author's personal copy...author's personal copy and multi-strata agroforests. shade coffee...

This article appeared in a journal published by Elsevier. The attachedcopy is furnished to the author for internal non-commercial researchand education use, including for instruction at the authors institution

and sharing with colleagues.

Other uses, including reproduction and distribution, or selling orlicensing copies, or posting to personal, institutional or third party

websites are prohibited.

In most cases authors are permitted to post their version of thearticle (e.g. in Word or Tex form) to their personal website orinstitutional repository. Authors requiring further information

regarding Elsevier’s archiving and manuscript policies areencouraged to visit:

http://www.elsevier.com/copyright

http://www.elsevier.com/copyright

Author's personal copy

ANALYSIS

A conjoint analysis of farmer preferences for communityforestry contracts in the Sumber Jaya Watershed, Indonesia

Bustanul Arifina, Brent M. Swallowb,⁎, S. Suyantoc, Richard D. Coeb

aUniversity of Lampung, Bandar Lampung, IndonesiabWorld Agroforestry Centre, Nairobi, KenyacWorld Agroforestry Centre, Bogor, Indonesia

A R T I C L E D A T A A B S T R A C T

Article history:Received 24 April 2008Received in revised form2 December 2008Accepted 4 December 2008Available online 20 January 2009

A wide range of policy instruments have been devised and applied to support the goals ofsustainable forestry management. Community forestry programs can contain elements ofseveral of those instruments. This paper considers the design of community forestrycontracts in the Sumber Jaya area of Indonesia where community forestry contracts areagreements between the Forestry Department and community groups that provide groupmembers with time-bound leasehold rights to protection forests, on the condition thatfarmers abide by specified land-use restrictions and pay any required fees. Farmers perceivethat the contracts represent a bundle of restrictions and inducements, some of which areexplicitly stated in the contract and others that are implied by the contract. Conjointanalysis was used to quantify farmers' tradeoffs among the explicit and implicit attributesof the contracts. The results of logit and ordered logit models show that farmers are mostconcerned about the length of the contract, and relatively unconcerned about requirementson tree density and species composition. An implicit attribute, greater access to forestry andagroforestry extension, emerges as an important implicit attribute. The results imply thatfarmers in this part of Indonesia would be willing to abide by fairly strict limitations on landuse, provided that they can be assured of long-term rights to the planted trees.

© 2009 Elsevier B.V. All rights reserved.

Keywords:Community forestrySocial forestryLand tenureIndonesiaConjoint analysisOrdered logitLogitContract design

1. Introduction

Tropical forested landscapes are typicalmulti-value resources.Portfolios of private, collective and public goods and servicesare produced by those landscapes: goods and services ofinterest to a range of stakeholders who live in the landscapesand external stakeholders who benefit directly or indirectlyfrom the goods and services. Research conducted across thehumid tropics shows that there are almost always tradeoffsbetween the extremes of intense production of tree or non-treeproducts and conservation of old-growth forests. Between the

extremes there is considerable scope for achieving outcomesof mutual benefit for different stakeholder groups. Agrofor-estry systems, in which farmers integrate trees into farmingsystems, have emerged as a promising alternative (Tomichet al., 2005).

An array of policy instruments has been devised forpromoting the management of tropical forest landscapes.These include “traditional” instruments such as state forestownership, timber harvest concessions, public tree plantingprograms and restrictions on commercial trade of forestproducts. They also include a variety of community and co-

E C O L O G I C A L E C O N O M I C S 6 8 ( 2 0 0 9 ) 2 0 4 0 – 2 0 5 0

⁎ Corresponding author. PO Box 30677, 00100 United Nations Avenue, Nairobi, Kenya. Tel.: +254 20 7224000; fax: +254 20 7224001.E-mail address: [email protected] (B.M. Swallow).

0921-8009/$ – see front matter © 2009 Elsevier B.V. All rights reserved.doi:10.1016/j.ecolecon.2008.12.007

ava i l ab l e a t www.sc i enced i rec t . com

www.e l sev i e r. com/ l oca te /eco l econ


management tenure arrangements: decentralization to localgovernment authorities, agroforestry extension, conservationtrust funds, product certification, and conditional paymentsfor water quality preservation or carbon sequestration (Cub-bage et al., 2007; Swallow et al., 2007).

In any particular context, governments and other organi-zations interested in the management of forest landscapesneed to make informed choices about which instruments toapply. One approach is to design the policy instrument tobalance the interests and perceptions of key stakeholdergroups. In this paper we show that conjoint analysis can be apowerful tool for understanding theway that farmers perceivethe attributes of community forestry contracts in Indonesia.Results from the study can inform the processes of negotiationbetween farmers and other key stakeholders in communityforestry. Depending upon the context, a similar study couldalso be done with representatives of other stakeholders orother communities in order to identify areas of similarity anddifference in preferences (e.g. Tsalikis et al., 2002).

Forest landscape management in Indonesia is an instruc-tive case of multiple values, multiple interests, and multiplepolicy instruments. After many decades of top-down regula-tion and central government control of the vast forest estate,Indonesia's forest policy began in 1998 to slowly evolvetoward decentralization and stronger rights for communitiesand indigenous peoples. Although the Forest Departmentcontinues to exert overall control over a large portion ofthe Indonesian land mass, some progress has been madein opening space for negotiation between communitiesand the government (Fay et al., 2000). Community forestrycontracts (HKm) provide one possible avenue for negotiatedsettlements.

This paper addresses the challenge of understandingfarmer perceptions of community forestry contracts inIndonesia. The paper focuses on a case study in the SumberJayawatershed in Lampung Province on the island of Sumatra.The contract can be conceived as a bundle of policy instru-ments, some of which are explicitly stated in the contract andothers implicitly implied by the contract. Conjoint analysis isused to assess the way that farmers value different policyinstruments — expressed as different attributes of a possiblecontract. The results, many of which were unexpected, haveimportant implications for community forestry contracts inIndonesia, and potentially other community forestry contextsin the developing world.

2. Background

2.1. Forest policy in Indonesia

Indonesia drew upon the Dutch colonial experience in landconsolidation when drafting the 1967 Forestry Law. In thatview, forests are viewed as strategic assets that a governmentshould protect and manage in order to generate income andsecure environmental services of general public benefit.Government agencies generate income from the sale oftimber concessions without ceding ownership rights. Basedon the Forest Act of 1967, the Ministry of Forests assertedcontrol of more than 70% of the total area of the country,

despite the fact that those areas were home for about90 million people (Fay and Michon, 2005; Colchester et al.,2005). This law, however, was not well implemented due tothe strength of local interest.

International development agencies began to press forreforms of the Forest Law in the late 1980s and modest gainswere achieved in the mid-1990s. The fall of President Suhartoin 1998 ushered in the era of Reformasi and the hope for muchgreater recognition of the rights of indigenous people and localcommunities. A new Forestry Act was passed in 1999 thatincluded provisions for areas of forest domain to be desig-nated as Special Purpose Areas and customary forests. The1999 Forestry Law reflected a considerable shift of power backto local interests. In 2001, the Minister of Forestry issueddecree 31/2001, which provided operational guidelines forcommunity forestry permits, Hutan Kemasyarakatan (HKm).Farmer groups interested in securing HKm permits arerequired to form recognized farmers' organizations and tofollow management guidelines that local forestry officialsapprove as being protective of the watershed functions of thelandscape. Five-year initial permits can be extended to amaximum of 35 years. As of 2005, the area under HKm permitscomprised 2100 km2, less than 0.2% of the country's forestestate (Colchester et al., 2005).

2.2. Study site

The conjoint analysis study was conducted in the Way Besaysub-watershed in the sub-district of Sumber Jaya in WestLampung, Indonesia. As of the mid-1990s the Sumber Jayaarea was known for its intense land-use conflicts.

The entire area was classified as state forest in 1922 duringthe Dutch colonial period. This classification was upheld bythe independent Government of Indonesia. Between 1980 and1990 the Government further demarcated the area into a smallarea of private farm land, conservation forests for biodiversityconservation, and large areas of protection forests forwatershed protection (Verbist et al., 2005). Although much ofthe protection forests had been deforested and converted tocoffee farms as early as the 1950s, government officialsperiodically evicted thousands of farmers who had settled inthe area, with large evictions in 1990–1, 1995 and 1996.Government agencies were concerned about the impacts ofcoffee farms on the quality and quantity of water in thetributaries of theWay Besay River. A run-of-river hydro-powerplant was built on the Way Besay River to increase energysupplies to southern Sumatra and surrounding areas. As of themid-1990s, there had been no serious attempt to resolve theland-use conflict through dialog or negotiation. The WorldAgroforestry Centre and several local non-governmentalorganizations became active in the area in 1998. Sumber Jayasub-district covers an area of 356 km2 and is home to about50,000 people (Verbist et al., 2005).

Forest cover and agroforestry are both very dynamic in theSumber Jaya area. The amount of forest declined from about60% in 1970 to 32% in 1978 and to 10% in 1990 and 2000. Overthe same period, the area covered by coffee-based agroforestrysystems increased from about 8% in 1970 to 20% in 1978 toabout 63% in 1990 to about 70% in 2000. Coffee is grown inthree production systems: monoculture coffee, shade coffee,

2041E C O L O G I C A L E C O N O M I C S 6 8 ( 2 0 0 9 ) 2 0 4 0 – 2 0 5 0


and multi-strata agroforests. Shade coffee and multi-strataagroforests have been expanding since 1984 and now occupyabout 36% of the study area (Verbist et al., 2005). The rate ofdeforestation peaked in the 1999–2000 period, when farmerstook advantage of the early days of Reformasi to expand coffeeproduction in the protection forest and national park. Despiteits higher conservation status under the Forestry Law, the rateof deforestation has been higher in the National Park than inthe protection forest (Ekadinata et al., no date). Studies by VanNoordwijk et al. (2007) and Verbist et al. (2005) have shownthat well-managed multi-strata agroforestry systems can beconsistent with good soil and watershed management.Suyanto et al. (2005) show that farmers with secure propertyrights are more likely to establish multi-strata agroforestrysystems than monoculture coffee systems.

Since the late 1990s, conflict and socio-economic tensionbetween coffee growers and government officers has relaxedas processes have become more open and governmentagencies have shown greater willingness to work with com-munities for better forest management. In the year 2000, thegovernment of the Province of Lampung introduced a tax onnon-timber forest products, including coffee, which was seenas implicitly recognizing farmers who cultivate coffee inprotection forests. The Forestry Law of 1999 and decentraliza-tion lawsof 1999 and 2004 increased the scope for local forestryofficials to negotiate land use agreements with local residents.Research organizations and non-governmental organizationsworking in the area have supported negotiations betweencommunity groups and various government agencies.

2.3. Application of community forestry in the Sumber Jayacase study area

In the Sumber Jaya area, coffee agroforests have beenestablished in areas designated as private farm land andareas designated as protected state forest land. Temporary 5-yearHutan Kemasyarakatan (HKm) contracts have been grantedto coffee growers that form a farmers' organization and meetthe criteria for community-based forestry managementaccording to the Decree of Ministry of Forestry Number 31/2001. As of 2005, 28% of the protected forest area wasmanagedthrough HKm contracts, 56% of the area was in process ofnegotiation for HKm contracts, while the remaining 16% hadno community forestry status.

Some elements of the HKm contracts are explicitly statedin the contracts that have been offered to farmers in theSumber Jaya area:

1. Conditional tenure security to utilize forest land for aprobationary period for 5 years. If the group of HKm farmerssatisfies all criteria and indicators of HKm requirement, thetenure could be extended to a maximum of 25 years.

2. Farmers obtaining HKm contracts are required to growtimber trees in the protection forest land. Differentgovernment programmes have required densities between400 and 1000 trees per hectare. The current HKm contractsin Sumber Jaya require aminimumof 400 trees per hectare.

3. HKmholders can continue growing coffee and accompany-ing shade trees as long as at least 30% of the trees are oftimber species. (The underlying rational for requiring

timber species was that they provide better anchoring forthe soil because of their deeper roots.)

4. Farmers have no right to sell and cut the timber trees thatthey plant.

5. HKm contracts in Sumber Jaya do not require the paymentof any annual fee from the farmers, although farmers livingin neighboring regions are required to pay an annual fee.

During open-ended interviews conducted in the preliminarystages of this research, some farmer groups that had attainedHKmcontracts indicated that theyhad receivedbenefits beyondthe explicit elements of their contracts. It was suggested that bygaining the HKm status, farmers had become more eligible forparticipation in forestry extension programs, and had morerights to ask for public infrastructure to be constructed in theHKm areas, and their families had more legitimate claims onpublic services available in nearby towns. The conjoint analysisstudy of farmers' preferences for the attributes of communityforestry contracts therefore considered both the explicit andimplicit elements of the contracts.

3. Applying conjoint analysis for assessingpreferences for the elements of communityforestry contracts

3.1. Application of the conjoint analysis approach

This study used conjoint analysis to evaluate farmers' prefer-ences over the explicit and implicit attributes of communityforestry contracts in the Sumber Jaya area of Indonesia. The term‘conjoint analysis' refers to an overall approach and group ofquantitative techniques that can be used to determine respon-dents' preferences for the attributes that make up a product orservice,with the totalworth of a product determined by the part-worths of its attributes (Sayadi, Roa andRequena, 2005). Over thelast fewdecades, conjoint analysis has been increasingly appliedin the field of resource economics for valuation of ecosystemcomponentsand for ascertaining thepreferencesof stakeholdersin environmental decisions involving tradeoffs not efficientlyrepresented inmarket transactions (Mackenzie, 1993; Boyle et al.,2001). Conjoint analysis has previously been used to supportnegotiations about the elements of labour contracts (eg. Green-halgh and Neslin, 1981) and the management of agro-ecosys-tems (Klosowski et al., 2001) in the United States.

We begin with the assumption that farmers derive utilitydirectly from the explicit and implicit attributes of the HKmcontracts, with each attribute represented by a small numberof levels.

The utility derived from HKm contracts is expressed as aquasi-concave, twice continuously differentiable utility func-tion (Eq. (1)):

Ui Phð Þ =U Zh;Xif g ð1Þ

where:

Ph is the h-th hypothetical package of HKm contractattributes;

Ui(Ph) is the utility that the i-th farmer derives from theh-thHKm contract;

2042 E C O L O G I C A L E C O N O M I C S 6 8 ( 2 0 0 9 ) 2 0 4 0 – 2 0 5 0


Zh is a vector of levels making up the attributes of theHKm contract, Ph;

Xi is a vector of characteristics of the i-th farmer.

The corresponding indirect utility function Vi(P) has asystematic component vi(P) and a random unobservablecomponent, ε, that is unique to each farmer. The indirectutility for the h-th HKm contract is given by Eq. (2).

Vi Phð Þ = v Phð Þ + eih ð2Þ

Field experiments were implemented in which a sample offarmers are asked to give ratings of hypothetical packages ofHKm contracts with predetermined attributes and levels ofattributes. Farmers were also asked to indicate whether theywould prefer the hypothetical contract or the existing HKmcontract. Previous conjoint analysis studies have shown thatrating and ranking response formats generally producesimilar empirical results (Boyle et al., 2001). Ratings wereused in this study as they provide both absolute and relativelevels of the desirability of contracts. Ratings also avoidrequiring respondents to order alternatives that they may beindifferent to.

A preliminary list of 18 explicit and implicit attributes ofthe HKm contracts and levels of those attributes was devel-oped on the basis of existing contracts (see Table 1), previousstudies of HKm in the area (Arifin, 2006; Suyanto et al., 2005),discussions with key informants, and group interviews at thesite. The explicit attributes were mostly derived from theMinisterial decree that established the HKm arrangement, aswell as other requirements included in existing HKm con-tracts. From the key informant and group interviews, wehypothesized that farmers associatedwith HKmcontracts hadeasier access to agroforestry extension services, access tobetter roads in the HKm areas, easier access to agriculturalcredit, easier access to agricultural marketing services, and

easier access to public services. These perceptions weresupported by interviews with Forestry Department officialswho revealed that they had deliberately included HKm groupsin forestry extension programmes, in part because the HKmgroup structure made it easier to reach large number offarmers. Another government agency made major renova-tions on at least one road into the HKm area after contractswere signed. Some key informants postulated that easieraccess to credit and marketing services could be built into thecontracts to make the contracts more popular with farmers.

The preliminary list of attributes was discussed with agroup of 30 researchers who attended a training workshop onthe conjoint analysis method at the University of Lampung onMarch 17–18th, 2004 and at the site with a group of 6researchers with long experience working on communityforestry issues at the Lampung site. A revised list of 14attributes and levels was then applied in a pre-test of thequestionnaire in the study site. A pre-test with 8 farmers in anarea near to the study site suggested that respondents weredisregarding some of the attributes and considering a smallernumber of attributes when making their choices. Studies ofconsumer choice in a range of conditions show that such anelimination strategy is common when people are presentedwith complex decisions (Iyangar and Lepper, 2000). The list ofattributes and levels was further discussed by team membersand at a seminar held at the offices of the Centre forInternational Forestry Research and World AgroforestryCentre in Bogor, Indonesia. Generally, we removed attributesthat farmers did not expect to be explicitly included or

Table 1 – Explicit and implicit attributes of HKm contractsin Sumber Jaya, Indonesia.

Attribute Requirements

Explicit attribute of contract1. Length of main contract 25 years, after 5 year probation2. Minimum density of trees perhectare

400 trees (5×5 meter spacing)

3. Composition of trees allowed At least 30% of timber trees4. Right to harvest and selltimber trees that you plant onHKm land

No right to cut and sell

5. Fee to be paid to localgovernment for HKm

Rp 0 per year

6. Required contribution of laborwithout pay for group activity(in addition to meeting)e.g.: forest guarding, treeplanting etc

Not specified in HKm contract,but indicated in groupconstitutions

Implicit attribute of contract7. Easier access to agroforestryextension services andseedlings

Not specified

8. Better roads into the HKm area Not specified

Table 2 – Attributes and levels of HKm contracts in SumberJaya, Indonesia.

Attribute Level 1 Level 2 Level 3

Explicit attribute of contract1. Length of main contract 15 years 25 years 35 years2. Required densityof trees/ha

400 (5 m) 600 (4 m) 1000 (3 m)

3. Composition of treesallowed

15%timbertreesrequired



4. Right to cut and sell timbertrees that you plant on HKmland

Haveright tosell

Don'thave rightto sell

5. Level of fee to be paid forHKm⁎

Rp 0 perhectareper year

Rp36,000 perhectareper year

Rp72,000 perhectareper year

6. Required contribution of laborwithout pay for group activity(in addition to meeting) such asforest guarding, tree planting etc

1 daypermonth

5 days permonth

Implicit attribute of contract7. Easier access to agroforestry/forestry extension services andseedlings

Yes No

8. Better roads into the HKm area Yes No

⁎The average rate of exchange between the US$ and IndonesianRupiah in 2005 was 9500 Rupiah/1 US$.



influenced by HKm contracts. For example, farmers did notexpect access to credit or to education services to be explicitlyincluded in HKm contracts. Table 1 lists the eight attributesthat were ultimately included in the study design, includingthe level of that attribute as indicated in the current contract.

The levels of attributes are selected to simplisticallycapture the existing level and realistic alternative levels. Forexample, levels on the length of main contract are set to varyfrom the minimum of 15 years, to the default of 25 years, andto the maximum 35 years. The levels on the attribute of rightto harvest and sell timber are set to binary options: “yes” (havethe right) or and “no” (don't have the right). Implicit attributesrelated to government programs such as access to forestry andagroforestry extension and access on better roads into HKmareas also have only two levels, yes or no. Table 2 lists theattributes and levels included in the final conjoint surveydesign, while Table 3 presents our hypotheses of how theattributes would affect preferences.

3.2. Field survey design

The surveywasadministered to twostrata ofhouseholds. Strata1 included households who already had HKm contracts, whileStrata 2 included households who had applied for, but not yetreceived, HKm contracts. The sampling and interviewmethodswere jointly designed to efficiently cover the range of contractdesign options described in Table 2. There are four attributeswith three levels (length of contract, required tree density,composition of trees allowed, fee) and 4 attributes with 2 levels(right to cut and sell timber trees, required labour contribution,access to agroforestry / forestry extension services, access toimproved roads). The total number of scenarios for the 8attributes is 24 ⁎ 34=1296. Frompre-tests of the conjoint analysismethod, we determined that 12 was an appropriate number ofscenarios to ask each respondent to rate. Fewer than thisrequires an unreasonably large sample, and during the pre-testit became apparent that respondents had difficulty comparingmore than 12. We therefore ascertained that 108 respondentscould handle the full set of scenarios, meaning that eachrespondent responded to a unique set of 12 scenarios (12scenarios×108 respondents=1296 scenarios). The 1296 scenar-iosweredividedamong the108 respondents inaway that gaveadesign as near orthogonal as possible. The problem of findingsuch a design is the same as the problem of designing anexperiment for a 24 ⁎ 34 treatments set in blocks of size 12. Weused the AlgDesign package from R statistical system (Wheeler,2008). Our sampling procedure thereforewas to randomly select108 respondents from the stratumof householdswhohadHKmcontracts and 108 respondents from the stratum of householdswhohadapplied forHKmcontracts. Each respondentwas askedto compare 12 scenarios, selected to give a near orthogonaldesign. A total of 216 interviews were therefore conducted,yielding a total sample size of 12 ⁎ 216=2592 observations acrossthe two strata.

Graduate students from Lampung University and experi-enced research assistants administered the surveys in face-to-face interviews with the respondents. The surveywas adminis-tered to the householdmember identified as the HKmmember,all whomweremale. Customs and religious beliefs in the studyarea indicate that negotiations with outsiders should be

conducted through male representatives (Agrawal et al., 2006).Thus all of the survey respondents were men.

The questionnaires were written in Bahasa Indonesia inorder to ease the process of data collection and improvecommunication between enumerators and respondents. Anintroduction of about 30 minutes was given before therespondent was asked to provide ratings and preferences onthe package of HKm contracts. The attributes and levels weresummarized in the questionnaire in a simple one-page table

Table 3 – Hypotheses regarding the effects of the attributesand household characteristics on respondents' ratings.

Attribute Hypothesized affect onratings

Explicit attribute of contract1. Length of main contract Positive and relatively important:

longer contracts will implygreater tenure security, thecentral concern of farmers.

2. Required density of trees/ha Negative and relativelyimportant: requirements forhigher density of trees will implyless discretion over land use forfarmers, particularly forintegrating trees with coffee andannual crops.

3. Composition of trees allowed Negative and relativelyimportant: requirements forhigher percentages of timbertrees will imply less discretionover land use for farmers.Expected an interraction withthe right to cut and sell timber.

4. Right to cut and sell timbertrees that you plant onHKm land

Negative and relativelyimportant: not having the rightto sell timber trees on HKm landwill imply less potential forfarmers to benefit from the land.

5. Level of fee to be paid for HKm⁎ Negative and relativelyunimportant. Although the levelof the fee imposes a cost onfarmers, the levels included inthe survey are small compared tothe potential earnings thatfarmers stand to make.

6. Required contribution of laborwithout pay for group activity(in addition to meeting) such asforest guarding, tree plantingetc

Negatively and relativelyunimportant. Althoughimposing a cost to the farmers,the cost would be considered aworthwhile investment.

Implicit attribute of contract7. Easier access to agroforestry/forestry extension services andseedlings

Negative and relativelyunimportant. Although betteraccess to extension would beconsidered an advantage, itwould confer minimal benefit tofarmers.

8. Better roads into theHKmarea Negative and relativelyimportant. Better roads into theHKm area could have majorimpacts on the economics ofproduction.



that could be easily understood by the respondents. Asexplained above, each respondent was asked to rate 12scenarios or hypothetical contracts. Respondents were askedto provide ratings of between 1 and 5 for each hypotheticalcontract, with 1 representing a low desirability and 5 represent-ing high desirability, as well as an indication of whether or notthey preferred the hypothetical contract over the currentcontract. Respondents were also asked to provide informationabout household characteristics that we hypothesized to haveimportant affects on preferences. These included age ofhousehold head, family size, ethnicity, years of education ofthe household head, knowledge of HKm, previous access totechnical assistance, and whether the HKm contract hadalready been granted or was still being processed. We collectedthese household data to test hypotheses about the effects ofhousehold characteristics on preferences. We hypothesizedthat: (1) preferences over length of contractwould be affected bythe household's HKm status; (2) preferences over the HKm feewould beaffectedby thehousehold's assets; (3) preferences overthe need to contribute to group activities would be affected byfamily size; and (4) preferences over tree density requirementswould be affected by previous access to technical assistance.

3.3. Logit model

Two statistical analyses was undertaken to evaluate farmers'preferences over the levels and attributes of the contracts.First, a logit or logistic regression model was used to assesshow the levels of the attributes affected whether the farmerpreferred the hypothetical contract over the current contract.

Expanding Eq. (2) from above for our case, we obtain Eq. (3):

Vi Phð Þ4 = blZ1h + N b8Z8h + c1X1 + N cRhXR + eih ð3Þ

Where: Vi(Ph)⁎ is an unobserved measure of the utility thatfarmer i derives from the attributes of the h-th HKm contract

Z1h … Z8h is a vector of levels of the observed attributes of theHKm contract

X1 … XR is a vector of the respondent's characteristicsb1 … b8 and c1h … cRh are unknown parametersεih is a random error term.

While Vi(Ph)⁎ cannot be observed, the binary choiceexperiment provides information about whether the farmerprefers or does not prefer the hypothetical contract to theexisting contract. The rational farmer will choose the existingcontract if the utility he or she expects to derive from theexisting contract is equal to or greater than the utility he or shewould expect to derive from the hypothetical contract. Inother words:

BCi = 0 if Vi4 P0ð ÞzVi4 Phð ÞBCi = 1 if Vi4 P0ð ÞbVi4 Phð Þ ð4Þ

Where: BCi=0 if farmer i prefers the existing contract andBCi=1 if farmer i prefers the hypothetical alternative contract(BC represents binary choice).

Pr BCi = 1½ � = Pr Ui4 Phð ÞNUi4 P0ð Þ½ �= Pr½ ei0 � eihð Þb b1Z1h + N + b8Z8h + c1hX1 + N + cRhXRð Þ� b1Z10 + N + b8Z80 + c10X1 + N + cR0XRð Þ�

ð5Þ

ZX is the vector of attributes of the HKm contracts andrespondent characteristics

β and c are vectors of parameters to be estimated

Assuming that the εs are independently and identicallydistributed, the appropriate functional form of (εi0 −εih)defines the appropriate estimation technique for estimat-ing the utility difference (Eq. (5)). Here we assume that(εi0 −εih) is distributed according to the logistic function,making logit fitted with maximum likelihood an appro-priate estimation technique. Use of the logistic distributionrather than common alternatives (such as the normal)generally makes very little difference to results. Howeverthe logit technique has the advantage that the parameterestimates are easily interpreted as the logarithm of theodds ratios. An odds ratio is the ratio of the odds of anevent occurring in one group to the odds of it occurring inanother group. For a given attribute, an odds ratio of oneimplies that the attribute has no effect on the odds thatrespondents will prefer the hypothetical contract to anexisting contract. An odds ratio less than one indicate thathigher levels of the attribute reduce the odds thatrespondents prefer the hypothetical contract to the exist-ing contract. An odds ratio greater than one implies thathigher levels of the attribute increase the odds thatrespondents will prefer the hypothetical contract com-pared to the existing contract.

3.4. Ordered logit model

Another statistical procedure was appropriate for theanalysis of the ratings data. The dependent variable is therating between 1 (least preferred) and 5 (most preferred),while the independent variables are the levels of the 8attributes and the characteristics of the respondents. Weassume that any contract that the farmer rates with ahigher number is preferred over any contract that he or sherates with a lower number, but we do not assume that theintervals between the ratings are equal. For example, we donot assume that the difference in preference betweenratings 1 and 2 is the same as the difference in preferencebetween ratings 4 and 5. The ratings therefore are char-acterized as discrete and ordered, but not ordered by equalinterval. Again assuming that the error terms are distrib-uted over the logistic function, ordered logit fitted bymaximum likelihood is an appropriate analytical approachfor this analysis (Greene, 1993).

The starting point for the ordered logit model is again Eq.(3) above, where the indirect utility that is derived from anHKm contract is a function of the attributes of the contractand the respondent's characteristics. While the indirectutility derived from a particular contract cannot beobserved, we observe the ratings (Rating) of between 1 to5, where:

Rating¼1 if Ph⁎Vl1

Rating¼2 if l1bPh4bl2

Rating¼3 if l2bPh4bl3



Rating¼ 4 if l3bPh4bl4

Rating¼ 5 if Ph4bl4

Where µ1 … µ4 are estimated cutoff points.The probability that the farmer will give a rating of j to the

h-th HKm contract is given as:

Phj = Pr Rating = j½ � = Pr cj�1b blZ1h + N + b8Z8h +c1hX1 + N +cRhXRð Þbcj

hð6aÞ

=L cj� blZ1h + N +b8Z8h +c1hX1 + N +cRhXRð Þ� �

�L cj�1� blZ1h + N +b8Z8h +c10X1 + N +cR0XRÞÞð�

ð6bÞ

Where L is the cumulative logistic density function andother symbols are as for the logit model above.

The marginal effect of each variable in either the logit orordered logit model is its influence on the response. The usualdefinition of the marginal effect of zk is

ApAzk

, where p=Pr[BCi=1] inthe case of the logit model and p=Pr[ratingb j] for some j in thecaseof theordinal logitmodel. Thevalueof thesemarginal effectsdepends on the value of ZX at which they are calculated. This istakenas thebaselineor currentHKmcontracts forZvariables andthe most common (discrete) or average (continuous) values forXvariables. For the ordered model, marginal effects are presentedfor j=3. While these marginal effects show the relative impor-tance of thevariables, theyarehard to interpret as actual changesin p due to the nonlinearity of the model. Hence we also presentchanges in p, defined as Δpk=pk−p0, where p0 is the value for thebaseline and pk is the value of p for a hypothetical contract inwhich only the k'th variable only is changed from the baseline.

4. Results

This section presents the research results, including descrip-tive statistics about the respondents and econometric analysison the attributes of HKm contracts.

4.1. Respondent characteristics

The average respondent of this survey was 43.7 years forhouseholds with HKm contracts and 42.6 years for householdswith HKm contracts in process. Respondents with the HKmcontracts had an average of 6.5 years of formal education,while those with the HKm in process had an average 5.9 yearsof education. Average family sizes were generally small, 4.0 inhouseholds with HKm contracts and 3.6 in households withHKm contracts in process. All of the survey respondents weremale.

The major ethnic groups of respondents with HKmcontracts were Javenese (45%) and Sundanese (43%), whilethose with HKm contracts in process were Javanese (43%) andSundanese (34%). The majority of respondents were migrantsfrom Java or children of migrants. Indigenous Lampungesewere a minority. The demographic characteristics of respon-dents are summarized in Table 4.

Additional characteristics of the respondents are presentedin Table 5. A large majority of the respondents listedagriculture as their primary occupation. The total value ofthe assets owned or controlled by the households was Rp57.5 million for households with HKm contracts and 61.5 mil-lion for households with contracts in process. This includesthe value of their house, land parcels, animals, and transpor-tation equipment. Respondents in households with HKmcontracts were generally aware of the contracts (84%), whileawareness was lower among households with contracts inprocess (48%). Access to formal credit was similar with the twogroups, while access to extension services was higher forhouseholds with HKm contracts (84%) compared to house-holds with contracts in process (59%). These differences inawareness of HKm and access to extension were the onlynotable differences between the two groups.

4.2. Preferences for hypothetical contracts compared toexisting contracts

Table 6 presents the results of the logit analysis on preferencesfor hypothetical contracts compared to the existing HKmcontracts. As shown in Eq. (5), the explanatory variables

Table 4 – Demographic characteristics of respondents tothe conjoint survey, Sumber Jaya, Indonesia.

Important variables HKm contract HKm process

Mean Std.dev

Mean Std.dev

1. Age (years) 43.7 12.8 42.6 12.42. Education (years) 6.5 3.1 5.9 3.33. Household size (number) 3.6 1.3 4.0 1.54. Ethnic group–Javanese (%) 45.4 4.8 42.6 4.8– Sundanese (%) 42.6 4.8 34.3 4.6– Lampungese (%) 12.0 3.2 23.2 4.1

5. Migration status– From Java 35.2 4.6 38.0 4.7– From other districts inLampung

19.4 3.8 17.6 3.7

– From other places in thecountry

8.3 2.7 7.4 2.5

6. Permanent migrant status (%) 98.2 1.3 99.1 0.97. Secondgenerationmigrant (%) 34.3 4.6 33.3 4.68. Third generation migrant (%) 2.8 1.6 3.7 1.8

Source: Calculated from field survey.

Table 5 – Social-economic characteristics of respondentsto the conjoint survey, Sumber Jaya, Indonesia.

Important variables HKMcontract

HKm process

Mean St.dev Mean St.dev

Agriculture as primary occupation(%)

97.2 1.0 97.2 1.6

Total asset owned (Rp million) ⁎ 57.5 2.0 61.5 2.3Awareness of HKm tenure (%) 84.3 3.5 48.2 4.8Access on technical assistance (%) 81.4 3.8 59.3 4.8Access on formal credits (%) 39.8 4.7 38.9 4.7

⁎The average rate of exchange between the US$ and IndonesianRupiah in 2005 was 9500 Rupiah / 1 US$.Source: Calculated from field survey.



include the contract attributes and characteristics of therespondent households. In addition, we considered fourvariables that represent the interactions between contractattributes and household characteristics. That is, we hypothe-sized: (1) that preferences over length of contract would beaffected by the household's HKm status; (2) that preferencesover the HKm fee would be affected by the household's assets;(3) that preferences over the need to contribute to groupactivities would be affected by family size; and (4) thatpreferences over tree density requirements would be affectedby previous access to technical assistance.

The results for the explicit attributes are generally con-sistent with expectation, with all attributes having statisti-cally significant effects on preferences except for treecomposition and labour contributions to collective activities.The odds ratios shown in the fifth column show that contractsthat require higher tree density, higher fruit tree percentages,and impose higher fees would all be less preferred than the

terms of the existing contract (those with odds ratios less than1). Length of contract has a very large impact on the odds of afarmer preferring the hypothetical contract over the existingcontract: a contract valid for 15 years would be 0.2 times aslikely to be chosen as a contract valid for 25 years, while acontract valid for 35 years would be 4.3 times more likely to bechosen than a contract valid for 25 years. A contract thatallowed farmers to harvest timber trees would be 2.2 timesmore likely to be chosen than a contract that did not allowharvesting of timber trees. The Δp parameter shows themarginal effect of a change in contract length for the basecase situation (discussed above). Everything else equal, short-ening the contract length from 25 to 15 years would reduce theprobability of the contract being selected by 35%, whileincreasing the contract length from 25 to 35 years wouldincrease the probability of the contract being selected by 29%.Allowing the harvest of timber trees would increase theprobability of the contract being selected by 18%.

Table 6 – Logit results on preferences for hypothetical compared to the existing HKm permit.

Definition Estimated coefficients(log odds ratio)

z Significance Oddsratio

Marginaleffect

Change,Δp

ConstantAttributes of the permit:• HKm period 15 years (baseline=25 years) −1.590 −9.05 b .001 0.204 −1.958 −0.351• HKm period 35 years (baseline=25 years) 1.454 9.29 b .001 4.281 1.791 0.289•Tree density 600 trees / ha (baseline=400 trees/ha) −0.637 −2.87 0.004 0.529 −0.785 −0.157• Tree density 800 trees / ha (baseline=400 trees / ha) −1.033 −4.64 b .001 0.356 −1.272 −0.247• Max fruit tree composition 15% (baseline=30%) 0.165 1.37 0.171 1.179 0.203 0.040• Max fruit tree composition 50% (baseline=30%) 0.011 0.09 0.93 1.011 0.014 0.003• Tree cutting rights (baseline=cutting) 0.803 8.02 b .001 2.233 0.989 0.181• Pay fees of 36,000 Rp/year (baseline=no fees) −0.933 −6.52 b .001 0.393 −1.149 −0.226• Pay fees of 72,000 Rp year (baseline=fees) −1.164 −7.99 b .001 0.312 −1.434 −0.274• Contribute 5 days per month to group activities(baseline=1 day)

−0.258 −0.92 0.357 0.772 −0.318 −0.064

• Access to extension services & inputs (baseline=no access)

0.863 8.59 b .001 2.369 1.063 0.193

• Access to roads in HKm area (baseline=no access) 0.383 3.87 b .001 1.466 0.471 0.092

Characteristics of respondent:• Age of household head 0.001 0.22 0.825 1.001 0.001 0.000• Ethnicity of Javanese (baseline=Lampung,Sumendo)

−0.03 −0.21 0.835 0.970 −0.037 −0.007

• Ethnicity of Sundanese (baseline=Lampung,Sumendo)

0.061 0.4 0.687 1.063 0.075 0.015

• Years of education −0.021 −1.14 0.254 0.980 −0.025 −0.005• Aware of HKm (baseline=not aware of HKm) −0.021 −0.17 0.867 0.979 −0.026 −0.005•Access to technicalassistance (baseline=noaccess) −0.496 −2.57 0.01 0.609 −0.611 −0.123• Assets in year 2000 0.003 1.62 0.105 1.003 0.003 0.001• HKm permit in progress (baseline=HKm permitalready issued)

−0.366 −2.34 0.019 0.693 −0.451 −0.091

• Family size in 2005 −0.055 −1.11 0.266 0.947 −0.068 −0.014

Interactions between attributes of permit and respondent characteristics:HKm period 15 years×HKm status 0.014 0.06 0.955 1.014HKm period 35 years×HKm status 0.748 3.31 b .001 2.114Pay fees of 36,000×Assets −0.001 −0.62 0.535 0.998Pay fees of 72,000×Assets −0.000 −0.21 0.833 0.999Contribute to group activities×Family size 0.061 0.89 0.375 1.063Tree density 600×access to technical assistance 0.327 1.24 0.215 1.387Tree density 800×access to technical assistance 0.691 2.61 0.009 1.995

Baseline p0 0.55Pseudo R-squared 0.291



The two implicit attributes — better access to extensionservices and access to better roads in the HKm area — havesignificant effects onpreferences. Contracts thatprovidedbetteraccess to forestry and agroforestry extension services would be2.4 times more likely to be selected than contracts that did notprovide such access, while contracts that provided better roadsinHKmareaswould be 1.5 timesmore likely to be selected thancontracts that did not provide that access. Even though theexisting contracts do not explicitly indicate that members ofHKm groups will obtain better access to extension services andseedlings, the pattern in the study site is that governmentextension providers have given priority to HKm groups.

The results show that respondent characteristics have littleeffect on preferences. Age of household head, ethnicity, house-

hold assets, and years of education all had insignificant effects onpreferences. Perhaps most surprising was that there were nosignificant differences in preferences between respondents whobelonged to groups with HKm contracts and respondents whobelonged togroupswithHKmcontracts inprogress (the twostrataused in the survey). Instead, the results indicate that respondentswho were not aware of the HKm contracts had somewhat lowerpreferences for the hypothetical contracts than respondentswhowere aware of the HKm contracts. They had odds of 0.29 ofpreferring the hypothetical contract to the actual contract. Alsorespondents who had received technical assistance had lowerpreferences for the hypothetical contracts than respondentswhohad not received technical assistance. They had odds of 0.32 ofpreferring the hypothetical contract to the actual contract.

Table 7 – Ordered logit results on ratings of hypothetical HKm Permits.

Definition Estimated coefficient(log odds ratio)

z Significance Oddsratio

Marginaleffect

Change,Δp

Attributes of the permit• HKm period 15 years (baseline=25 years) −1.474 −10.84 b .001 0.229 −4.98 −0.335• HKm period 35 years (baseline=25 years) 1.936 14.63 b .001 6.93 6.54 0.188• Tree density 600 trees / ha (baseline=400 trees / ha) −0.306 −1.82 0.069 0.736 −1.03 −0.058• Tree density 800 trees / ha (baseline=400 trees / ha) −0.553 −3.3 b .001 0.575 −1.87 −0.111• Max fruit tree composition 15% (baseline=30%) 0.138 1.51 0.132 1.148 0.47 0.023• Max fruit tree composition 50% (baseline=30%) −0.083 −0.91 0.365 0.920 −0.28 −0.015• Tree cutting rights (baseline=cutting) 0.609 8.06 b .001 1.838 2.05 0.090• Pay fees of 36,000 Rp / year (baseline=no fees) −0.933 −8.52 b .001 0.393 −3.15 −0.201• Pay fees of 72,000 Rp year (baseline=fees) −1.399 −12.47 b .001 0.247 −4.72 −0.317• Contribute 5 days per month to group activities(baseline=1 day)

−0.286 −1.34 0.179 0.751 −0.97 −0.054

• Access to extension services & inputs (baseline=no access)

0.921 12.02 b .001 2.513 3.11 0.123

• Access to roads in HKm area (baseline=no access) 0.496 6.59 b .001 1.642 1.67 0.076

Characteristics of respondent• Age of household head −0.00205 −0.59 0.555 0.998 −0.01 0.000• Ethnicity of Javanese (baseline=Lampung,Sumendo)

0.146 1.35 0.176 1.157 0.49 0.025

• Ethnicity of Sundanese (baseline=Lampung,Sumendo)

0.153 1.35 0.178 1.165 0.52 0.026

• Years of education −0.0157 −1.15 0.25 0.984 −0.05 −0.003• Aware of HKm (baseline=not aware of HKm) 0.1691 1.77 0.077 1.184 0.57 0.028• Access to technical assistance (baseline=no access) −0.213 −1.45 0.146 0.809 −0.72 −0.040• Assets in year 2000 0.0026 2.1 0.036 1.003 0.01 0.000• HKm permit in progress (baseline=HKm permitalready issued)

0.01 0.08 0.938 1.01 0.03 0.002

• Family size in 2005 −0.0541 −1.44 0.151 0.947 −0.18 −0.010

Interactions between attributes of permit and respondent characteristicsHKm period 15 years×HKm status −0.188 −1 0.315 0.829HKm period 35 years×HKm status −0.001 0 0.998 0.999Pay fees of 36,000×assets −0.002 −1.39 0.165 0.998Pay fees of 72,000×assets −0.002 −1.2 0.232 0.998Contribute to group activities×Family size 0.061 1.17 0.244 1.063Tree density 600×access to technical assistance 0.024 0.12 0.904 1.025Tree density 800×access to technical assistance 0.058 0.29 0.773 1.059

Cut-point 0/1 −3.393Cut-point 1/2 −0.322Cut-point 2/3 1.319Cut-point 3/4 3.216

Baseline p0 0.77Pseudo R-squared 0.291



The results at the bottom of Table 6 show that two of theinteraction terms had statistically significant effects onpreferences. That is, households whose HKm contract wasstill be processed placed greater importance on the length ofcontract than households whose HKm contracts had alreadybeen granted. While statistically significant, the odds ratiowas only 1.04. Households that had access to technicalassistance placed greater importance on tree density thanhouseholds whose HKm contracts had already been granted.The odds ratio on this interaction term was only 1.002.

4.3. Factors affecting farmers' ratings of HKm contracts

Results from the ordered logit analysis of farmer ratings ofthe hypothetical contracts are presented in Table 7. Inqualitative terms, the results of the ordered logit model arevery similar to those generated by the logit model of farmers'preferences of hypothetical contracts compared to theexisting HKm contract. The ordered logit results also indicatestrong preferences over the length of the contract period, therequired density of timber trees, the right to cut the timbertrees, and the level of the annual fee. Rules on thecomposition of trees and required contributions to groupactivities had insignificant effects on farmer ratings. The twoimplicit attributes, access to extension services and roads inthe HKm area, also had significant effects on preferences.The only respondent characteristics that had any effect onpreferences were awareness of the contracts and previousaccess to technical services, both of which reduced farmers'ratings over the hypothetical contracts. The odds ratiosconfirm that length of contract is by far the most importantvariable affecting preferences. None of the interaction termshad significant effects on preferences.

5. Discussion and conclusions

This conjoint analysis study of farmers' preferences for theattributes of community forestry contracts in the Sumber Jayawatershed in Lampung Province applied a logit model toanalyze farmers' comparisons of hypothetical contracts withexisting contracts and ordered logit to analyze farmers' ratingsover hypothetical contracts. Results from the two analyseswere remarkably similar, suggesting that we can be veryconfident in the validity of the findings. The results indicatethat in Indonesia and elsewhere in the developing world,conjoint analysis can be effective in eliciting farmers' pre-ferences over the attributes of community forestry contracts.To our knowledge, this is the first application of conjointanalysis to elicit farmers' preferences for the attributes ofcommunity forestry contracts. Results such as those gener-ated in this study can be used to support conflict resolutionand negotiation between farmers and foresters.

Results from both the logit and ordered logit modelsindicate that the length of the HKm contract is the attributeof greatest concern to farmers in the Sumber Jaya area. Theresults suggest that farmers would be willing to acceptcontracts with many land use and tree planting restrictions,provided that they have certainty that they and their familieswill be able to stay on the land for a relatively long period.

There are two likely reasons why land tenure securityemerged as so important. First, most of the survey respon-dents identify themselves as first or second generationmigrants; as migrants they are less able to claim customaryrights and are particularly interested in being able to benefitfrom long-term investments. Second, the Sumber Jaya areahas a history of conflict over farmers' use of the protectionforests for coffee production (Verbist et al., 2005).

The survey participants also responded very favourably tohypothetical contracts that gave them the right to harvesttimber from their HKm plots. The existing rules and regula-tions on HKm do not allow any farmer to cut the timber in thestate forests, except for some domestic or non-commercialuses. The Forest Department should carefully consider theimplications of allowing farmers to harvest timber forcommercial use on some of protected forest land. Resultsfrom previous studies by Verbist et al. (2005) and Suyanto et al.(2005) suggest that farmers in the area are increasinglyadopting multi-strata agroforestry systems and that landtenure security prompts greater investment in the multi-strata systems. Having the right to harvest timber mightfurther strengthen these investment incentives.

The results also suggest that farmers are very concernedabout the presence and level of any fee levied on the HKmcontracts. This result was surprising for two reasons. First,preliminary interviews suggested that farmers might prefer topay a fee for their contracts so that they would have clearerevidence of government endorsement of those contracts.Second, the fees of Rp 36.000 or Rp 72.000 appear to be verylow ($US 4–8) compared to the potential returns that thefarmers can earn from growing coffee in the HKm plots.Unpublished analyses by the authors indicate that farmers inthe Sumber Jaya area generate about US$231 of revenue andUS$96 of profit from each hectare of coffee garden. A study inthe same area by Arifin (2006) suggests that farmers feel thatthey have paid enough retribution and fees during the earlierstages of obtaining the HKm contract. Farmers also may feelthat the imposition of a fee reduces their tenure security,giving them a status more akin to annual renting than long-term leasing. An approximate analysis based on fitting mixedmodels, with random effects for respondents (results notshown), suggested that farmers concern about length ofcontract and rights to cut trees were the most variablebetween individuals in the sample. This variability betweenindividuals could not be attributed to any of the social char-acteristics considered.

This study went beyond consideration of the explicitattributes of the community forestry contracts to also considertwo implicit attributes that farmers seemed to associate withthe HKm contracts. The descriptive results confirmed thatfarmers with HKm contracts were more likely to get access toforestry and agroforestry extension. Results from the conjointanalysis showed that farmerswould indeed place higher valueon contracts that ensured them better access to extension. Aslong as the norm is for farmers to associate better access toforestry extension with the HKm contracts, the governmentmay not need to explicitly incorporate this into the contract.Results from this work have been formally and informallypresented and discussed with policy makers and otherstakeholders in the community forestry in Indonesia. Most



directly, the results have been considered in the developmentof indicators for the implementation of HKm contracts.Additional efforts need to be undertaken to promote widerdissemination and use of the results on farmers' preferences.A natural follow-up of this study of farmers' preferenceswould be an empirical analysis of policy makers' preferences.Such a study might illuminate some of the other pressuresinfluencing the shape of community forestry contracts.

Acknowledgements

The study reported in this paper was conducted under aproject entitled: “Property rights, Environmental ServiceMechanisms and Poverty in Indonesia.” The project wassupported by the Strengthening Input Systems (BASIS) Colla-borative Research Support Program (CRSP), funded by theUnited States Agency for International Development. Finan-cial support was also provided by the European Union and theInternational Fund for Agricultural Development. The con-tents of this publication are the sole responsibility of theauthors and can in no way be taken to reflect the views of thefunding agencies. The study benefited from useful commentsfrom Meine Van Noordwijk, Peter Frost, John Kerr, JohnPender, members of the faculty of the Department of Agri-cultural Economics at Lampung University, and two anon-ymous reviewers. Edwin Jonson, Noviana Khususiyah,Hutman Kariadi and Rio Jaladri administered the survey andassisted with data management.

R E F E R E N C E S

Agrawal, A., Yadama, G., Andrade, R., Bhattacharya, A., 2006.Decentralization and Environmental Conservation: GenderEffects from Participation in Joint Forest Management. CAPRiWorking Paper, vol. 53. IFPRI, Washington DC.

Arifin, B., 2006. Transaction cost analysis of lowland-uplandrelations in watershed services: lessons from community-basedforestry management in Sumatra, Indonesia. Quarterly Journalof International Agriculture 45, 359–373.

Boyle, K.J., Holmes, T.P., Teisl, M.F., Roe, B., 2001. A comparison ofconjoint analysis response formats. American Journal ofAgricultural Economics 83 (2), 441–454.

Colchester, M., Ekadinanta, A., Fay, C., Pasya, G., Indriani, E.,Situmorang, L., Sirait, M., van Noordwijk, M., Cahyaningsih, N.,Budidarsono, S., Kusters, K., Manalu, P., Gaveau, D., 2005.Facilitating agroforestry development through land and treetenure reforms in Indonesia. ICRAF Southeast Asia WorkingPaper No. 2005_2, World Agroforestry Center, Bogor, Indonesia.

Cubbage, F., Harou, P., Sills, E., 2007. Policy instruments to enhancemulti- functional forest management. Forest Policy andEconomics 9, 833–851.

Ekadinata, A., Dewi, S., Hadi, D.P., Nugroho, D.K., no date. CanSecure Tenure Help Reduce Deforestation? Lessons Learntfrom Sumberjaya Watershed, Lampung, Indonesia. WorldAgroforestry Centre, Bogor, Indonesia.

Fay, C., Michon, G., 2005. Redressing forestry hegemony: when aforestry regulatory framework is best replaced by an agrarianone. Forests, Trees and Livelihoods 15, 193–209.

Fay, C., Sirait, M., Kusworo, A., 2000. Getting the boundaries right:Indonesia's urgent need to redefine its forest estate. InternationalCentre for Research in Agro-Forestry, Bogor, Indonesia.

Greene, W.H., 1993. Econometric Analysis. MacMillan Publishing,New York.

Greenhalgh, L., Neslin, S.A., 1981. Conjoint analysis of negotiatorpreferences. The Journal of Conflict Resolution 25, 301–327.

Iyangar, S.S., Lepper, M.R., 2000. When choice is demotivating: canone desire toomuch of a good thing? Journal of Personality andSocial Psychology 79 (6), 995–1006 2000.

Klosowski, R., Stevens, T., Kittredge, D., Dennis, D., 2001.Economics incentives for coordinated management of forestland: a case study of southern New England. Forest Policy andEconomics 2, 29–38.

Mackenzie, J., 1993. A comparison of contingent preferencemodels. American Journal of Agricultural Economics 75,593–603.

Sayadi, S., Roa,M.C.G., Requena, J.C., 2005.Rankingversusscale ratingin conjoint analysis: evaluating landscapes in mountainousregions in Southeastern Spain. Ecological Economics 55, 539–550.

Suyanto, S., Permana, R.P., Khususiyah, N., Joshi, L., 2005. Landtenure, agroforestry adoption, and reduction of fire hazard in aforest zone: a case study from Lampung, Sumatra, Indonesia.Agroforestry Systems 65, 1–11.

Swallow, B., Kallesoe, M., Iftikhar, U., van Noordwijk, M., Bracer,C., Scherr, S., Raju, K.V., Poats, S., Duraiappah, A., Ochieng, B.,Mallee, H., Rumley, R., 2007. Compensation and rewards forenvironmental services in the developing world: framingpan-tropical analysis and comparison. ICRAF Working Paper32, World Agroforestry Centre, Nairobi, Kenya.

Tomich, T., Cattaneo, A., Chater, S., Geist, H.J., Gochowski, J.,Kaimowitz, D., Lambin, E.F., Lewis, J., Ndoye, O., Palm, C.A.,Stolle, F., Sunderlin, W.D., Valentim, J.F., van Noordwijk, M.,Vosti, S.A., 2005. Balancing agricultural development andenvironmental objectives: assessing tradeoffs in the humidtropics. In: Palm, C.A., Vosti, S.A., Sanchez, P.A., Ericksen, P.J.(Eds.), Slash-and-Burn Agriculture: The Search for Alternatives.Columbia University Press, New York, pp. 415–440.

Tsalikis, J., Seaton, B., Tomaras, P., 2002. A new perspective oncross-cultural ethical evaluations: the use of conjoint analysis.Journal of Business Ethics 35, 281–292.

VanNoordwijk, M., Leimona, B., Emerton, L., Tomich, T.P., Velarde,S.J., Kallesoe, M., Sekher, M., Swallow, B., 2007. Criteria andindicators for environmental service compensation and rewardmechanisms: realistic, voluntary, conditional and pro-poor.CRES Scoping Study Issue Paper 2, ICRAF Working Paper 37,World Agroforestry Centre, Nairobi, Kenya.

Verbist, B., Eka Dinata, A., Budidarsono, S., 2005. Factors driving landuse change: effects onwatershed functions ina coffeeagroforestrysystem in Lampung, Sumatra. Agricultural System 85, 254–270.

Wheeler, B., 2008. The AlgDesign Package. http://cran.r-project.org/.


author's personal copy...author's personal copy and multi-strata agroforests. shade coffee...

Documents