” knowledge. For the 34% who self-reported “some knowledge” or higher, 25% cited learning about PESfrom a friend or acquaintance. At least ten percentmentioned obtaininginformation from the following: local organizations, publications, televi-sion, and hearing about Costa Rica's national program through a varietyof sources.

3.2.2. Do Landowners Support PES Development, and If Yes, Which Type ofProgram Focus (e.g., Forest Conservation, Reforestation, Agroforestry)would be most attractive?

In terms of the three PES scenarios presented, 78% of respondentsexpressed interest in the agroforestry scenario, 53% in the forest conser-vation scenario, and 42% in the reforestation scenario (Fig. 2). Respon-dents' ranking of first choice followed the same order: agroforestry

ken out by site (Boquete and Renacimiento). Significant differences between the twoan indication of the effect size or the strength of the relationship between the variables

ientoa Comparing the two communities

Test statisticb p value Systemic measureb

5.04* b0.001 0.265**

5.36 b0.05 −0.126***

2.07* b0.05 0.113***

13.09 b0.01 0.197

.as follows: *t value, **eta (minimal ≥ 0.1 typical ≥ 0.243, substantial ≥ 0.371), ***phi.

$600–$699 per month.

Table 3Key farm characteristics reported for the entire sample (combined) as well as broken out by site (Boquete and Renacimiento). Respondents selected from three farm size options asdetailed blow. Those who owned their land also self-reported information about land tenure (official documentation of ownership). Those who reported “Don't know” may not haveany legal claim to their land. Steep slopewas also self-reported. Significant differences between the two communities on a given variable are identified by p values, and the systemicmea-sure offers an indication of the effect size or the strength of the relationship between the variables (minimal, typical, or substantial).

Comparing the two communities

Variable Combined Boquete Renacimiento Test statistic1 p value Systemic measure2

Farm sizeb5 ha 45% 47% 43% 2.29 0.318 0.0825–30 ha 37% 33% 41%N30 ha 18% 19% 16%

Land tenureTitle 81% 93% 67% 35.89 b0.001 0.328Property rights 17% 5% 28%Don't know 3% 2% 4%

SlopeNo steep slope 80% 74% 86% 7.11 b0.01 −0.144***

Steep slope 20% 26% 14%

1The test statistic is chi squared and 2the systemic measure is Cramer's V unless indicated as follows: ***phi. (minimal ≥ 0.1 typical ≥ 0.3, substantial ≥ 0.5).

49E.A. Duke et al. / Ecological Economics 103 (2014) 44–55

(53%), forest conservation (38%), and reforestation (9%). Respondentsin the Renacimiento site expressed more interest in enrolling theirland in the reforestation scenario than did respondents in the Boquetesite (χ2 = 7.45, df = 1, Phi = .147, p b 0.01).

The mean amount of land respondents expressed interest in enroll-ing in each of the scenarios is as follows: forest conservation (mean =15 ha; SD = 37), reforestation (mean= 13 ha; SD = 36), and agrofor-estry (mean = 11 ha; SD = 28). These means are semi-windsorized(Vaske, 2008) in order to account for skewing by outliers since a few re-spondents expressed interest in enrolling relatively large amounts ofland (e.g., 2–4% of respondents stated an interest in enrolling≥100 ha, and one respondent was willing to enroll 800 ha). Responsesat more than 3 standard deviations from the mean were replaced withthe next valid value.

In terms of contract length, respondents preferred a mean of15 years for the forest conservation and reforestation scenarios, and10 years for the agroforestry scenario. Respondents in the Boquete sitepreferred significantly longer contracts then those in Renacimiento(mean = 15 years instead of 10 years) for the reforestation scenario(t = 2.25, p b 0.05). The two most common responses for preferredorganization to work with as the PES program administrator were “nopreference” or “NGO”, with the following percentages for each scenario:reforestation (46% no preference, 36% NGO), forest conservation (43%

100%

78%80%

53%57%60%

42%42%40%

20%

10%

0%0%Agroforestry

PES ScenarioReforestationForest Conservation

Per

cen

t o

f re

spo

nd

ents

Fig. 2. Percent of respondentswho indicated interest in participating (dark grey) in eachofthe PES scenarios comparedwith the percent of respondentswho ranked each scenario astheir first choice (light grey).

no preference, 39% NGO), and agroforestry (35% no preference, 49%NGO).

3.2.3. Which Landowner Characteristics Affect Stated Interest to Participatein a Future PES Program?

We used a simultaneous logistic regression analysis to identify fac-tors for each of the PES scenarios that predicted respondents' stated in-terest to participate (Fig. 3). All variableswere entered into themodel atthe same time. Two out of the threemodelswere significant overall: for-est conservation (χ2= 84.48, p b 0.001) and reforestation (χ2= 28.01,p b 0.001). The agroforestry model was not significant (χ2 = 6.66, p =0.353), because it was affected by the limited variance in the dependentvariable (stated interest to participate), given that a high percentage(78%) of respondents expressed interest in participating in this scenario.

Among these six independent variables selected for inclusion in themodel, we checked formulticollinearity prior to running the full regres-sion model and found only two variables with a correlation of practicalsignificance at the substantial level (r = 0.366). Separate regressionswere run for these two variables (farm size and past conservation pro-gram participation), and the results were substantially the same in

Fig. 3. The unstandardized logistic regression coefficient B is the log odds of variables thatpredict interest in participating in the PES scenarios (forest conservation, reforestation,and agroforestry). A positive B represents that an increase in the given variable resultsin an increase in the likelihood of respondents being interested in participating; a negativeB represents that an increase in the given variable results in a decrease in the likelihoodofrespondents being interested in participating. The Nagelkerke R2 approximates thevariance predicted by the model (predictive power). Figure 3 Footnote: B = unstandard-ized logistic regression coefficient (the log odds) of variables which predict interest inparticipating in the three PES scenarios.

Fig. 4.Respondents’ stated interest in participating (dark grey; c2 = 1.815, p= 0.404,df = 2, Cramer’s V = 0.073) compared to ability to enroll in the context of eligibilityrequirements for land ownership and land title (light grey; c2 = 33.687, p b 0.001,df = 2, Cramer’s V = 0.314) across low, medium, and high socioeconomic status.

50 E.A. Duke et al. / Ecological Economics 103 (2014) 44–55

terms of significance. For the sake of parsimony, we ultimately chose toinclude all variables in a single regression model.

For the forest conservation scenario, significant positive predictorsof respondents' interest in participating were household income,amount of land enrolled previously in other conservation programs,and farm size, while age was a significant negative predictor (Fig. 3).For the reforestation scenario, farm size and amount of land enrolledpreviously in other conservation programs had a significant positive re-lationship with interest in participating, while land tenure had a signif-icant negative relationship (Fig. 3).

3.2.4. HowWould Decisions About Specific Program Design Elements(e.g., Landowner Eligibility Criteria) Affect the Ability of LandownersFrom Different Socioeconomic Classes to Participate in a PES Scheme?

3.2.4.1. Socioeconomic Status Index Testing. We assessed the reliability(an analysis of the intercorrelation of response patterns) of our con-structed socioeconomic status index using the internal consistencymethod, which resulted in a Cronbach's alpha reliability coefficient of0.61, which exceeds the minimum of 0.60 (Vaske, 2008). TheCronbach's alpha decreased if any items were removed (Table 4).

In the resulting index, those in the low socioeconomic status groupwere less likely to have a land title (mean = 0.67, with 0 = no titleand 1 = title), as compared to the high socioeconomic status group inwhich nearly all respondents had a land title (mean = 0.94; Table 4).Farm size increased as socioeconomic status increased. Respondentswith low and medium socioeconomic status levels were more likelyto have smaller farms (low socioeconomic statusmean= 1.53,mediumsocioeconomic status mean = 1.65; both indicating a farm sizeof ‘b5 ha’) than those in the high socioeconomic status category(mean = 2.03, indicating a farm size of ‘5 to 30 ha’).

3.2.4.2. Socioeconomic Status and Relationship to Respondents' StatedInterest and Potential Eligibility to Participate. We did not find a signifi-cant relationship between socioeconomic status and respondents' stat-ed interest to participate in at least one of the proposed PES scenarios(χ2= 1.815, df=2, Cramer's V= 0.073, p= 0.404). As socioeconomicstatus decreased, the percentage of people interested in participatingremained approximately the same (Fig. 4). Note again, however, thatthis is an expression of respondents' interest, though not necessarilyan ability or actual decision to participate.

We found a significant positive relationship between potential land-owner eligibility and socioeconomic status (χ2 = 33.687, df = 2,Cramer's V = 0.314, p b 0.001; Fig. 4). In other words, the higher arespondent's socioeconomic status, the more likely he or she was tomeet the hypothetical eligibility criteria including having a land title,

Table 4Socioeconomic status index level based on scale attributes and related variables.

ANOVA Socioeconomic status index levela

Variables in index Low MediumEducation (years of schooling) 5.58 8.02Income of household 3.03 ($200–$299) 5.63 ($500–$599)Electricity 2.58b 2.98b

Plumbing 2.31c 2.98c

Cooking fuel 2.36d 2.96d

Mean comparisonRelated variablesLand title 0.67e 0.83e

Size of farm 1.53f 1.65f

a Cell entries are mean scores for respondents.b 1 = ‘No’, no electricity in house; 3 = ‘Yes’, electricity in housec 1 = ‘Latrine’, no indoor plumbing; 3 = ‘Bathroom’, yes indoor plumbingd 1 = ‘Wood’; 2 = ‘Kerosene’ or ‘coal’; 3 = ‘Gas’ or ‘electric’e 0 = no title and 1 = titlef 1=‘ less than 5 ha’; 2 = ‘5 to 30 ha’; 3 = ‘more than 30 ha’

owning the land, and owning a minimum of 5 ha. For example, only26% of those in the low socioeconomic status group were eligible com-pared to 64% of respondents in the high socioeconomic status group.

3.2.4.3. Potential Landowner Participation Curves Under Eligibility andLand Targeting Constraints. Fig. 5 shows the percent of landownerswho stated interest to enroll a given number of hectares in each of thethree PES scenarios. We present the number of hectares all respondentswere interested in enrolling and the number of hectares that just thosein the low socioeconomic status groupwere interested in enrolling. Wefurther looked at how our hypothetical eligibility requirements andtargeting of steeply-sloped lands would affect the number of hectaresenrolled (meaning not just that a landownerwas interested in enrollingland but that the landowner actuallymet the hypothetical eligibility andsteep slope requirements). Dashed vertical lines additionally illustratehow adding two plausible minimum land enrollment requirements of5 ha or 10 ha (as was originally used in Costa Rica; Hope et al., 2005)would affect potential land enrollment. Each graph in Fig. 5 containsthree lines representing the following categories: (1) “all participants,”which shows the percent of all respondents interested in enrolling agiven number of hectares without program constraints; (2) “eligibleparticipants,” which bounds landowner enrollment by requiring that

High F p value eta12.19 174.13 b0.001 0.73010.18 ($900–$999) 159.00 b0.001 0.7263.00b 24.03 b0.001 0.3673.00c 49.18 b0.001 0.4953.00d 41.88 b0.001 0.462

0.94e 14.11 b0.001 0.2922.03f 13.87 b0.001 0.287

Fig. 5. Potential landowner participation curves showing the maximum number of ha that landowners are interested in enrolling in each PES scenario, in the context of no constraints aswell as constraints related to eligibility requirements (land ownership and land title) and steep-slope targeting. Curves are shown for all respondents (panels A, C, and E), and just forrespondents in the lowest socioeconomic group (panels B, D, and F). Each panel contains three lines: “All Participants” shows the number of ha respondents reported that they wouldenroll; “Eligible Participants” shows the reported number of hectares that respondents would enroll who meet eligibility requirements; “Eligible Participants with Slope” shows thereported number of hectares by respondents who would enroll and meet both eligibility requirements and steep-slope targeting criteria. The vertical dashed lines at five and 10 hamake it easier to see how many people would be included (or excluded) if program requirements included these representative minimum land enrollment sizes.

51E.A. Duke et al. / Ecological Economics 103 (2014) 44–55

they meet land ownership and tenure criteria; and (3) “eligible partici-pants with slope,” which further bounds landowner enrollment by re-quiring that they also meet the steep slope criterion.

This analysis highlighted three main results. First, a notable drop-offin potential landowner participation can be seen between 1 and 5 haand between 5 and 10 ha across all curves. The many landowners whowere interested in enrolling only small parcels of land become ineligibleeven with relatively small land enrollment requirements. Second, landtenure contributes most of the variability between the “all participants”

line and the “eligible participants” line. The eligible participants line fallsanywhere from 23 to 28% below the all participants line for the all-inclusive graphs and from 36 to 50% below for the low socioeconomicstatus graphs, meaning that requiring land ownership and title wouldmake programs inaccessible to these landowners. Third, the restrictiveimpacts of potential program eligibility requirements are greater forlandowners in the low socioeconomic status category than for otherlandowners. For example, minimum land enrollment requirementscould greatly impact program participation among poorer households,

52 E.A. Duke et al. / Ecological Economics 103 (2014) 44–55

with only 7% of low socioeconomic status respondents expressing inter-est in enrolling 10 ha ormore in the reforestation scenario compared to21% of respondents overall. Also, requiring land title and ownershipdrops those interested and able to enroll 1 ha in the reforestation pro-gram by 27% overall and by 50% for those in the low socioeconomic sta-tus group. The results were similar across the forest conservation andagroforestry scenarios.

4. Discussion

Multiple factors must be taken into account when considering howto develop a PES program so as not to unintentionally excludelandowners who would contribute to the program achieving its envi-ronmental, and social where applicable, objectives (Engel et al., 2008).At the initial design stage of a PES program, the implications of variousprogram design factors on potential landowner participation shouldbe analyzed to determine themarginal impact of each factor andwheth-er it would affect a programbeing able to achieve its objectives. Tomakethis possible, program objectives must be clearly defined at theoutset along with rules and requirements to reinforce these objectives(Rodríguez et al., 2011). How eligibility requirements (e.g., land tenure,minimum land enrollment) and other design parameters are set can bea powerful bottleneck affecting landowner involvement, both in termsof their interest and their ability to participate (Pagiola et al., 2007;Wunder, 2008). We developed a framework, informed by the literatureand experience from existing PES programs, to investigate potentiallandowner interest in participating in a future PES scheme. We appliedthis framework to our study region inwestern Panama to provide stake-holderswith information about howprogramdesign factors could influ-ence supply-side participation.

4.1. Demographic Variations, Preferences, and the Landowner Context

One important component of programdesign is selectingwhichpro-gram modalities to offer. Our analysis found that agroforestry was themost preferred scenario, whichmay be because it best fits with respon-dents' current agricultural operations. Forest conservation was the sec-ond most popular option, possibly due to the lack of upfront costs orneed to make any major changes in land use. Reforestation was theleast preferred option, possibly because it would require the greatestand most expensive changes to current operations.

While our analysis did not showwidespreaddifferences in landown-er responses across study sites, we did find some differences that wouldlikely lead to differential impacts on landowner participation in a futurePES program that encompasses both sites. First, landowners in theRenacimiento site were 26% less likely to possess a legal land title thanin the Boquete site, which would affect participation if land title wereincluded as an eligibility requirement. Second, landowners in theRenacimiento site reported 12% fewer steeply-sloped land holdingsthan in the Boquete site, which would affect prioritization for a PEScontract if slope targeting were included in a future program. Third,we found an association between farm location and conservationengagement, with greater landowner involvement in conservation-related activities reported for the Boquete site. This could suggest ahigher rate of future participation in PES given the influence of pastbehavior on future behavior, though we also found that more respon-dents from the Renacimiento site expressed interest specifically in en-rolling in the reforestation scenario compared to those from theBoquete site. Perhaps this is due to the fact that less conservation actionsuch as reforestation has already taken place in Renacimiento. Interestin participating in the other two scenarios was about equal betweenthe two sites. Overall, the level of conservation engagement in bothsitesmay be overly represented in our study population due to our sam-pling procedure which selected names from lists of program partici-pants from MIDA, members of coffee growing cooperatives, and other

such lists of farmers working with these groups and participating intheir programs.

4.2. Predicting Landowner Interest in Participating

Several studies have found that the environmental attitudes and af-filiations, as opposed to economic factors,were the strongestmotivationfor landowners to participate in PES and other conservation programson private lands (Jolejole et al., 2009; Kosoy et al., 2008; Lichtenbergand Zimmerman, 1999). Our results provide support for a growingbody of research showing that past behaviors and some socio-demographic variables can successfully predict or explain landowners'interest in participating in a PES program. Farm size is generally foundto be positively correlated with actual PES participation (e.g., Zbindenand Lee, 2005), and our research further supports this finding in thecontext of behavioral intention, or expressed interest in program enroll-ment. It is suggested in the literature that demographics are not the bestpredictors of behaviors and related constructs (Ajzen and Fishbein,1980; Dietz et al., 1998). Therefore, it is not surprising that they wereless significant predictors of interest in enrolling across the three PESscenarios for our study region. Instead, psychological variables (e.g. cog-nitions including values, attitudes, and norms) are often found to form amore meaningful basis for behavior (Sorice and Conner, 2010). Anotherkey indicator of future behavior is past behavior (Eagly and Chaiken,1993). Data from Kramer and Jenkins' (2009) survey of North Carolinafarmers showed that landowners who were already participating inconservation programs were predisposed to willingness to participatein a PES program. Similarly,while farm size and land tenurewere impor-tant influences, past participation in conservation-related activities forour respondents explained the most variation in interest to enroll inthe forest conservation scenario and was also a significant contributorto the reforestationmodel. Kramer and Jenkins' explained a similar find-ing in their research a follows: “This may be due to their familiarity andsatisfactionwith programs they are enrolled in. This may also reflect theinfluence of a stronger conservation ethic that predisposes some indi-viduals to consider enrolment in such programs” (Kramer and Jenkins,2009, p.24). A landowner's environmental ethic and other non-financial factors can play a key role in conservation decisions, becauseproducersmaygain direct personal satisfaction from the improved envi-ronmental quality (Chouinard et al., 2008).

4.3. Eligibility Trade-offs

A key component of our studywas to investigate howprogram eligi-bility criteria would affect landowners' ability to participate across dif-ferent levels of socioeconomic status, recognizing that livelihood andpoverty alleviation objectives are becoming an increasingly importantconsideration for PES programs, particularly in developing countries(Li et al., 2011). While we found that households across the socioeco-nomic status spectrum shared a common interest in participating inPES (no statistical difference by socioeconomic status level), the intro-duction of certain eligibility requirements would result in significantprogram enrollment restrictions for lower socioeconomic classes. Thishighlights the importance of eligibility criteria as a place for programdesigners to focus attention to be aware of possible intended or unin-tended impacts that may result for poorer households.

Setting minimum land enrollment requirements is one illustrationof how eligibility requirements can impact the poor and, consequently,the social objectives of PES. If a programwere to set aminimum land en-rollment requirement, thenwhere this number is set could have amajorimpact on the eligibility of a substantial fraction of landowners. Giventhat almost half of our respondents owned 5 ha or less, ensuring thatPES designers are aware of the number of small farms in the area willbe important to evaluating the trade-offs (ecological, economic, andsocial) in setting a minimum enrolled land size requirement. Settingminimum land enrollment requirements too high can also have

53E.A. Duke et al. / Ecological Economics 103 (2014) 44–55

consequences for a PES program meeting its environmental objectives.The diverse agroecological systems that are often favorable to ecosys-tem service production are typically managed by smallholders and in-digenous farmers. For example, studies using spatial analyses indeveloping countries indicate a high level of overlap between thelands that poor people inhabit and the lands that generate key ecosys-tem services (Nelson and Chomitz, 2004; Sunderlin et al., 2007). Policiesthat favor smallholders, promote diverse farming landscapes, and sup-port dissemination of traditional practices and agroecological knowl-edge are generally beneficial for biodiversity conservation (Castilloand Toledo, 2000). Accordingly, excluding smaller farms could proveto be problematic for achieving the environmental objectives of PESprograms.

Our potential landowner participation curves provided a more in-depth investigation of the relationship between landowner interest inparticipating in a PES program, potential eligibility and other program-matic targets, and socioeconomic status (Fig. 5). The goal of theseanalyses was to guide program designers toward a first-level of under-standing about where a “sweet spot” could be achieved between com-peting design components to produce a favored outcome betweenextremes.

Low socioeconomic status respondents were most severely impact-ed by land title and ownership requirements. Inequity in probable pro-gram access was demonstrated by the greater difference between allparticipants and eligible participants for the low socioeconomic statuscurves (Fig. 5b,d,f) when compared to the curves without any socioeco-nomic status restriction (Fig. 5a,c,e). We included title because almostall PES programs either currently require participants to have a title totheir land or did so at one time but have since relaxed the requirement.This variablemay have a greater impact on participation by landownersin other parts of Panama compared to our study region, which has al-ready benefited from the PRONAT titling project funded by the GlobalEnvironmental Facility and the World Bank.

Land ownership in our study contributed little to potential programeligibility differences, since 98% of respondents reported owning theirland. We aimed to only survey those who owned land; however, ourlists did end up containing a small number of land renters. We includedownership in the eligibility analysis because it can be difficult to designprograms that include renters, and including squatters may be undesir-able (e.g., it may encourage encroachment into protected areas;Wunder, 2005). However, special arrangements could be designedwith renters in mind if program decision-makers feel that includingrenters is important either because a large percentage of land in a spe-cific area ismanaged by renters or land of especially high ecosystem ser-vice value is managed by renters.

Including a steep slope targeting requirement in our analysis dra-matically reduced eligibility across all socioeconomic status levels. Asnoted earlier, we considered a steeply-sloped land requirement becausesuch lands are more negatively impacted by agricultural intensification(e.g., soil erosion), and cumulative effects can be especially negativeacross a watershed (Nelson and Chomitz, 2004). Indeed, these are chal-lenges facing communities in our study region, where farming onsteeply-sloped lands has expanded recently and is believed to be caus-ing increased problems with soil erosion and landslides. Interestingly,there is some evidence in the literature that poorer farmers tend tohave more marginal lands including steeply-sloped agricultural lands,so this criterion in some places could actually facilitate both pro-poorand conservation targeting (Okwi et al., 2007).

Adding the steep slope targeting factor reduced potential participa-tion by as much as 86% overall. Slope was self-reported and only pro-vides a first approximation of the actual land attribute, but there issubstantial slope variability in the region, so it could be a key programfeature. Offering higher payments for steeply-sloped land enrollmentsmight be a way to target a program to address erosion and landsliderisk without excluding other potential participants. Other targeting ef-forts might offer higher payments for land directly bordering protected

areas or land within certain watersheds. However, as with all suchchoices, trade-offs and possible perverse incentives must be carefullyconsidered to determine the most appropriate PES structure.

4.4. Study Limitations

Limitations to this study should be kept in mind when interpretingour findings. First, it is important to note that the three hypotheticalPES scenarios explored through this project address additionality tovarying degrees. As a subsidy approach, PES is particularly vulnerableto the potential pitfalls presented by a lack of additionality or makingpayments for land uses that have already been adopted or would havebeen adopted anyway, thus reducing funds available to induce land-use change elsewhere (Engel et al., 2008). While the reforestation sce-nario incentivizes the generation of new forest cover, the forest conser-vation scenario does not, and the agroforestry scenario may or may not.However, the question of whether or not existing forest in the area isthreatened must be considered as well, which relates to the discussionof credits for Reducing Emissions from Deforestation and Forest Degra-dation (REDD). Notably, deforestation is a concern in our study region.Oestreicher et al. (2009) report that, from 1992 to 2000, PILA had thesecond highest level of loss in mature forest cover out of the tenPanamanian protected areas examined.

Second, as mentioned previously, the number of hectares that re-spondents expressed interest in enrolling provides only an initial indi-cation of the supply-side land potential. As program design advancesand factors such as payment levels and eligibility criteria are defined,landowners will have more refined information from which to makeprogram participation and land enrollment decisions. Since paymentamounts were not provided in our exploratory survey, and thus netbenefits were unknown, the relative order of the options selected by re-spondents might bemostly an illustration of their preferences based ontheir private costs. In particular from a financial perspective, land-owners' willingness to participate in a future PES program will bestrongly influenced by their opportunity costs of partial or full foregoneagricultural activities. As opportunity costs go up (down), landownerswill be less (more) willing to participate, all else equal.

Many factors, including conservation value, buyer willingness topay, landowner willingness to accept, andmonitoring costs together in-fluence the determination of PES payment amounts. Monitoring costsmay be of particular concern to potential participants if they might beresponsible for covering some or all of these costs. However, our resultsprovide a reasonable starting point from which to understand the con-texts in which landowners may need greater incentives.

5. Conclusions: Implications for PES Program Design

PES program design is an integral factor influencing the degree towhich a broad spectrumof landowners are able to receive PES contracts,and therefore whether they and the communities they live in have theopportunity to benefit from a new program (though we note that par-ticipation does not necessarily result in a net benefit (Bennett, 2008;Pagiola et al., 2005). Appropriate design requires combining site-specific information with lessons learned from existing PES programs.Within the context of lessons learned, we interpret the results of ourlandowner survey to provide specific program design and policy-relevant insights for the future development of a PES program in west-ern Panama. By proactively investigating these issues, we can helpensure that PES programs that specify social objectives, in addition toenvironmental objectives, are designed to increase the likelihood ofbroad landowner eligibility and the delivery of positive livelihood ben-efits to participating landowners. Of course, socioeconomic informationmust be considered in tandemwith ecological data to ensure that a pro-gram can achieve its environmental objectives, or does so withminimalefficiency loss if the specification of social objectives leads to an efficien-cy trade-off.

54 E.A. Duke et al. / Ecological Economics 103 (2014) 44–55

For the study region, our results contribute to an important first stepin supply-side information gathering, consortium building, and collabo-ration, which could ultimately lead to the development of a pilot PESprogram in the study region. The information gathered from our find-ings could help to inform ongoing policy discussions to propose a lawin Panama with the intent of developing PES capacity at the nationallevel, while supporting local and regional efforts. Based upon our re-sults, we provide the following recommendations for how to advancePES development in this region in a way that takes landowner needsand preferences into account, while also enabling access for low socio-economic status participants, which is a plausible program objectivein this region: (1) offer contract options that allow enrollment ofworking land uses, such as sustainable agriculture and livestock man-agement; (2) base someprogramoptions onmodifying existing agricul-tural production practices to meet specific guidelines for shade growncoffee or encourage ongoing forest conservation in areaswhere defores-tation pressure is high instead of focusing only on substantial land-usechanges such as full reforestation (reforestation was the least favoredof the three PES scenarios); (3) engage a trusted non-governmental or-ganization to negotiate enrollment, monitor compliance, and managepayments; and (4) carefully weigh participation/eligibility trade-offsprior to setting criteria such as a minimum land enrollment size, re-quirement for land title, steep slope targeting, or other factors. Becausemost farms and ranches in this region remain small, higher land enroll-ment requirements result in a rapid decline in eligible participants asshown in Fig. 5. As a first approximation, these results reveal important“cut-off points” that can provide valuable information to guide programdesign decisions to increase the likelihood of achieving desired levelsand types (e.g., across the socioeconomic status spectrum) of land-owner participation.

The next important steps for this research program include identify-ing key ecosystem service beneficiaries and assessing their potential toact as PES buyers; spatial ecological mapping for targeting purposes; aninstitutional capacity analysis; and an economic valuation analysis. Ourteam has already begun to investigate institutional capacity for PES pro-gram implementation in the region and to identify key user groups. Aneconomic analysis that builds upon the results of our study is a neces-sary step to further refineour preliminary predictions of landowner par-ticipation, as well as to understand the value of ecosystem services tobeneficiaries. This work should take place once further narrowing ofthe scope of a future PES program is agreed upon by stakeholders,such as which ecosystem services and/or land use types are beingtargeted. Through our research, we aimed to contribute a valuablefirst step towards this process in western Panama, and to provide abroader framework for investigating landowners' interest to participatein PES in regions where program development is at the initial stage ofexploration.

Acknowledgments

We thank Colorado State University's (CSU) Center for ProtectedArea Management and Training, FUNDICCEP, FUNDAVISAP, Ministeriode Desarrollo Agropecuario de Panamá, local farming cooperatives,Banco de Desarrollo Agropecuario, and The Nature Conservancy for re-search support. We thank CSU's Center for Collaborative Conservation,Green Dream Team, and Environmental Governance Working Groupfor funding. We benefited from insights from R. Reid, P. Taylor,A. Dietsch, and an anonymous reviewer who provided constructivecomments on an earlier version of this manuscript. We also thankC. Huayhuaca and L. Tyler for assistance with Spanish translation, andA. Miller for assistance with data entry and management.

Appendix A. Supplementary data

Supplementary data to this article can be found online at http://dx.doi.org/10.1016/j.ecolecon.2014.04.013.

References

Ajzen, I., Fishbein, M., 1980. Understanding Attitudes and Predicting Social Behavior.Prentice-Hall, Englewood Cliffs, NJ.

Aldrich, J.H., Nelson, F.D., 1984. Linear Probability, Logit, and Probit Models. Sage, BeverleyHills, CA.

Altman, D.G., 1995. Sustaining interventions in community systems: on the relationshipbetween researchers and communities. Health Psychol. 14 (6), 526.

Bennett, M.T., 2008. China's sloping land conversion program: institutional innovation orbusiness as usual? Ecol. Econ. 65 (4), 699–711.

Castillo, A., Toledo, V.M., 2000. Applying ecology in the third world: the case of Mexico.Bioscience 50 (1), 66.

Chouinard, H.H., Paterson, T., Wandschneider, P.R., Ohler, A.M., 2008. Will farmers tradeprofits for stewardship? Heterogeneous motivations for farm practice selection.Land Econ. 84 (1), 66–82.

Cinner, J., Daw, T., McClanahan, T., 2008. Socioeconomic factors that affect artisanal fish-ers’ readiness to exit a declining fishery. Cons. Bio. 23 (1), 124–130.

Cinner, J., Sutton, S., Bond, T., 2007. Socioeconomic thresholds that affect use of customaryfisheries management tools. Cons. Bio. 21, 1603–1611.

Clark, T.W., Ashton, M.S., Dixon, L., 2006. Innovation and appraisal of sustainability effortsin La Amistad, Bocas del Toro, Panama and Talamanca, Costa Rica Region — a synthe-sis. J. Sustain. For. 22 (1), 183.

Cooke, M., 2005. The first nations community well-being index (CWB): a conceptualreview. Strategic Research and Analysis DirectorateIndian and Northern Affairs,Canada.

de Koning, F., Aguinaga, M., Bravo, M., Chiu, M., Lascano, M., Lozada, T., Suarez, L., 2011.Bridging the gap between forest conservation and poverty alleviation: theEcuadorian Socio Bosque program. Environ. Sci. Pol. 14 (5), 531–542.

Dietz, T., Stern, P.C., Guagnano, G.A., 1998. Social structural and social psychological basesof environmental concern. Environ. Behav. 30 (4), 450–471.

Eagly, A.H., Chaiken, S., 1993. The Psychology of Attitudes. Wadsworth Publishing Co Inc,Fort Worth, TX.

Engel, S., Pagiola, S., Wunder, S., 2008. Designing payments for environmental services intheory and practice: an overview of the issues. Ecol. Econ. 65 (4), 663–674.

Hope, R.A., Porras, I.T., Miranda, M., 2005. Can payments for environmental services con-tribute to poverty reduction? A Livelihoods Analysis From Arenal, Costa Rica. Unpub-lished project report (DFID FRP R8171)Newcastle University, UK.

Jolejole, M.C., Swinton, S., Lupi, F., 2009. Incentives to supply enhanced ecosystem servicesfrom cropland. Text presented at the 2009 Annual meeting of the Agircultural andApplied Economics AssociationMilwaukee, Wisconsin (July).

Kosoy, N., Corbera, E., Brown, K., 2008. Participation in payments for ecosystem services:case studies from the Lacandon rainforest, Mexico. Geoforum 39 (6), 2073–2083.

Kramer, R., Jenkins, A., 2009. Ecosystem Services, Markets, and Red Wolf Habitat: ResultsFrom a Farm Operator Survey. Conservation Economics White Paper. Defenders ofWildlife.

Li, J., Feldman, M., Li, S., Daily, G., 2011. Rural household income and inequality under theSloping Land Conversion Program in western China. Proceedings of the National Acad-emy of Sciences, Early Edition pp. 1–6. http://dx.doi.org/10.1073/pnas.1101018108.

Lichtenberg, E., Zimmerman, R., 1999. Farmers' willingness to pay for groundwater pro-tection. Water Resour. Res. 35 (3), 833–841.

Lichtenfeld, M., 2007. Improving the supply of carbon sequestration services in Panama. J.Sustain. For. 25 (1 & 2), 43–73.

Liu, J., Li, S., Ouyang, Z., Tam, C., Chen, X., 2008. Ecological and socioeconomic effects of China'spolicies for ecosystem services. Proc. Natl. Acad. Sci. 105 (28), 9477.

McClanahan, T.R., Cinner, J.E., Maina, J., Graham, N.A.J., Daw, T.M., Stead, S.M., Wamukota,A., et al., 2008. Conservation action in a changing climate. Conserv. Lett. 1, 53–59.

Milder, J.C., Scherr, S.J., Bracer, C., 2010. Trends and future potential of payment for eco-system services to alleviate rural poverty in developing countries. Ecol. Soc. 15 (2), 4.

Muñoz-Piña, C., Guevara, A., Torres, J.M., Braña, J., 2008. Paying for the hydrological ser-vices of Mexico's forests: analysis, negotiations and results. Ecol. Econ. 65 (4),725–736.

Nelson, A., Chomitz,, K.M., 2004. The Forest-Hydrology-Poverty Nexus in Central America:An Heuristic Analysis World Bank Policy Research Working Paper, 3430.

Oestreicher, J.S., Benessaiah, K., Ruiz-Jaen, M.C., Sloan, S., Turner, K., Pelletier, J., Guay, B.,et al., 2009. Avoiding deforestation in Panamanian protected areas: an analysis ofprotection effectiveness and implications for reducing emissions from deforestationand forest degradation. Glob. Environ. Chang. 19 (2), 279–291.

Okwi, P.O., Ndeng'e, G., Kristjanson, P., Arunga, M., Notenbaert, A., Omolo, A., Henninger, N.,et al., 2007. Spatial determinants of poverty in rural Kenya. Proc. Natl. Acad. Sci. U. S. A.104 (43), 16769–16774.

Pagiola, S., Arcenas, A., Platais, G., 2005. Can payments for environmental services help re-duce poverty? An exploration of the issues and the evidence to date from LatinAmerica. World Dev. 33 (2), 237–253.

Pagiola, S., Rios, A., Arcenas, A., 2007. Poor household participation in payments for envi-ronmental services: lessons from the Silvopastoral Project in Quindío, Colombia.September 6 MPRA Paper (Retrieved April 18, 2009, from http://mpra.ub.uni-muenchen.de/4794/).

Rai, A., Sharma, S., Sahoo, P., Malhotra, P., 2008. Development of livelihood index for dif-ferent agro-climatic zones of India. Agr. Econ. Res. 21, 173–182.

Reid, R., Nkedianye, D., Said, M., Kaelo, D., Neselle, M., Makui, O., Onetu, L., et al., 2009.Evolution of models to support community and policy action with science: balancingpastoral livelihoods and wildlife conservation in savannas of East Africa. Proc. Natl.Acad. Sci. (www.pnas.org/).

Rodríguez, L.C., Pascual, U., Muradian, R., Pazmino, N., Whitten, S., 2011. Towards a unifiedscheme for environmental and social projection: learning from PES and CCT experi-ences in developing countries. Ecol. Econ. 70, 2163–2174.

55E.A. Duke et al. / Ecological Economics 103 (2014) 44–55

Sayer, J., Campbell, B., Petheram, L., Aldrich, M., Perez, M.R., Endamana, D., Zacharie-L, N.D.,et al., 2007. Assessing environment and development outcomes in conservation land-scapes. Biodivers. Conserv. 16 (9), 2677–2694.

Sorice, M.G., Conner, J.R., 2010. Predicting private landowner intentions to enroll in an incen-tive program to protect endangered species. Hum. Dimens. Wildl. Int. J. 15 (2), 77–89.

Sunderlin, W.D., Dewi, S., Puntodewo, A., 2007. Poverty and forests: multi-country analy-sis of spatial association and proposed policy solutions. CIFOR Occasional PaperNumber 47Center for International Forestry Research, Bogor, Indonesia ([online]URL: http://www.cifor.cgiar.org/Knowledge/Publications/Detail?pid=2294).

Tacconi, L., 2012. Redefining payments for environmental services. Ecol. Econ. 73, 29–36.The Nature Conservancy, 2007. Amistad Biosphere Reserve, Costa Rica and Panama: Parks

in Peril End-of-Project Report. The Nature Conservancy, Arlington, Virginia, USA.Turpie, J.K., Marais, C., Blignaut, J.N., 2008. The working for water programme: evolution

of a payment for ecosystem services mechanism that addresses both poverty andecosystem service delivery in South Africa. Ecol. Econ. 65, 788–798.

UNESCO, 2008. World Heritage Centre — 32nd session of the Committee: Reactive mon-itoring mission to the Talamanca Range La Amistad Reserves/La Amistad NationalPark — PILA, Costa Rica and Panama. About World Heritage (Retrieved May 10,2009, from http://whc.unesco.org/en/sessions/32COM/documents/).

Vaske, J., 2008. Survey Research and Analysis: Applications in Parks, Recreation andHuman Dimensions. Venture Publishing Inc., State College, Pennsylvania.

Wagle, U., 2008. Multidimensional Poverty Measurement: Concepts and Applications.Springer, New York, NY.

Wunder, S., 2005. Payments for Environmental Services: Some Nuts and Bolts. Center forInternational Forestry Research, Bogor, Indonesia.

Wunder, S., 2008. Payments for environmental services and the poor: concepts and pre-liminary evidence. Environ. Dev. Econ. 13 (03), 279–297.

Zbinden, S., Lee, D., 2005. Paying for environmental services: an analysis of participationin Costa Rica's PSA program. World Dev. 33 (2), 255–272.