Privatecostsforenvironmentalgoodsprovisionina
developmentcontext:
labandfieldtestsofanovelcost-revealingmechanism
PreparedforCBEAR-M
AAP2017
SamuelD
.BellOregonStateU
niversityOctober15,2017
Roadmap
1.Motivation
•Setting
•ResearchO
bjectives2.
RandomQuantityM
echanism(RQ
M)
•Features
•Mechanics
3.Experim
entalTestoftheRQM
•Experim
entaldesign•Results
4.Sm
allholderSupplyofEnvironmentalG
oods:EvidencefromAgroforestry
•Settingandexperim
entaldesign•Results
Setting&M
otivation
•Paym
entsforenvironmentalservices,w
ith:•Heterogeneousprivatecosts
•Inform
ationasymmetries
•Threebroadoptionstoim
proveefficiency•Targetoncharacteristicscorrelatedw
ithprivatecostsofprovision•Procurem
entauctionsforallocatingcontracts•Offerscreeningcontractstailoredtodistributionofprivatecosts
•Betterinform
ationonprivatecostscanimproveefficiency
•Random
quantitymechanism
(RQM):
•Truthfullyrevealsprivatecosts
•Canestim
ateheterogeneoustreatmenteffects
Researchobjectives
1.Testperform
anceoftheRQMusinginducedcostsinexperim
entallab;
2.Im
plementtheRQ
Minafieldsetting;
•sm
allholderagroforestryinZambia
3.AssessthepotentialoftheRQ
Masacontractingm
echanismand/or
researchtool.
WhatistheRQ
M?
Cost-revealingextensiontotheBecker -DeG
root-Marschak
mechanism•Providesquasi -experim
entalvariationintreatment(contract
allocation)•Providesrandom
variationincontractterms
•Enablesdirectestim
ationofwillingnesstoaccept(W
TA)across
intensivemarginsusingarepeatelicitationform
atØ
Allow
sestimationofcoststructuresandsupply
•Enablesestim
ationofheterogeneoustreatmenteffects
RQMProcedure
Threesteps:A.
Transfervalue->B.Quantityoffer
->C.Randomquantitydraw
•ContractiffC≤BØ
Optim
alstrategyistoofferquantityforwhichtotalcostofproductionequals
transfer
•Sim
ilardecisiontasktotheBDM(in’reverse’).
•ManyofBD
McharacteristicsapplytoRQ
M
•Profit=transfervalue-privatecostofdraw
nquantity
Inducedcostlabexperiment
•Procurem
entsetting-participantsmakeproductiondecisionsbased
oninducedcostsofproductionandcontract value/transfer•Repeatelicitationacrossfivefirm
s•5differentcoststructures,hypotheticalhom
ogeneousgood.
•Participantsm
adequantityoffersinresponseto3randomly-draw
ntransferpaym
entswithineachfirm
•Eachparticipantm
ade15offersintotal(3pereachof5firms)
•Marketclearingonceallofferroundscom
plete,tolimitlearning
•Onecontractperfirm
implem
ented(randomdraw
oftransferandQ)
•20participants,$34averageearnings
Labexperimentresults
Efficiencymeasures:aggregateandindividualexpectedpayoffs
•17outofthe20participant’soffersresultedinoptim
alexpectedpayoffs•79%
ofthepooledoffersresultedinoptimalexpectedpayoffs
Conclusions:•Experim
entaltestsupportsincentivecompatibility
•Costrevealingandefficientinexpectedpaym
ents•Caveat:sam
plesizequitesmall
RQMfieldexperim
ent
•Sm
allholderfarmers(n=223)inZam
bia•Agroforestrytreeplanting-environm
entalgoodwithpublicand
privatebenefits
•Testnew
mechanism
inthefield–doestheRQ
Mwork?
•Constructaggregatesupplycurves•Com
pareRQMagainst(sim
ulated)procurementauction
•Exploredeterm
inantsofWTA
Fieldimplem
entation
•Sm
allholderisinformedofthecontractpaym
entandaskedhowm
anytreeshe/shew
ouldplantandmaintain
•Repeatedfor5differenttransfers
[20,40,70,100,140]in‘000ZMK
•Onetransferrandom
lydrawn,thenquantitydraw
nfrom[12,25,37,50,75]
•Quasi-random
variationincontractterms
•Paym
entstructure•50%
ofpaymentm
adefortransplantingallseedlings(activity-based)•50%
paidpro-rataonsurvival1yrfollowingplanting(outcom
e-based)
•PartnerN
GOm
onitoredtransplantingandoutcomes.
•60%
ofparticipantsreceivedcontracts
DeterminantsofW
TA
•Regressquantityoffersonsurvivalexpectationsandobservables,controllingfortransfervalues•Survivalexpectationsinteresting-m
ayaffecteffortandoutcomes
(directionambiguousex-ante)
•HighersurvivalexpectationsreducesW
TA–i.e.low
ercostfarmershave
higherexpectations•Possibilityoftargetinglow
-costfarmersonobservables,ifcorrelations
strong•Sellingfoodsandcrafts,Lengthoffoodinsecurity(m
onths)havesignificantpositiveeffects.
•Fem
aleheadedhouseholdstrongpositiveeffect(lowerW
TA),fem
aleisnegativelycorrelatedandm
arginallysignificant.Intra-householddecisionmakingeffect?
Comparisonof:
•RQ
Maggregatesupply(bytransfer)
•RQ
Maggregatesupply(pooled)
•Single-bidm
ulti-unitauctionaggregatesupply
Whatifthesesupplycurvesare
usedtoinformpostedprice?
•Illustrativebudgetconstraint$600:
RQMvs.G
eneralizedVickrey
Quantitytarget:
Matchingthefield
experiment(3,677)
RQMvs.G
eneralizedVickrey
•RQ
Misausefulresearchtool,lesseffectiveasanallocation
mechanism
:•Providesprecisem
easureofWTAallow
ingdirect,non-parametricestim
ationofsupply(likeauctions)•Allocatescontractsw
ithpositiveprobability,allowingotherw
iseout-of-sam
plepredictionsandprovidesexogenousvariationincontractallocationandterm
s,allowingestim
ationofheterogeneoustreatmenteffects
•Random
variationinpricepaid,conditiononmeetingparticipation
constraints•Helpsinseparatingselectioneffectsfrom
causalimpactofpaym
ents
Determinantsofcontractperform
ance
Results–contractoutcom
es
•Largesttransferpaym
entresultsin1.2to3.4extratreessurvivingcom
paredtosmallesttransfer
•Largestquantitydraw
resultsin2to3.9fewertreessurviving
•Survivalexpectationshave(sm
all)negativeeffectonsurvivaloutcom
es•Low
costfarmershavelow
ersurvival.•“Prem
ium”constructedtoexploretheim
pactofhigherpayments
conditionalonagivenWTA.
•Causaleffectofpayment:paym
entsaboveminim
umW
TA
significantlypositivelyimpactsoutcom
es.
Privatecostsforenvironmentalgoodsprovisionina
developmentcontext:
labandfieldtestsofanovelcost-revealingmechanism
PreparedforCBEAR-M
AAP2017
SamuelD
.BellOregonStateU
niversityOctober15,2017