25
1 UAA Department of Economics Working Paper 1 Title: Indirect Reciprocity, Resource Sharing, and Environmental Risk: Evidence from Field 2 Experiments in Siberia 3 Short Title: Risk and Reciprocity: Field Experiments in Siberia 4 Authors: E. Lance Howe 1* , James J. Murphy 1,2,3 , Drew Gerkey 4 , Colin Thor West 5 5 1 Department of Economics and Public Policy, University of Alaska Anchorage, 3211 Providence Dr., 6 Anchorage, AK 99508; 2 Institute of State Economy, Nankai University, 94 Weijin Road, Tianjin, 7 China 300071; 3 Economic Science Institute, Chapman University, One University Drive, Orange, CA 8 92866; 4 Department of Anthropology, School of Language, Culture & Society, Oregon State 9 University, 218 Waldo Hall, Corvallis, OR, 97330; 5 Department of Anthropology, University of North 10 Carolina, 301 Alumni Building, Chapel Hill, NC, 27599 11 * Corresponding author 12 E‐mail: [email protected] 13 14

1 UAA Department of Working Paper

  • Upload
    others

  • View
    1

  • Download
    0

Embed Size (px)

Citation preview

1

UAADepartmentofEconomicsWorkingPaper1 

Title:IndirectReciprocity,ResourceSharing,andEnvironmentalRisk:EvidencefromField2 

ExperimentsinSiberia3 

ShortTitle:RiskandReciprocity:FieldExperimentsinSiberia4 

Authors:E.LanceHowe1*,JamesJ.Murphy1,2,3,DrewGerkey4,ColinThorWest55 1DepartmentofEconomicsandPublicPolicy,UniversityofAlaskaAnchorage,3211ProvidenceDr.,6 Anchorage,AK99508;2InstituteofStateEconomy,NankaiUniversity,94WeijinRoad,Tianjin,7 China300071;3EconomicScienceInstitute,ChapmanUniversity,OneUniversityDrive,Orange,CA8 92866;4DepartmentofAnthropology,SchoolofLanguage,Culture&Society,OregonState9 University,218WaldoHall,Corvallis,OR,97330;5DepartmentofAnthropology,UniversityofNorth10 Carolina,301AlumniBuilding,ChapelHill,NC,2759911 

*Correspondingauthor12 

E‐mail:[email protected] 

14 

2

Abstract1 

Integratinginformationfromexistingresearch,qualitativeethnographicinterviews,and2 

participantobservation,wedesignedafieldexperimentthatintroducesidiosyncratic3 

environmentalriskandavoluntarysharingdecisionintoastandardpublicgoodsgame.Conducted4 

withsubsistenceresourceusersinruralvillagesinremoteKamchatkaRussia,wefindevidence5 

consistentwithamodelofindirectreciprocityandlocalsocialnormsofhelpingtheneedy.When6 

experimentsallowparticipantstodevelopreputations,asisthecaseinmostsmall‐scalesocieties,7 

wefindthatsharingisincreasinglydirectedtowardindividualsexperiencinghardship,good8 

reputationsincreaseaid,andrisk‐poolingbecomesmoreeffective.Ourresultshighlightthe9 

importanceofinvestigatingsocialandecologicalfactors,beyondstrategicrisk,thataffectthe10 

balancebetweenindependenceandinterdependencewhendevelopingandtestingtheoriesof11 

cooperation.12 

13 

14 

3

Introduction1 

Researchoncooperationandcollectiveactionoftenfocusesonstrategicrisks—thecostsand2 

benefitsofcooperatingordefecting—andassociatedfree‐ridingbehaviorinasingledomain(1,2).3 

Expectedpayoffsareafunctionofindividualandgroupchoices,andinteractionsaretypically4 

limitedtoaproductionorinvestmentdomain.Previousstudieshavefoundthatrewardscan5 

mitigatethecostsofcooperation(3,4),punishmentcanincreasethecostsofdefection(5–8),6 

reputationscanfacilitatepositiveassortmentamongcooperators(9),andculturalnormsand7 

institutionscanstructureincentivesandexpectationsinwaysthatsustaincooperation(1,10).8 

Insmall‐scaleresourcedependentcommunities,cooperationcanoftenoccurinmultiple9 

domains,suchascontributingtoapublicgood,harvestingfromasharedresource,punishing10 

defectors,rewardingcooperatorsorsharingwiththosewhoexperienceahardship(11–13).These11 

domainsusuallyinteractwhichreflectsthefactthatbenefitsofcooperationcanextendbeyonda12 

singleperiod,domain,orstateofnature.Cooperationmaybepreferredtonon‐cooperation13 

preciselybecausefuturestatesofnatureareuncertaininoneormorelinkeddomains.Assuch,14 

environmentalrisk—definedasthespatialandtemporalfluctuationsinbioticandabiotic15 

componentsoftheenvironmentthataffectaccesstoresources,health,andothermeasuresof16 

humanwell‐being—couldincreaseinterdependenceand,asaresult,long‐termsuccessdepends17 

uponcooperationinmultipledomains.18 

Idiosyncraticenvironmentalriskcreatesuncertaintyaboutfuturepayoffsinacollective19 

actionproblem.Individualharvestingsuccessmaybestochastic,harvestedresourcesmayspoil,20 

animalsmaydestroystoredfood,oraninjurymaypreventtheindividualfromparticipatingin21 

collectiveaction.Insubsistencecommunities,whenanindividualexperiencesahardship,ora22 

“shock,”hisorhersurvivaldependsuponthelargesseofothers.Thus,thedecisionsaboutsharing23 

subsistenceresourcesmaydependupontheindividual’sreputationforcooperatinginother24 

4

domains.Althoughenvironmentalriskcanincreasevariationintheproductiondomain,sharing1 

amongindividualsandhouseholdscancompensatefortheseshort‐termproductiondeficits,linking2 

strategicdynamicsandcooperationacrossthetwodomains(14–17).3 

Inthispaperwepresentresultsfromaframedpublicgoodsexperiment,conductedin4 

subsistence‐dependentcommunitiesinSiberia,designedtotesthowidiosyncraticenvironmental5 

riskinteractswithstrategicrisktoaffectcooperationwithinandbetweentheproductionand6 

sharingdomains.Consistentwithamodelofindirectreciprocity, ourresultsindicatethatdecisions7 

inthesharingdomainareconditionedonreputationsforcooperationintheproductiondomain.We8 

alsofindevidenceforrisk‐pooling,asindividualssharemorewiththoseinneed.Further,when9 

reputationsforcooperationextendacrossmultiplerounds,theaidprovidedtocooperators10 

increasessubstantiallyandrisk‐poolingbecomesmoreeffective—aresultthathighlightsthe11 

importanceoflocalsocialnormswhichemphasizeresourcesharingandhelpingtheneedy(18,19).12 

However,therewardsfromsharingareinsufficienttoimprovecooperationintheproduction13 

domain.Similarly,wefindcooperationintheproductiondomainisunaffectedbyenvironmental14 

riskthatisunavoidable,consistentwiththeoreticalpredictions.15 

EnvironmentalRiskandCooperation16 

Becauseenvironmentalriskintroducesvariabilityinresourceacquisition,itcanbedifficultor17 

impossibleforasolitaryindividualtoconsistentlyacquiresufficientresourcestosurvive.Thus,18 

environmentalriskscanaffecttherelativeviabilityofindependentversuscooperativebehavior.19 

Previousresearchshowsthatenvironmentalriskaffectscooperationoverrivalrousgoodsinsmall‐20 

scale,resource‐dependentcommunities.Intheoreticalstudies,environmentalriskoruncertainty21 

canincreaseordecreasecooperationinsocialdilemmas(20,21).Experimentalstudiesgenerally22 

findthatincreasingthevariabilityofreturnstoeitherthegrouporprivateaccountreduces23 

5

cooperationintheriskierdomain(22–24);incontrasttheidiosyncraticshockinourdesignis1 

unavoidable,itaffectsboththegroupandprivateaccountsequally.2 

Althoughenvironmentalriskshavereceivedrelativelylessattentioninresearchon3 

cooperationandcollectiveaction,theoreticalandempiricalstudiesofrisk‐poolinginanthropology4 

(25,26)andeconomics(27,28)haveexploredinteractionsbetweenstrategicandenvironmental5 

risks.Smith(29)suggestsrisk‐poolingislikelytooccurwhenanindividual’ssuccessinresource6 

acquisitionexhibitsstochasticvariationthatisasynchronousamongindividuals,creating7 

opportunitiesforindividualstoreduceenvironmentalriskbysharingresources.Relatedeconomic8 

studieshaveidentifiedtheuseofnon‐marketmechanisms—includinginformalloans,remittances,9 

andsocialnetworks—topoolriskandminimizethenegativeeffectsofconsumptionvariability(16,10 

30,31).11 

Smoothingconsumptionbypoolingresourcescanenhanceoddsofsurvival,butthisformof12 

cooperationentailsstrategicriskinboththeproductionandsharingdomains.Assuch,the13 

insuranceprovidedbypoolingresourcespresentsanothersocialdilemmasincefree‐ridersmay14 

benefitwithoutcontributing,underminingthelong‐termstabilityofrisk‐pooling.Related15 

theoreticalmodelshaveshownastrongcommitmentdeviceisneededtofacilitateeffectiverisk‐16 

pooling,insuringthelong‐termbenefitsofparticipationexceedtheshort‐termgainsofleavinga17 

network(32).18 

Experimentalresearchechoestheresultsofthesemodels.Studieshaveexplored19 

commitmentinthecontextofendogenousgroupformation(33,34).Forexample,BarrandGenicot20 

(33)foundindividualsweremostlikelytoformrisk‐poolinggroupsinthepresenceofastrong,21 

exogenouslyenforced,commitmentdevice.CharnessandGenicot(35)findstrongevidenceforrisk‐22 

poolingwithalimitedcommitmentdevice;however,directreciprocityisacentralfeatureoftheir23 

6

incentivestructure.Incontrast,directreciprocityisnotanexplicitfeatureofourdesign,which1 

allowsustotestforrisksharingintheabsenceofcommitmentdevices.2 

Ourstudycomplementsexistingresearchbyintegratinginsightsonrisk‐poolingwithmore3 

generaltheoreticalandempiricalresearchoncooperation.Specifically,wecombinefactorsthat4 

increaseinterdependenceandencouragerisk‐pooling—stochasticresourceacquisitionand5 

voluntaryresourcesharing—withfactorsthatamplifystrategicrisksofdefection—rewardsand6 

reputations.Weutilizemethodologicaltoolsfromanthropologyandeconomicstodesignaseriesof7 

fieldexperimentsinvolving136participantsfrom3villageslocatedontheKamchatkaPeninsulain8 

NortheastSiberia.PeoplelivinginKamchatka’sremotevillagesmustcontinuallycopewith9 

strategicandenvironmentalrisks,withlimitedsupportfromformalinstitutions(36).Priortothe10 

experiments,weconductedqualitativeethnographicinterviewsandparticipantobservationto11 

identifytheparticularstrategicandenvironmentalrisksthatpeopleinKamchatkaface.These12 

insightsinformedthedesignofourexperiments.13 

ResearchDesign14 

OurfieldexperimentswereconductedinthreesmallcommunitiesintheKaraginskiiregionof15 

KamchatkaoverafourdayperiodineachcommunityduringSpring2011.Thisisalarge,remote16 

region(40,600km2)withasmallpopulation(4,824people)thatisdependentuponharvestinglocal17 

resourcesforsubsistence.Approximately85%ofexperimentparticipantswereindigenousandhad18 

livedintheareaformostoftheirlives.19 

Subjectswererecruitedthroughbulletinboardannouncements,door‐to‐doorvisits,andby20 

alocalcommunitycoordinator.ExperimentswereconductedinRussianandallsupporting21 

materialswerepresentedinRussian.Participantsreadaconsentformpriortothestartofthe22 

experimentandprovidedverbalaffirmationofinformedconsentpriortoparticipation.Signatures23 

werenotcollectedsinceourstudywasdeterminedtobeofminimalrisk,subjectsexperiencedrisk24 

7

similartothatencounteredineverydaylife,andsignatureswouldhaveunnecessarilylinked1 

subjectstothestudy.Investigatorcontactinformationwasprovidedtosubjectsandleftwithvillage2 

mayorsandcommunitycoordinators.Researchersreturnedtocommunitiestwoyearslaterto3 

reportrelatedresearchfindingstoparticipantsandcommunitymembers.Ourstudyandconsent4 

procedureswereapprovedbytheUAAInstitutionalReviewBoard(projectid#216266).The5 

protocolwaspre‐testedwithnativeRussian‐speakingstudentsattheUniversityofAlaska6 

Anchorage.InstructionswerereadaloudandaccompaniedbyPowerPointslidesprojectedontoa7 

screen.InstructionsinEnglishandRussian,fieldprotocol,andanimageofinformationdisplayedto8 

subjects,canbefoundatthedatareviewurlincludedinoursubmission.9 

Eachsessionlastedapproximatelythreehours,duringwhichsubjectsplayedamodified10 

versionofalinearpublicgoodsgame.Experimentswerehand‐run,withtheaidofasinglelaptop11 

computerandaprojector.Foreachround,decisionswerewrittenonslipsofpaper,collectedby12 

oneoftheexperimenters,andenteredintoaspreadsheet.Resultswereprojectedontothescreen,13 

andsubjectswrotetheoutcomesonarecordsheet.Oncethisprocesswascompleted,another14 

roundfollowed.15 

Subjectswererandomlyassignedtooneoftwofive‐persongroups.In4ofthesessions16 

therewasone5persongroupandone4persongroup.Intheremaining10sessionstherewere217 

fivepersongroups.Inalltreatments,individualswereidentifiedbyaletterknownonlybythe18 

individualandtheexperimenter.Thus,althoughparticipantsknewthecompositionofeachgroup,19 

therewasnowayforothergroupmemberstolinkanindividualtohisorherdecisions.Moreover,20 

withoneexception(describedlater),eachindividual’sletterrandomlyvariedeveryround.This21 

methodeliminatedthepossibilityofusinginformationaboutaparticulargroupmember’sactions22 

inpriorroundsandpreventedindividualsfromdevelopingreputations.23 

8

Subjectswerepaidincash,withaverageearningsof610rubles(about$22USdollarsatthe1 

time),equivalenttoatypicaldailywage.Inaddition,allparticipantsreceiveda200rubleshow‐up2 

payment.3 

Themodifiedpublicgoodsgamewasframedasteamsubsistenceproduction(37,38),and4 

consistedoftwostages.Stage1wasidenticalforallgroupsandconsistedofastandardlinear5 

publicgoodsgameforfiverounds.ThisservesasourBaselineTreatment.Eachround,every6 

individualstartedwithaninitialendowmentof50“hours”whichhadtobeallocatedbetweenan7 

individualandagroupactivity.Theactivitywasframedas“fishing,hunting,orcollecting8 

mushroomsandberries…”where“sometimesyoudotheseactivitiesonyourown”(theindividual9 

productionactivity)but“sometimesyoudothemwithotherpeople”(thegroupproduction10 

activity).Eachhourallocatedtotheindividualactivityyieldedaprivatereturnof10rubles.Time11 

allocatedtothegroupactivityyielded20rublesperhour,because“peopleoftengetmoredone12 

whenworkingtogether.”Returnsfromthegroupactivityweredividedequallyamongallgroup13 

members,regardlessofthetimeallocated.Attheendofeachround,theallocationdecisionofeach14 

groupmemberwaspubliclyrevealed(identifiedonlybyaletter).15 

Stage2consistedofeightadditionalroundsunderoneoffoursharingtreatments.16 

Treatmentsvariedintermsofthepresenceofenvironmentalriskandincentivestocooperateas17 

determinedbytheinformationavailabletosubjectswhenmakingdecisions.Inalltreatments,18 

subjectsfirstmadethesametimeallocationdecisionasinStage1.Afterthedecisionsweremade,19 

someinformationwasrevealed,thensubjectsweregiventheopportunitytosharerubleswith20 

othergroupmembers.Theinstructionsemphasizedthevoluntarynatureofsharingandusedthe21 

Russianverbpodelit’sya(“toshare”).Therewasnorestrictiononthenumberoffellowgroup22 

memberswithwhomanindividualcouldshare.Toavoidsharingcommitmentsinexcessofan23 

individual’searnings,thetotalamountsharedbyanindividualwaslimitedto250rubles.Table1.24 

summarizeskeyinformationforeachtreatment.25 

9

Table1.ExperimentalDesign1 

RiskandSharingTreatment InformationRevealedPriortoSharingDecision

Treatment NIdiosyncratic

RiskVoluntarySharing

PlayerShocked

Allocationdecisionsincurrentround

Alldecisionsinpriorrounds

Baseline 136 no no ‐‐ ‐‐ ‐‐

Reward 40 no yes ‐‐ yes no

Shock 29 yes yes yes no no

NoReputation 38 yes yes yes yes no

WithReputation 29 yes yes yes yes yes

 2 

RewardTreatment.ThefirsttreatmentwasidenticaltotheBaselineexceptthataftertime3 

allocationdecisionswerecompletedandpubliclyrevealed,subjectsmadeasharingdecision.4 

Becauseindividualtimeallocationdecisionswerecommonknowledge,subjectscouldusesharing5 

asamechanismtorewardothersforcontributingtothegroupactivityinthecurrentperiodorto6 

indirectlypunishnon‐cooperatorsbywithholdingsharing,increasingthecostofdefectionrelative7 

totheBaselineTreatment.Becausesharingwasazero‐sumtransfer,ithadnoimpactongroup8 

earnings.Aftersharingdecisionswerecollected,theamountssharedandreceivedwererevealedto9 

thegroup.NotethatinallStage2treatments,onlyaggregatesharingoutcomeswererevealed;the10 

amounttransferredbetweentwoparticularplayerswasnotdisclosed.Thistreatmentissimilarto11 

theRewardTreatmentinSefton,ShuppandWalker(39).12 

RiskTreatment.Thesecondtreatmentintroducedidiosyncraticenvironmentalrisk.Afterthetime13 

allocationdecisionsweremade,butbeforetheywererevealed,oneindividualfromeachgroupwas14 

randomlyselectedbytherollofadietoincura“shock”whichwasdescribedas“notcatchingany15 

fish,gettingsick,orhavingallthefoodyou’vegatheredspoil.”Theindividualwhoincurredthe16 

shocklostallearningsfromboththegroupandindividualactivities.Onlytheamountreceivedfrom17 

voluntarysharingbyothersdeterminedtheindividual’searningsforthatround.Theletterofthe18 

individualincurringtheshockwasannouncedtothegrouppriortothesharingdecision.Afterthe19 

10

sharingdecisionswerecollected,boththetimeallocationandsharingdecisionsofallgroup1 

memberswererevealed.2 

Risk/RewardTreatment.ThethirdtreatmentwasidenticaltotheRiskTreatment,exceptthat3 

priortothesharingdecision,boththeletteroftheindividualshockedandtheallocationdecisions4 

ofallgroupmemberswererevealed.Thisallowedsharingtobebasedonwhetheranindividual5 

wasshockedand/ortheindividual’stimeallocationinthecurrentperiod.Afterthesharing6 

decisionswerecollected,theindividualsharingandtimeallocationdecisionswererevealedtothe7 

group.8 

Risk/Reward/ReputationTreatment.Inthefinaltreatment(whichwewillrefertoasthe9 

ReputationTreatment),individualplayerletterswereconstantacrossrounds,butotherwise10 

followedthesamerulesastheRisk/RewardTreatment.Holdingplayerlettersconstantcreatedan11 

opportunityforindividualstodevelopareputationforcooperativebehaviornotonlyinthe12 

productiondomain,butalsothesharingdomain.Thisallowedothergroupmemberstocondition13 

sharingonthesereputations.TheReputationTreatmentbringstheexperimentclosertonaturally14 

occurringcontextsofcooperationinsmall‐scalesocieties,whereindividualshaveaccesstoand15 

utilizereputations.16 

Individualcashearningsweredeterminedbyasingleroundthatwasrandomlyselectedby17 

adierollattheendoftheexperiment(13,35).Selectingasingleroundeliminatedthepossibility18 

forsubjectstopoolearningsovertime,whichwouldhavebeenanalogoustoindividuallyinsuring19 

againstshocks.OurdesignchoiceparallelsfieldconditionsinnorthernKamchatkawherethereis20 

substantialseasonalvariationinweatherandresourceavailabilityanditisdifficulttoself‐insure21 

againstshockstosubsistenceharvests.22 

11

RelatedStudiesandHypotheses1 

ThedesignofourexperimentismostsimilartoacomputerizedlaboratoryexperimentbyCherry,2 

Howe,andMurphy(13)butdiffersintermsoftheframing,thesourceoftheshock,thenatureof3 

sharing,andtheamountofinformationrevealed.Theyfindstrongevidenceforrisk‐pooling4 

withoutacommitmentdevice.Incontrast,ourdesignintroducesunavoidableidiosyncraticriskand5 

allowsustotesttheeffectofreputationsonsharingandcooperationdecisions.6 

Ineachofourtreatments,thestaticNashequilibriumallocationstothegroupactivityand7 

tosharingarebothzero.Further,becausedirectreciprocitywasnotpossibleinourgame8 

environmentgivenactualdecisions,sharingarrangementsarenotself‐enforcing;thatis,the9 

expectedfutureindividualgainfromcooperatingbysharingdoesnotexceedthecurrentbenefitof10 

defecting.EssentialfeaturesofthisdecisionenvironmenthavebeenmodeledbyNowakand11 

Sigmund(9)whoexplorecooperationviaindirectreciprocity.Inourdecisionenvironment,indirect12 

reciprocityisdefinedasthesharinggiventoanindividualthatisconditionedontheobserved13 

cooperationofthatindividualwithothergroupmembersinboththeproductionandsharing14 

domainswhenpossible(9,40).15 

Agrowingnumberofexperimentalstudiesprovidesupportfortheimportanceof16 

reputationandtheroleofindirectreciprocityincooperationandcollectiveaction(4,9,40–42).In17 

thecontextoftwolinkedcooperativedomains,Panchanathan&Boyd(43)suggestindirect18 

reciprocitydependsontwoconditions:1)reputationsformedbyactionsinthefirstdomain19 

increasebenefitsreceivedintheseconddomainand2)thebenefitsofagoodreputationinthe20 

seconddomainexceedthecostsofcooperationinthefirstdomain.Weinvestigatehow21 

environmentalriskaffectsthesestrategicdynamicsofreputationandindirectreciprocity.22 

BycomparingdecisionsacrossthefourtreatmentsinStage2,wecantesttheextentto23 

whichtimeallocationandsharingdecisionsareinterlinkedandhowtheyrespondtorisk.Basedon24 

12

thePanchanathan&Boyd(43)modelofindirectreciprocity,wehypothesizethatsharingdecisions1 

willbeconditionedonobservablebehavior,andpeoplewhoexhibitmorecooperationinthese2 

domainswillreceivemoresupport.Thisimpliesthat intheRewardTreatment,subjectswilluse3 

sharingtorewardcooperationandwillpunishnon‐cooperatorsbywithholdingsharing(H1).In4 

theRiskTreatment,thoseexperiencingahardshipwillreceiveadditionalsupport,butitwillbe5 

independentofallocationandsharingdecisionsbecausetheseareunobservable(H2).Inthe6 

Risk/RewardTreatment,sharingwillbedirectedtowardstheindividualwhowasshockedand7 

sharingwillincreasewiththeshockedindividual’sgroupallocationdecisioninthecurrentround8 

(H3).IntheReputationTreatment,theamountsharedwithashockedindividualshouldincrease9 

withbothhisorherallocationdecisioninthecurrentperiodandhisorhersharingdecisioninthe10 

previousperiod(H4).Finally,ifweobservesharingthatisconditionedonallocationstothegroup11 

activityinthefinaltwotreatments,thenweexpecttheamountoftimeallocatedtothegroup12 

activitywillbegreaterthanintheRisktreatment,whichdoesnotfacilitateconditionalsharing13 

(H5).14 

Results15 

Sharing.IntheRewardTreatment,theaverageamountreceivedfromsharingwas96rubles.Inthe16 

threetreatmentswithidiosyncraticrisk,theaverageamountreceivedwasnotsubstantially17 

different;however,sharingwasoverwhelminglydirectedtowardthoseexperiencingahardship.18 

Moreover,themoreashockedindividualcooperatedintheproductiondomain,themoreheorshe19 

receivedfromsharing.WeexplorethisresultwithfourrandomeffectsregressionmodelsinTable20 

2,oneregressionforeachofthefourStage2treatments.Themodelsallusethesamebasic21 

structure:Yit = 0 + 1it + 2t + i + it,whereinTable2.Yitisthetotalamountreceivedin22 

sharingbysubjectiinroundt[6,13],itisasetofindependentvariablesthatcontrolforwhether23 

eachindividualwasshocked,theamountsharedinthepreviousperiod,theamountallocatedtothe24 

13

groupactivityinthecurrentperiod,andinteractionsofthesevariables,icapturesunobserved1 

individualsubjectcharacteristicsanditrepresentsthecontemporaneouserrorterm.Because2 

subjectsparticipatedinmultipleroundsofasingletreatment,subject‐specificheterogeneityis3 

modeledasarandomeffect.WeuseaHuber(44)andWhite(45)robustestimateofvariance.4 

Table2.IndividualAmountReceivedfromSharing(Stage2)5 

Reward Risk Risk/Reward ReputationAmountSharedt–1 0.09(0.06) 0.14(0.12) 0.07(0.06) 0.03(0.07)GroupActivityt 1.15***(0.39) ‐0.31(0.25) ‐0.25(0.36) ‐0.22(0.52)Shockedt 108.24*** (24.53) 82.18** (32.99) ‐40.70(42.69)ShockedtXAmountSharedt–1 ‐0.14(0.19) ‐0.19(0.13) 1.16**(0.55)ShockedtXGroupActivityt 2.48(1.51) 2.28** (1.14) 5.66***(1.73)Periodt ‐0.92(1.43) ‐1.18(1.76) ‐2.78** (1.16) ‐2.30(1.87)Constant 72.71***(22.64) 45.60** (22.27) 72.87*** (16.62) 73.13**(33.38)N 280 161 210 161Robuststandarderrorsinparenthesesareclusteredatthegroup‐level.Dependentvariableisthe6 

amountreceivedinsharing.Statisticalsignificance:***:p<0.01;**:p<0.057 

ThefirstmodelshowsresultsfortheRewardTreatment,whichdoesnotincludeashock8 

andthereforerelatedvariablesarenotincluded.ConsistentwithH1,theGroup_Activitytcoefficient9 

ispositiveandstatisticallysignificant.Conversely,whethertheindividualsharedresourcesinthe10 

previousroundisunknownand,asexpected,theAmount_Sharedt–1variableisnotsignificant.Thus,11 

consistentwithSefton,Shupp,andWalker(39),individualsusedthesharingmechanismtoreward12 

cooperativebehaviorinthegroupactivitydecision.However,themagnitudeoftheeffectis13 

relativelymodest.Eachhourallocatedtothegroupactivityyieldedareturnof5.15rubles—1.1514 

receivedfromsharingplus4rublesfromthegroupactivity—whichisonlyabouthalfofthe1015 

rublereturnfromanhourallocatedtotheindividualactivity.Recallthateveryhourallocatedtothe16 

groupactivityyields4rublesfortheindividual,aswellaseachoftheothergroupmembers(2017 

rublesperhourwhichisevenlydividedamongallfivegroupmembers).18 

Thenextthreemodelsincludeinteractionsofwhethertheindividualwasshocked,the19 

allocationtothegroupactivityinthecurrentround,andtheamountsharedinthepreviousround.20 

14

Inalltreatments,individualsconditionedsharingdecisionsontheavailableinformation,and1 

unavailableinformationisnotsignificant,asexpected.IntheRiskTreatment,participantsreceived2 

anaverageof108.24rublesjustforincurringashock,supportingthehypothesis(H2)thatpeople3 

usesharingtoassistthoseinneed.Withmoreinformationaboutothergroupmembers’behaviorin4 

theRisk/RewardandReputationTreatments,sharingwasstilldirectedtowardthoseinneed,and5 

theamountreceivedincreasedforthoseindividualswithhigherlevelsofcooperationinthegroup6 

activity(Figure1.).IntheRisk/Rewardtreatment,thosewhoexperiencedashockcontinuedto7 

receivesomesupportthatwasindependentoftheiractions(82.18rubles),butsubjectsreceivinga8 

shockalsoreceivedanadditional2.28rublesperhourallocatedtothegroupactivity(consistent9 

withH3).IntheReputationtreatment,behaviorinboththecurrentandpreviousroundswas10 

commonknowledge.InsupportofH4,sharingwasconditionedonboththeshockedindividual’s11 

mostrecentsharingdecision(periodt–1)andthemostrecentgroupactivitydecision(periodt).12 

Foreachrubleshared,shockedplayersreceived1.16rublesinsharing.Foreachhourallocatedto13 

thegroupactivity,shockedplayersreceived5.66rublesinsharing.InboththeRisk/Rewardand14 

ReputationTreatments,sharingwasnotusedtorewardcooperationindependentoftheshock.15 

Instead,sharingwasdirectedonlytowardthoseinneedandwasconditionedontheircooperation. 16 

15

Figure1.Predictedindividualamountreceivedinsharing,conditionedonwhetherthe2 individualreceivedashock,usingcoefficientsinTable2.3 

GroupActivity.Inthefirstfiverounds(Stage1),allgroupsparticipatedintheBaselineTreatment.4 

InStage1,averageallocationstothegroupactivity(about40%oftheinitialendowment)were5 

consistentwithresultsfromotherlinearpublicgoodsgames(2,46).Table3presentstheresultsof6 

tworandomeffectsmodelsforthegroupactivitydecisioninStage2(rounds6‐13only).Inthese7 

regressionsthedependentvariableYitistheindividualallocationtothegroupactivityofsubjectiin8 

roundt.Model2addsindividualcharacteristicsandfixedeffectsforthecommunities.Toprotect9 

subjectconfidentialityandtomakedatapublicallyavailableforreplicationwehavenotidentified10 

specificcommunities,gender,orraceintheregressionresults.Weexploitthewithin‐subjectdesign11 

byusingtheindividual’saveragegroupallocationoverallfiveroundsofStage1Baselineasan12 

independentvariable(BaselineGroupActivity).Sincetheidiosyncraticenvironmentalriskwas13 

unavoidable,cooperationintheRiskTreatmentshouldbeunaffectedbyrisk,whichisprecisely14 

whattheresultsinTable3suggest.However,contrarytoH5,theabilitytosharefailedtoincrease15 

01

002

003

004

005

00A

mo

unt r

ece

ive

d fr

om s

harin

g

0 10 20 30 40 50Allocation to the group activity

Risk/Reward (Not Shocked) Reputation (Not Shocked)Risk/Reward (Shocked) Reputation (Shocked)

16

cooperationinboththeRisk/RewardandReputationtreatments.Althoughresultsindicatesharing1 

withthoseexperiencingtheshockisconditionedontheindividual’sallocationtothegroupactivity2 

(Table2),thelevelsofsharingareinsufficienttoinduceanincreaseincooperation.Ifaperson3 

receivesashockintheReputationTreatment,thereturnfromanhourallocatedtothegroup4 

activitywas9.66rubles(5.66asrewardforanallocationviasharingplus4fromthegroup5 

activity),whichisstilllowerthantheperhourreturnof10rublesfromtheindividualactivity.6 

17

Table3.IndividualAllocationtoGroupActivity(Stage2only,rounds6‐13)1 

Model1 Model2

RewardTreatment omitted omitted

RiskTreatment ‐1.642(1.76) ‐1.849(1.60)

Risk/RewardTreatment ‐0.961(1.74) ‐0.947(1.77)

ReputationTreatment 0.380(2.86) 0.642(2.54)

Round ‐0.095(0.16) ‐0.095(0.16)

BaselineGroupActivity 0.775***(0.10) 0.754***(0.11)

Gender1 ‐1.525(1.52)

Age 0.099**(0.04)

Race1 ‐0.206(2.07)

Community1 ‐3.427**(1.42)

Community2 ‐1.595(1.73)

Community3 omitted

Constant 3.616(2.70) 2.375(2.93)

N 1088 1072

Robuststandarderrorsareclusteredatthegroup‐level.BaselineGroupActivityisthemeanofthe2 

individual’sdecisionsintheStage1Baselinetreatment(rounds1‐5).Dependentvariableisthe3 

individualallocationtothegroupactivity.Statisticalsignificance:***:p<0.01;**:p<0.054 

Thus,wefindsomesupportforPanchanathan&Boyd’s(43)modelofindirectreciprocity.5 

Individualsinneeddoreceivesubstantialsupport,and,whenpossible,thissupportisconditioned6 

ontheirreputationsforcooperation.However,thebenefitsfromapositivereputationdidnot7 

exceedthecostsofparticipatinginthegroupactivity,andasaresult,theabilitytosharedidnot8 

increasecooperation.9 

18

Discussion1 

Wesystematicallyexaminedtheinteractionsofstrategicandenvironmentalrisksamongpeoplein2 

Kamchatkawhofacethesechallengesrepeatedlyinthepost‐Sovietera(47).Introducing3 

idiosyncraticenvironmentalriskinthesocialdilemmaincreasedinterdependence,andpeople4 

respondedbychannelingresourcestothoseinneed,rewardingindividualsforcooperation,and5 

punishingindividualswhodidnotcooperate.Theabilitytoshareasatooltomitigate6 

environmentalriskincreasedtheinterdependenceamonggroupmembers.Asaresult,highlevels7 

ofsharingwereachievedwithoutdirectreciprocityorastrongcommitmentdevice.Observed8 

sharingis,however,consistentwithlocalsharingnorms.Wefindstrongevidenceforsharing,even9 

withoutreputations,whichisconsistentwithamodelofpro‐socialbehavior(andrelated10 

experimentalresults)inwhichpreferencesforkeepingsocialrulesarethedrivingforcebehind11 

pro‐socialbehavior(48).12 

Whencurrentorpastbehaviorwasobservable,sharingwasconditionedonobserved13 

cooperativebehavior.IntheRewardTreatment,individualswhoparticipatedmoreinthegroup14 

activityreceivedmorefromsharing,consistentwithpreviousstudiesthatemphasizethe15 

importanceofrewards,punishments,andreputationsfortheemergenceofcooperation(42,49).16 

Thepositiverelationshipidentifiedbetweensharingandallocationstothepublicgoodinthe17 

Risk/RewardandReputationTreatmentssuggeststhatwhenbothstrategicandenvironmental18 

risksarepresentinasocialdilemma,theeffectsofstrategicrisksdependonenvironmentalrisks.19 

Theseresultshaveimportantimplicationsforresearchonrisk‐pooling,theroleofreputations,20 

rewards,andpunishmentsintheoriesofcooperation,andmoregenerally,theroleof21 

environmentalvariabilityinhumanadaptationandresilience.22 

Ethnographicresearchonrisk‐poolingemphasizestheimportanceofsupportingthosein23 

needandmechanismsofreputationtomaintaincooperation(25,50).Labexperimentsinspiredby24 

19

thisresearchhavedemonstratedthathigh‐varianceresourcesandreputationscanplayakeyrole1 

intheemergenceofrisk‐pooling,dramaticallyincreasingreciprocalexchangesamongindividuals2 

relativetolow‐varianceresources(51)andthatrisk‐poolingstrategiescanincreaseindividualand3 

pair‐wisesurvivalinenvironmentswithhighdegreesofrisk(52).Similarly,agent‐based4 

simulationshaveshownincreasedenvironmentalharshness—whichcanbemitigatedvia5 

cooperation—canamplifycooperation(53).Eachofthesestudiesemphasizestheimpactof6 

interdependenceontheemergenceofcooperation.Wecontributetothisworkbydemonstrating7 

howasymmetriesofneedcausedbystochasticenvironmentalrisksor“shocks”interactwiththe8 

strategicriskstiedtorewards,punishments,andreputationstoincreaseinterdependenceand9 

enhancerisk‐pooling.InboththeRisk/RewardandReputationTreatments,individualswho10 

contributemoretothepublicgoodreceivemoreviasharing,butonlywhentheysufferashock.11 

Theseinteractionsbetweenstrategicandenvironmentalriskssuggeststrategicrisksremain12 

importantforpreciselythoseindividualswhobenefitmostfromrisk‐pooling,discouraging13 

defectorsandfree‐riders.Indeed,wefoundtheeffectivenessofrisk‐poolingincreasedwhenpeople14 

hadtheabilitytomonitorandactuponreputationsacrossmultiplerounds.Whileprevious15 

researchhasemphasizedtheimportanceofexogenouscommitmentdevices,formalinstitutions,16 

endogenousgroup‐formation,anddirectreciprocityforeffectiverisk‐pooling,ourexperiments17 

showthatrisk‐poolingcanemergefromendogenousreputationdynamicsandindirectreciprocity.18 

Althoughtheinteractionofstrategicandenvironmentalriskenhancedtheeffectivenessof19 

risk‐pooling,wedidnotobservesystematicincreasesinthegroupactivityreportedbyprevious20 

studieswhererewardsareofferedinthecontextofasocialdilemma(5,39,42).Oneexplanationis21 

thatthebenefitsofgoodreputationsforcooperatorsneverexceedthecostsofcontributingtothe22 

publicgood.Previousstudieswithasimilartwo‐dilemmadesignamplifytheimpactofreputations23 

byincreasingtherelativecostsandbenefits(i.e.efficiency)ofrewardsand/orpunishments,often24 

withratiosashighas1:3(4–7,54).Thus,increasinglevelsofcooperationobservedinprevious25 

20

experimentsmaynotbeduetothepresenceorabsenceofrewardsandpunishmentsperse,butthe1 

presenceofhighlyefficientrewardsandpunishments(55–57).Whilehighlyefficient2 

reward/punishmentmechanismshavebeenshowntoincreaselevelsofcooperationin3 

experiments,itislessclearhowoftensuchmechanismsareavailableinnaturallyoccurring4 

contextsofcooperation(58).Indeed,thewayparticipantsconditionaidtoneedyplayersbasedon5 

cooperationreflectslocalnormsofindirectpunishment,whicharemorecommonlyobservedinour6 

studyregionthannormsofdirect,individualcostlypunishment.7 

Inadditiontoaddressingindividualstrategicbehavior,ourstudyhighlightstheimportant8 

roleoffactorsthatincreaseinterdependenceamongindividuals.Weinvestigatedonefactor—9 

stochasticresourceacquisition—thatincreasesinterdependencebycreatingconsumptiondeficits10 

thatcanbeovercomebypoolingresourcesthroughsharing.Suchdeficitsmightalsoarisefrom11 

differencesinindividual/householdproductivecapacityandconsumptiveneeds(59)orstochastic12 

differencesinharvestsduetopoorhealthorothermisfortunes(60,61).Ourexperiments13 

incorporateconsumptiondeficitsviastochasticshocks,providingaspecificfactorforamplifying14 

theimpactofreputationsrelativetothehighlyefficientrewardandpunishmentmechanisms15 

utilizedinpreviousstudies.16 

Scholarsstudyingprocessesofcontemporaryhumanadaptationtounprecedentedforcesof17 

globalclimatic,economic,political,andculturalchangehaveemphasizedthecrucialroleof18 

strategiesthatmitigateenvironmentalrisks(62).Manycomponentsofcontemporaryadaptation—19 

includingtheroleoftraditionalecologicalknowledge,socialnetworks,institutions,andotherforms20 

ofsocialcapital—dependoncooperationamongindividualstomaintainresilienceinthefaceof21 

shocksandperturbations(63).Therefore,understandinghowenvironmentalrisksinteractwith22 

strategicriskstoaffecttheemergenceandstabilityofcooperationcanimproveourattemptsto23 

adapttothechallengeswefaceincontemporaryenvironments.Ourresearchsuggeststheoriesof24 

21

cooperationcancontributetothisgoalbyinvestigatingabroaderrangeoffactorsthatincrease1 

interdependence.2 

Acknowledgements3 

ThisresearchwassupportedbytheNationalScienceFoundation(#0729063and#1019303).DG4 

wassupportedbyapost‐doctoralfellowshipfromtheNationalSocio‐EnvironmentalSynthesis5 

Center(SESYNC)underfundingfromtheNationalScienceFoundation(DBI‐1052875).Wewishto6 

thankallthepeopleinKamchatkawhoparticipatedinourresearch,aswellasDr.Viktoria7 

Petrasheva,TatianaDegai,andDanielAllenforresearchassistance.WeespeciallythankCristina8 

GainaBlantonwhoassistedintranslations,transcriptions,andconductingfieldexperiments.We9 

thankMarthaMadsenatExploreKamchatkaforoutstandinglogisticssupport. 10 

22

References1 

1. OstromE,GardnerRJ,WalkerJM(1994)Rules,Games,&Common‐PoolResources(University2 ofMichiganPress,AnnArbor).3 

2. ChaudhuriA(2010)Sustainingcooperationinlaboratorypublicgoodsexperiments:a4 selectivesurveyoftheliterature.ExpEcon14:47–83.5 

3. DreberA,RandDG,FudenbergD,NowakMA(2008)Winnersdon’tpunish.Nature452:348–6 51.7 

4. RandDG,DreberA,EllingsenT,FudenbergD,NowakMA(2009)Positiveinteractions8 promotepubliccooperation.Science325:1272–5.9 

5. FehrE,GächterS(2000)Cooperationandpunishmentinpublicgoodsexperiments.AmEcon10 Rev90:980–994.11 

6. FehrE,GächterS(2002)Altruisticpunishmentinhumans.Nature415:137–140.12 

7. HenrichJetal.(2006)Costlypunishmentacrosshumansocieties.Science312:1767–70.13 

8. GurerkO,IrlenbuschB,RockenbachB(2006)Thecompetitiveadvantageofsanctioning14 institutions.Science312:108–111.15 

9. NowakMA,SigmundK(2005)Evolutionofindirectreciprocity.Nature437:1291–8.16 

10. OstromE(2005)UnderstandingInstitutionalDiversity(PrincetonUniversityPress,17 Princeton).18 

11. JanssenMA,AnderiesJM,CardenasJ‐C(2011)Head‐endersasstationarybanditsin19 asymmetriccommons:Comparingirrigationexperimentsinthelaboratoryandthefield.Ecol20 Econ70:1590–1598.21 

12. SchottS,BuckleyNJ,MestelmanS,MullerRA(2007)Outputsharinginpartnershipsasa22 commonpoolresourcemanagementinstrument.EnvironResourEcon37:697–711.23 

13. CherryTL,HoweEL,MurphyJJ(2015)SharingasRiskPoolinginaSocialDilemma24 Experiment.EcolSoc20:68.25 

14. Fienup‐RiordanA(1986)WhenOurBadSeasonComes:ACulturalAccountofSubsistence26 Harvesting&HarvestDisruptionontheYukonDelta(AlaskaAnthropologicalAssn,27 Anchorage,AK).28 

15. AlvardMS,NolinDA(2002)Rousseau’swhalehunt ?CurrAnthropol43:533–559.29 

16. FafchampsM,LundS(2003)Risk‐sharingnetworksinruralPhilippines.JDevEcon71:261–30 287.31 

23

17. FafchampsM,GubertF(2007)Theformationofrisksharingnetworks.JDevEcon83:326–1 350.2 

18. FehrE,FischbacherU(2004)Socialnormsandhumancooperation.TrendsCognSci8:185–3 90.4 

19. BielA,ThøgersenJ(2007)Activationofsocialnormsinsocialdilemmas:Areviewofthe5 evidenceandreflectionsontheimplicationsforenvironmentalbehaviour.JEconPsychol6 28:93–112.7 

20. BoucherV,BramoulléY(2010)Providingglobalpublicgoodsunderuncertainty.JPublic8 Econ94:591–603.9 

21. BramoulleY,TreichN(2009)Canuncertaintyalleviatethecommonsproblem?JEurEcon10 Assoc7:1042–1067.11 

22. RapoportA,SuleimanR(1992)Equilibriumsolutionsforresourcedilemmas.GrDecisNegot12 1:264–294.13 

23. WitA,WilkeH(1998)Publicgoodprovisionunderenvironmentalandsocialuncertainty.14 EurJSocPsychol28:249–256.15 

24. GangadharanL,NemesV(2009)Experimentalanalysisofriskanduncertaintyin16 provisioningprivateandpublicgoods.EconInq47:146–164.17 

25. CashdanEed.(1990)RiskandUncertaintyinTribalandPeasantEconomies(WestviewPress,18 Boulder).19 

26. WiessnerP(1982)inPoliticsandHistoryinBandSocieties,edsLeacockE,LeeR(Cambridge20 UniversityPress,Cambridge,UK),pp61–84.21 

27. FafchampsM(2003)RuralPoverty,RiskandDevelopment(EdwardElgarPublishing22 Limited).23 

28. TownsendRM(1994)RiskandInsuranceinVillageIndia.Econometrica62:539.24 

29. SmithEA(1988)inHuntersandGatherers:History,Evolution,andSocialChange,edsIngold25 T,RichesD,WoodburnJ(Berg,Oxford),pp222–252.26 

30. UdryC(1994)Riskandinsuranceinaruralcreditmarket:anempiricalinvestigationin27 NorthernNigeria.RevEconStud61:495–526.28 

31. DeWeerdtJ,DerconS(2006)Risk‐sharingnetworksandinsuranceagainstillness.JDev29 Econ81:337–356.30 

32. GenicotG,RayD(2003)Groupformationinrisk‐sharingarrangements.RevEconStud31 70:87–113.32 

24

33. BarrA,GenicotG(2008)Risksharing,commitment,andinformation:anexperimental1 analysis.JEurEconAssoc6:1151–1185.2 

34. BarrA,DekkerM,FafchampsM(2012)Bridgingthegenderdivide:anexperimentalanalysis3 ofgroupformationinAfricanvillages.WorldDev40:2063–2077.4 

35. CharnessG,GenicotG(2009)Informalrisksharinginaninfinite‐horizonexperiment.EconJ5 119:796–825.6 

36. GerkeyD(2011)Abandoningfish:Thevulnerabilityofsalmonasaculturalresourceina7 post‐Sovietcommons.AnthropolWorkRev32:77–89.8 

37. CrosonRTA(2001)Feedbackinvoluntarycontributionmechanisms:Anexperimentinteam9 production.ResExpEcon8:85–97.10 

38. EckelCC,GrossmanPJ(2005)Managingdiversitybycreatingteamidentity.JEconBehav11 Organ58:371–392.12 

39. SeftonM,ShuppR,WalkerJM(2007)Theeffectofrewardsandsanctionsinprovisionof13 publicgoods.EconInq45:671–690.14 

40. BoltonGE,KatokE,OckenfelsA(2005)Cooperationamongstrangerswithlimited15 informationaboutreputation.JPublicEcon89:1457–1468.16 

41. EngelmannD,FischbacherU(2009)Indirectreciprocityandstrategicreputationbuildingin17 anexperimentalhelpinggame.GamesEconBehav67:399–407.18 

42. MilinskiM,RockenbachB(2012)Ontheinteractionofthestickandthecarrotinsocial19 dilemmas.JTheorBiol299:139–43.20 

43. PanchanathanK,BoydR(2004)Indirectreciprocitycanstabilizecooperationwithoutthe21 second‐orderfreeriderproblem.Nature432:499–502.22 

44. HuberPJ(1967)Thebehaviorofmaximumlikelihoodestimatesundernonstandard23 conditions.ProcFifthBerkeleySympMathStatProbab1:221–233.24 

45. WhiteH(1980)Aheteroskedasticity‐consistentcovariancematrixestimatorandadirect25 testforheteroskedasticity.Econometrica48:817–838.26 

46. HenrichJetal.(2004)Foundationsofhumansociality:economicexperimentsand27 ethnographicevidencefromfifteensmall‐scalesocieties(OxfordUniversityPress,Oxford).28 

47. GerkeyD(2013)Cooperationincontext:publicgoodsgamesandpost‐Sovietcollectivesin29 Kamchatka,Russia.CurrAnthropol54:144–176.30 

48. KimbroughEO,VostroknutovANormsmakepreferencessocial.JEurEconAssoc.31 

25

49. BalafoutasL,NikiforakisN,RockenbachB(2014)Directandindirectpunishmentamong1 strangersinthefield.ProcNatlAcadSciUSA111:15924–15927.2 

50. WinterhalderB,LuF,TuckerB(1999)Risk‐senstiveadaptivetactics:Modelsandevidence3 fromsubsistencestudiesinbiologyandanthropology.JArchaeolRes7:301–348.4 

51. KaplanHS,SchniterE,SmithVL,WilsonBJ(2012)Riskandtheevolutionofhuman5 exchange.ProcBiolSci279:2930–5.6 

52. AktipisCA,CronkL,AguiarR(2011)Risk‐poolingandherdsurvival:anagent‐basedmodel7 ofaMaasaigift‐givingsystem.HumEcol39:131–140.8 

53. SmaldinoPE,SchankJC,McElreathR(2013)Increasedcostsofcooperationhelpcooperators9 inthelongrun.AmNat181:451–63.10 

54. HerrmannB,ThöniC,GächterS(2008)Antisocialpunishmentacrosssocieties.Science11 319:1362–1367.12 

55. AmbrusA,GreinerB(2012)Imperfectpublicmonitoringwithcostlypunishment:an13 experimentalstudy.AmEconRev102:3317–3332.14 

56. EgasM,RiedlA(2008)Theeconomicsofaltruisticpunishmentandthemaintenanceof15 cooperation.ProcBiolSci275:871–8.16 

57. RobertsG(2013)Whenpunishmentpays.PLoSOne8:e57378.17 

58. GualaF(2012)Reciprocity:Weakorstrong?Whatpunishmentexperimentsdo(anddonot)18 demonstrate.BehavBrainSci35:1–15.19 

59. HooperPL,GurvenM,WinkingJ,KaplanHS(2015)Inclusivefitnessanddifferential20 productivityacrossthelifecoursedetermineintergenerationaltransfersinasmall‐scale21 humansociety.ProcBiolSci282:20142808.22 

60. GurvenM,Allen‐AraveW,HillK,HurtadoM(2000)“It’saWonderfulLife”signaling23 generosityamongtheAcheofParaguay.EvolHumBehav21:263–282.24 

61. SugiyamaLS(2004)Illness,injury,anddisabilityamongShiwiarforager‐horticulturalists:25 implicationsofhealth‐riskbufferingfortheevolutionofhumanlifehistory.AmJPhys26 Anthropol123:371–89.27 

62. TuckerB,Tsimitamby,HumberF,BenbowS,IidaT(2010)Foragingfordevelopment:a28 comparisonoffoodinsecurity,production,andriskamongfarmers,forestforagers,and29 marineforagersinSouthwesternMadagascar.HumOrgan69:375–386.30 

63. BerkesF,FolkeC,ColdingJ(1998)Linkingsocialandecologicalsystemsmanagement31 practicesandsocialmechanismsbuildingresilience(CambridgeUniversityPress,Cambridge).32 

33