PublicDisclosureAuthorizedPublicDisclosureAuthorized

PolicyResearchWorkingPaper11016

ConservingWildlifethroughDemandReductionandSupplyAlternatives

TwoExperimentsinRestaurantsinKinshasa

AbdoulayeCisseGabrielEnglanderDanielJ.Ingram

WORLDBANKGROUP

DevelopmentEconomics

DevelopmentResearchGroupJanuary2025

ReproducibleResearchRepository

Averifiedreproducibilitypackageforthispaperisavailableat

,clickherefordirectaccess.

PolicyResearchWorkingPaper11016

Abstract

HighaggregatelevelsofwildlifeconsumptioninCentralAfricancitiesmaybedepletingwildlifepopulations.Thisstudyexplorestheimpactsofdemandandsupply-sideinterventionsonwildmeatconsumptionthroughtworandomizedcontroltrialsinrestaurantsinKinshasa,Dem-ocraticRepublicoftheCongo:ademand-sideexperimentandasupply-sideexperiment.Inthedemand-sideexperi-ment,544subjectsweregivenacoupontotheirrestaurantofchoiceandrandomlyassignedtovieweitheratreatmentvideodiscouragingwildmeatconsumptionoracontrolvideounrelatedtowildmeat.Treatmentgroupsubjectsare31%lesslikelytoorderwildmeatthancontrolgroupsubjects,thoughthisdifferenceisnotstatisticallysignifi-cantandmaybeaffectedbysocialdesirabilitybias.Inthesupply-sideexperiment,theresearchquestioniswhetherrandomlyreducingthepriceofMoambeChicken,a

potentialalternativetowildmeat,affectsrestaurants’totalwildmeatsales.Theelasticityestimateindicatesthata1%reductioninthepriceofMoambeChickenreducestotalwildmeatsalesby0.91%.Althoughthisrelationshipisnotstatisticallysignificant,itsuggeststhatinterventionsincreasingtheavailabilityandaffordabilityofalternativestowildmeatmayreducewildmeatconsumption.Theexperimentsadvancepreviouswildmeatresearchbyuti-lizingactualconsumptiondataratherthanself-reporteddata,assessingsocialdesirabilitybias,andpre-registeringallstatisticalspecificationstoenhanceresearchintegrity.Thispaperprovidespreliminaryevidencethatbothwildmeatdemandreductionthroughsocialmarketingcampaignsandsupplyexpansionviaaffordablealternativescouldcontrib-utetoeffectivewildlifeconservationinCentralAfrica.

ThispaperisaproductoftheDevelopmentResearchGroup,DevelopmentEconomics.ItispartofalargereffortbytheWorldBanktoprovideopenaccesstoitsresearchandmakeacontributiontodevelopmentpolicydiscussionsaroundtheworld.PolicyResearchWorkingPapersarealsopostedontheWebat

/prwp

.Theauthorsmaybecontactedataenglander@.Averifiedreproducibilitypackageforthispaperisavailableathttp://,clickherefordirectaccess.

R

Y

C

I

L

R

A

E

S

E

O

P

H

C

S

TRANSPARENT

P

E

R

W

O

R

K

I

ANALYSIS

A

NGP

ThePolicyResearchWorkingPaperSeriesdisseminatesthefindingsofworkinprogresstoencouragetheexchangeofideasaboutdevelopmentissues.Anobjectiveoftheseriesistogetthefindingsoutquickly,evenifthepresentationsarelessthanfullypolished.Thepaperscarrythenamesoftheauthorsandshouldbecitedaccordingly.Thefindings,interpretations,andconclusionsexpressedinthispaperareentirelythoseoftheauthors.TheydonotnecessarilyrepresenttheviewsoftheInternationalBankforReconstructionandDevelopment/WorldBankanditsaffiliatedorganizations,orthoseoftheExecutiveDirectorsoftheWorldBankorthegovernmentstheyrepresent.

ProducedbytheResearchSupportTeam

ConservingWildlifethroughDemandReduction

andSupplyAlternatives:TwoExperimentsin

RestaurantsinKinshasa

AbdoulayeCisse*,GabrielEnglander?,DanielJ.Ingram?

Keywords:biodiversityconservation,wildmeatconsumption,CentralAfrica,randomizedcon-troltrials

JELClassification:Q57,O13,Q21,Q56,D12,C93

*AgriculturalandResourceEconomics,UniversityofCalifornia,Berkeley;California,USA

?DevelopmentResearchGroup,TheWorldBank;Washington,USA.Correspondence:

lr.nservationandEcology,UniversityofKent;Canterbury,CT27NR,UK

2

1Introduction

Millionsofpeopleintropicalregionsoftheworlddependonwildlifeasasourceoffoodandmeansofacquiringincome(Coadetal.,2019;Wellsetal.,2024).Simultaneously,overexploita-tion,includingfromhuntingthatexceedspopulationgrowthrates,hasemergedasthemostsig-nificantthreattomanyspecies(IPBES,2019),impactingwildlifepopulationsandthreateningecosystemfunctioning.Theconsumptionofthemeatofwildanimals,“wildmeat”,isprevalentinruralareas,andthereisalsosubstantialdemandforwildmeatamongpeoplelivinginsomeurbanareas(CarignanoTorresetal.,2022;Edderai&Dame,2006;Ingrametal.,2021;Simoetal.,2024).Consumptionofwildmeatintownsandcitiesisdrivenbyavarietyoffactors,includingculture,tastepreference,theperceptionthatitisthemostnaturalmeat,andtherela-tiveavailabilityandpricecomparedtodomesticatedmeats(Chaussonetal.,2019;vanVliet&Mbazza,2011;Wilkieetal.,2016).Wherewildmeatismoreexpensivethandomesticatedmeatalternativesitcanalsobeconsideredaluxuryproductthatsignifiesstatus(Sandaljetal.,2016).Whendemandforwildmeatinurbanareasdriveshuntingtounsustainablelevels,itposesauniquechallengeandopportunityforconservationefforts.Addressingthischallengerequiresinnovativeandjustsolutionsthatrespectlocallivelihoodsandtraditions,protectwildlife,andcultivatetherelationalvaluesbetweenpeopleandnature(Chanetal.,2018;Ingram,2020).

Therearetwoprimaryapproachestotackleoverexploitationofwildmeatinurbanareas:reducingthedemandforwildlifeproductsorlimitingthesupplyofwildlifetothemarket.

Recentyearshaveseenasurgeinwildlifeproductdemandreductioncampaigns(Verssimo

&Wan,2019;Willisetal.,2022).Forinstance,WildAid,anon-governmentalorganization(NGO),investsheavilyinsocialmarketingcampaignstoreducedemandforwildlifeproducts(WildAid,2020).TheWorldBank’sGlobalWildlifeProgramhasadvocatedforincreasedinvestmentinsuchinterventions(Sobrevila,2016).Alongsideeffortstoreducedemand,supply-

3

sideinterventionsseektolimitthequantityofwildlifesuppliedtothemarket.Forexample,theSustainableWildlifeManagementProgrammeaimstoregulatehuntingandexpandpoultryproductioninsomeofthe15countriesinwhichitoperates(FoodandAgricultureOrganizationetal.,2019).

Bothapproacheshavestrengthsandlimitations.Demand-sideinterventionscancreateawarenessandshiftsocietalnorms,butattemptstochangepreferencesindomainsotherthanwildmeatconsumptionoftenfail(MacFarlaneetal.,2022).Carefuldesignandyears-longcampaignsmayberequiredtochangedeeplyingrainedpreferences.Butifsuccessful,demand-sideinterventionscanprovidealastingbenefittowildlifepopulations:evenifwildmeatwereavailable,fewerpeoplewouldchoosetoconsumeorpurchaseit,reducingtheincentivetohuntwildlife.Supply-sideinterventions,ontheotherhand,canhavemoreimmediateimpacts.Bymoredirectlyreducingthequantityofwildlifekilledforfood,theycanquicklybenefitwildlifepopulations.However,supply-sideinterventionsalonemaynotbesufficienttoreducewildmeatconsumptiontosustainablelevels,becausepersistentdemandleavestheeconomicincen-tiveforhuntingunaddressed.DespitethenumerousandextensiveinitiativesundertakenbygovernmentsandNGOs,theeffectivenessofbothdemand-sideandsupply-sideinterventionsinchangingactualwildlifeconsumptionhabits,especiallyinurbansettings,remainsunderex-plored(Ingrametal.,2021;MacFarlaneetal.,2022;Traversetal.,2021;Verssimo&Wan,2019).

InNovember2023,weimplementedtwocomplementaryrandomizedcontroltrialsinKin-shasa.Ourresearchquestionsare:(1)Doesademand-sideinterventionreducewildmeatcon-sumptioninrestaurants?(2)Doesasupply-sideinterventionreducewildmeatconsumptioninrestaurants?ThestudyfocusedonrestaurantsinKinshasa,thecapitalcityoftheDemo-craticRepublicoftheCongo(DRC)andoneofthefastestgrowingmegacitiesintheworldwithapopulationof~17millionpeople.TheDRCisthelargestcountryinCentralAfrica,

4

comprises61%oftheregion’sforests,andishighlybiodiverse(Granthametal.,2020).Kin-shasaislocatedontheCongoRiver,southoftheworld’ssecondlargestrainforest,andithasapproximately3,000wildmeat(“bushmeat”)restaurants(Faetal.,2019).Wildmeatentersthecitybyroad,river,andplane,anditissoldinmarketsandrestaurantsacrossthecity(Lucasetal.,2022).Arecentstudyestimatedthat8,592wildmeatdishesweresoldinrestaurantsinKinshasaeachday,equatingto1,254tonnes(byliveweight)ofwildmeatannually(Wrightetal.,2022).Primatesandungulateswerethemostpopulartypesofwildmeat,whereasgen-erallythemostfrequentlysolddishesacrossthecitycontainedeitherfish,beansorchicken.WhilstthereappearstobenosingledriverofwildmeatconsumptioninKinshasa,traditionandculture,socialstatus,taste,freshness,cost,andaccessibilityallplayarole(Trefon,2023;Yocumetal.,2022).Furthermore,wildmeatcanbemoreexpensivethandomesticatedmeatinrestaurants,suggestingitcouldbeconsumedmorebywealthierresidents(LaCerva,2016;Wrightetal.,2022).RestauranteursinKinshasastatethattheysellwildmeatbecauseofhighdemand,itsprofitability,tomaintainCongoleseculture,andmenudiversification(Wrightetal.,2022).Indeed,sometraditionalCongolesedishesaremadepredominantlywithwildmeat.Inrestaurantswithmenus,wildmeatisoftenopenlylistedforsale.LawspertainingtowildmeatintheDRCrefertotheactofhunting,requiringapermit,andtothespeciesthatareprotected.Nospecificlawsregulatetheconsumptioninurbansettingsortradeofmeatobtainedbyhunting(SustainableWildlifeManagementProgramme,2024).

1.1ConceptualFrameworkandContributions

Relationalvaluesrefertothepreferences,principles,andvirtuesassociatedwithrelationships,bothinterpersonalandasarticulatedbypoliciesandsocialnorms,thatcontributetoamean-ingfullife(Chanetal.,2016).Unlikeinstrumentalvalues(natureforhumanuse)orintrinsicvalues(natureforitsownsake),relationalvaluesemphasizetheconnectionsandresponsibili-

5

tiesbetweenpeopleandnature(Klainetal.,2017).Ourstudyissituatedwithintherelationalvaluesframework,focusingonhowhumanbehaviors,culturalties,andeconomicincentivesimpactconservationoutcomes.Relationalvaluesguideboththedesignandanalysisofourexperiments.Thedemand-sideexperimentemploysaculturallyresonantvideothatconnectsprideinCongolesewildlifetotraditionalculinaryheritagewithoutwildmeat.Thisvideoaimstostrengthenpeople’sculturalbondstowildlife,inordertoshiftconsumerbehaviorthrougharenewedsenseofstewardshipandidentity(Westetal.,2018).Likewise,inthesupply-sideexperiment,wetestedhowofferingaculturallyrelevantalternative—MoambeChicken—atalowerpricecouldreducewildmeatconsumption.Thisinterventionrespectsexistingfood-relatedculturalpracticeswhileencouragingsustainableconsumption,aligningwithrelationalvaluesthatprioritizeharmonioushuman-environmentinteractions(Schulzetal.,2017).

Ourstudyofferspracticalinsightsforpolicymakersandconservationpractitionersonde-signingculturallysensitiveinterventionsthatresonatewithlocalvaluesandtraditions.Byus-ingactualconsumptiondataandrobustmethodologies—includingrandomizedcontroltrials,pre-registration,andresponsebiasmeasurement—weadheretobestpracticesinconservationresearch(Cisseetal.,2023).Internationally,thedemandandsupply-sideinterventionsweevaluatecouldbetailoredtourbanareasbeyondKinshasawherewildlifeconsumptionposesconservationchallenges.Ourstudy’stheoreticalcontributiontotherelationalvaluesframeworkisdemonstratinghoweconomicincentivescanbeintegratedwithculturalvaluestoeffectbe-havioralchange.Byaddressingthesocio-culturaldriversofwildmeatconsumption,ourworkhighlightsthecriticalroleofintegratingrelationalvaluesintoconservationeffortstofostersustainablebehaviorsthatarebothecologicallybeneficialandculturallyacceptable.

6

2Methods

2.1DesignofDemand-SideExperiment

Afundamentalchallengeinevaluatingtheeffectivenessofdemand-reductioninterventionsisthedifficultyincontrollingexposuretospecificmessages.Mediumslikebillboardsortelevi-sioncommercialsofferlimitedinsightintowhoviewsthemessageand,morecritically,whoconstitutesthecontrolgroup(e.g.,thosenotexposedtothemessage).Establishingacrediblecontrolgroupisnecessaryasitactsasacomparisonpointorcounterfactual,illustratingwhatthewildmeatconsumptionlevelmighthavebeenintheabsenceoftheintervention.Thus,ourdemand-sideexperimentwasdesignedwithadualfocus:first,tomanageanddocumenteachparticipant’sexposuretotheinterventionmessage;andsecond,tolinkthisexposuretotheirsubsequentconsumptionchoice.

RestaurantsinKinshasapresentanidealsettingforourexperimentduetothereal-worlddecision-makingenvironmenttheyprovide.Whenpatronsdineatarestaurant,theymakerealchoicesbasedontheirpreferencesandbudget,spendingtheirownmoneyonthedishestheychoose.Thissettingfacilitatesthemeasurementofchoicesrepresentativeofactual,ratherthanhypothetical,consumptionbehavior.WerecruitedfourrestaurantsinKinshasatoparticipateinourstudy(Figure1),basedontwocriteria:theyneededtoofferbothwildmeatandnon-wildmeatoptionsdaily,ensuringavarietyofchoicesforcustomers;andtherestaurantownersagreed,viacontract,toreporteachparticipant’sordertous,allowingustolinktheirchoicestotheinterventionmessagetheyreceived.

Intheexperiment,enumeratorssetupfourtables,eachrepresentingoneoftheparticipat-ingrestaurants.Thesetableswerepositionedtwoorthreeblocksawayfromtheirrespectiverestaurants,ensuringthatthestudyandtherestaurantswereoutofsightfromeachother.Thisconfigurationservedadualpurpose:itpreventedparticipantsfrombeinginfluencedbythedi-

Latitude

4.31°S

4.33°S

4.35°S

RestaurantARestaurantBRestaurantCRestaurantD

4.37°S

15.26。E15.28。E15.30。E15.32。E15.34。E15.36。E

Longitude

Figure1:ApproximateLocationsofParticipatingRestaurants.Toensureanonymity,theactuallocationsoftherestaurantswererandomlyadjustedbyuptoapproximately3km.Thedashedcircles,centeredonthesemodifiedcoordinates,indicatethatthetruelocationsoftherestaurantsaresituatedsomewherewithinthecorrespondingcircles.TheinsetmapontheleftshowstheDemocraticRepublicoftheCongo(greenpolygon)withinAfrica(grey),whiletheinsetontherighthighlightsKinshasa(greentriangle)withintheDemocraticRepublicoftheCongo(grey).

7

8

rectvisibilityoftherestaurantsduringtheexperimentandensuredthatparticipantswerenotvisibletoenumeratorswhenvisitingtherestaurantsandorderingadish.Theplacementalsoguaranteedthateachsubjecthadconvenientaccesstoatleastonerestaurant.

OvereightdaysinNovember2023,enumeratorsateachtableencouragedpassersbytopar-ticipateinasurvey,offeringacoupontotheirrestaurantofchoiceasanincentive.Participationwaslimitedtoadults(18yearsorolder)whoareactivewildmeatconsumers,definedassub-jectswhoreportedeatingbushmeatatleastonceinthepastmonth.Subjectswererequiredtousetheircouponwithintwoweeks.Enumeratorsfollowedastrictprotocol,attendingtoeachparticipantindividuallytopreventinteractionbetweenparticipantsthatcouldbiastheirchoices.

Subjectswererandomlyassignedtowatchoneoftwovideosonatabletwithheadphones.Thetreatmentgroupwatchedapubliclyavailable90-secondclipfromthe2019multimediacampaign“Let’seatlessbushmeatinthecity”(“Mangeonsmoinsdeviandedebrousseenville”),ledbytheGovernmentofneighboringRepublicoftheCongowiththeaimofreducingwildmeatconsumptioninthecityofPointeNoire.Inthevideo,twochildhoodfriendsre-unite,andwhenonesuggestscookingwildmeat,theydiscusstheenvironmentalconsequencesofurbanwildmeatconsumptiononthedepletionofforestwildlife(Ministredel’EconomieForestire-Congo,2019).TheydecidetoavoidsuchconsumptioninthefutureandinsteadembraceCongolesedishesthatdonotcontainwildmeat(AppendixB).Ratherthanaddressalldriversofwildmeatconsumptionsimultaneously,thevideoweusedfocusedonmessagesthattargetprideinCongoleseforestsandwildlife,whilstalsosupportingtraditionalCongolesecuisine.Thesecouldbeconsideredastargetingdriversprimarilylinkedtocultureandtradition.Thecontrolgroupvieweda90-secondclipfromaCongolesesoapoperaunrelatedtowildmeatortheenvironment.

Afterwatchingthevideo,subjectschoseoneoffourrestaurantsfora$5coupon(2,500CongoleseFrancs).Weprovidedsubjectswiththename,location,andasamplemenufrom

9

eachrestaurant.Wecompensatedrestaurantownersforeachredeemedcoupon.Subjectswereresponsibleforpayingthedifferencebetweenthepriceoftheirmealandthecouponvalue.Forexample,asubjectwouldhavetopay$4fora$9meal($9?$5=$4).Eachcouponhadauniqueidentifier,linkingsubjects’orderstotheirtreatmentassignmentstatus.Subjectswereinformedduringthesurveythattheyshouldorderwhateverdishtheypreferred,regardlessofwhetheritwaswildmeatornot.

Immediatelyfollowingthevideoviewingandrestaurantselection,subjectscompletedasurveyontheirattitudestowardbushmeatandthefrequencyofconsumptionbypeopleintheirsocialnetwork(AppendixC).Theattitudequestionsaskedsubjectstoexpresstheirlevelofagreementwithstatementsthatwildmeatissustainable,fresh,tasty,cool,legal,healthy,andconnectsthemtotheirplaceoforigin.Theseattitudeswereselectedbasedonotherstudies(Chaussonetal.,2019;Wilkieetal.,2016),withtheexceptionoflegalandsustainablewhichweaddedoutofinterest.Additionally,enumeratorsaskedsubjectsabouttheirprideintheenvironmentoftheDRC.Thepurposeoftheattitudinalquestionswastoprovideinsightintowhyanypotentialtreatmenteffectsoccurred.Additionally,subjectsanswered13questionsrelatedtosocialdesirability,enablingustoassessresponsebias(Dharetal.,2022).

554subjectsparticipatedinthedemand-sideexperiment,ofwhom59%usedtheircouponwithintwoweekstoorderamealattheirrestaurantofchoice.

2.2AnalysisofDemand-SideExperiment

2.2.1AnalysisofWildMeatConsumptionandCouponUsage

Tounderstandtheeffectsofthevideoaimedatdiscouragingwildmeatconsumption(treat-ment),ourpre-specifiedanalysesemployordinaryleastsquaresregressiontoestimateequations

10

ofthefollowingform(Cisseetal.,2023):

Yijt=βTi+αXi+κj+ζt+?ijt.(1)

Inthisequation,Yijtisanoutcomevariable,suchaswhetherasubjectorderedwildmeat.Thesubscriptidenotesthesubject,jrepresentsthetablewherethesubjectparticipatedintheexperiment,andtdenotesthedateofparticipation(notthedateatwhichtheyorderedadishintheirrestaurantofchoice).Tiindicateswhetherthesubjectwasinthetreatmentgroupthatwatchedthewildmeatdemandreductionvideo(Ti=1)orinthecontrolgroup(Ti=0).Xiisamatrixofeightindividualsubjectcharacteristics,suchasage,gender,educationlevel,salaryandbusinessownershipstatus.κjandζtarethe“fixedeffects”,orindividualdummycontrolvariablesforeachtable(κj)andindividualdummycontrolvariablesforeachdate(ζt).Includingthesefixedeffectsobviatestheneedforaninterceptterm,asitwouldbecollinearwitheitherκjorζt.Thevariable?ijtistheerrorterm,whichrepresentsfactorsthataffecttheoutcomevariablebutwhicharenototherwiseaccountedforintheequation.Thecoefficientofinterestisβ,whichmeasurestheeffectoftreatmentassignmentontheoutcomevariable.

Thestandarderrorsofallregressioncoefficientsareclusteredatthesubjectleveltomatchthelevelatwhichtreatmentisassigned(Abadieetal.,2023).Thismethodaccountsforpotentialdifferencesinthevariabilityofresponsesacrosssubjects.Ourclusteredstandarderrorsarealso“heteroskedasticity-robust”.Wedonotsolelyrelyonp-valuesininterpretingourresults;wealsoassessthemagnitudesofthetreatmenteffects,andwecomparetheirconsistencyandinter-pretivevalidityacrossthedifferentpartsofourexperiments(McShaneetal.,2019;Wasserstein&Lazar,2016).Whenwedoassessstatisticalsignificance,weusetheconventionalp-valuethresholdof0.05.

Theinclusionofindividual-levelcontrolsandfixedeffectsisaimedatincreasingthepre-

11

cisionoftheestimatedimpactofthetreatment,β.Theyarenotnecessaryforestimatingthemagnitudeofthetreatmenteffectbecauserandomassignmentensuresthatanydifferencesinoutcomesbetweenthetreatmentandcontrolgroupcanbeattributedtothetreatmentvideo(Ru-bin,1974).InEquation1,Xiincludesanindicatorforwhetherthesubjectreportsusuallyeatingwildmeatatformalorinformalrestaurants,thenumberofdaysthesubjecthaseatenwildmeatinthelast30days,theageofthesubjectinyears,anindicatorvariableforwhetherthesubjectismale,anindicatorforwhetherthesubjecthasattaineduniversity(post-secondary)education,thetotalnumberofyearsofeducationthesubjecthas,anindicatorvariableforwhetherthesubjectwasemployedforpayintheprevious7days,andanindicatorvariableforwhetherthesubjectworkedforpayastheownerofabusinessintheprevious7days.Foranymissingvaluesinthesecontrolvariables(exceptforindicatorvariables),weimputethevaluewiththemeanamongallnon-missingvalues.Forexample,ifageismissingfor44subjects,weassumetheiragetobethemeanageamongthe500subjectswithrecordedages.Wechosetheseeightcontrolvariablestocapturekeybehavioral,demographic,andeconomicfactorsthatmayinfluencewildmeatconsumption,whilemaintainingsurveybrevity(Chaussonetal.,2019;WildAid,2021).Othervariables,suchasethnicityordurationofresidencyinKinshasa,mayalsopredictwildmeatconsumption.

Theeightexplanatoryvariableswechosecanbegroupedconceptuallyintothreecategoriesofwildmeatconsumptionpredictors.Weexpectedthefirsttwo,whethersubjectsusuallyeatwildmeatatrestaurantsandthenumberofdayssubjectshaveeatenwildmeatintheprevious30days,topositivelycorrelatewithsubjectsusingtheircoupontoorderwildmeat.Wechosethenextfour—age,sex,whetherthesubjectattaineduniversityeducation,andyearsofeducation—ascapturingthemostessentialdemographicinformation,includingcharacteristicsthatpredictwildmeatconsumption.Forexample,inPointeNoire,RepublicoftheCongo,youngerpeopleeatlesswildmeat,andmaleseatmorewildmeat(Chaussonetal.,2019).Finally,weexpected

12

thelasttwovariables—inpaidemploymentandbusinessowner—tobepositivelycorrelatedwithsubjectsusingtheircoupon,becausewethoughtrichersubjectswouldbemoreabletopaytheportionoftherestaurantmealnotcoveredbythecoupon.

2.2.2AnalysisofSelectionofWildMeat-IntensiveRestaurants

Thedecisionofwhatdishtoorderatarestaurantfocusesonthequantityofwildmeatconsumed,ignoringtheprecedingchoiceofwhichrestauranttovisit.Ourtreatmentvideo,anddemand-sideinterventionsmorebroadly,mightinfluencenotjusttheselectionsmadewithinaneatingenvironmentbutalsothechoiceoftheeatingenvironmentitself.Withthefourrestaurantsinourstudyofferingvaryingnumbersofwildmeatdishesandthesedishesconstitutingdifferentproportionsoftheirtotalmenus,wehavetheopportunitytoassesswhethersubjectsexposedtothetreatmentaremorelikelytochoosecouponstorestaurantsthatfeaturelesswildmeat.

Forthisanalysis,wecategorizerestaurantsbasedontheproportionofwildmeatdishesfeaturedontheirsamplemenus,asshowntosubjectsduringthedemand-sideexperiment.Ourpre-specifieddefinitionidentifies“wildmeat-intensive”restaurantsasthosewiththehighestproportionofwildmeatdishes(23%and21%,comparedto7%and9%).Alongsidetheorigi-nalcontrolvariablesinEquation1,weincludeascontrolsthedistancefromthesubject’stabletothechosenrestaurantandtheaveragepriceofdishesontherestaurant’ssamplemenu.Weincludedthesevariablestoenhancetheprecisionofourestimateofthetreatmenteffectbycon-trollingforfactorslikeconvenienceandcost,whichmightaffectasubject’schoiceofrestaurantindependentlyoftreatmentassignment.

2.2.3AnalysisofAttitudesTowardsWildMeat

Tounderstandthemechanismsbehindthepotentialchangesinwildmeatconsumption,wein-vestigatehowthetreatmentvideoaffectssubjects’viewsonvariousattributesofwildmeat.

13

Wepre-specifiedthisanalysistodiscernwhetherchangesinwildmeatconsumptionareduetoshiftsinperceptioncausedbythetreatmentvideo.Afterwatchingthevideoandchoosingtheirrestaurant,subjectsansweredeightquestionsmeasuringdifferentaspectsoftheirviewsofbushmeat(AppendixC).Toavoidprimingtherespondentsortheenumeratorsintochoosingthefirstresponseoptions,werandomlyvariedtheorderwithwhichwedisplayedtheresponseoptions(fromcompletelyagreetocompletelydisagreevs.fromcompletelydisagreetocom-pletelyagree).

Weaggregateresponsestothesequestionsintoasingleindex,codingresponsestoindicateanegativeviewof