PDF Download 3777416.pdf 26 March 2026 Total Citations: 2 . Total Downloads: 168 . Latest updates: hps://dl.acm.org/doi/10.1145/3777416 . . Published: 18 November 2025 . . Accepted: 20 October 2025 RESEARCH-ARTICLE Revised: 24 September 2025 DAO Governance: Voting Power, Participation, and Controversy - A Received: 26 January 2025 . Review and an Empirical Analysis . Citation in BibTeX format . . MARKUS JUNGNICKEL, Imperial College London, London, U.K. . FERDA ÖZDEMIR SÖNMEZ, University of West London, London, U.K. . CATHY MULLIGAN, Imperial College London, London, U.K. . WILLIAM J. KNOTTENBELT, Imperial College London, London, U.K. . . . Open Access Support provided by: . Imperial College London . University of West London . Distributed Ledger Technologies: Research and Practice hps://doi.org/10.1145/3777416 EISSN: 2769-6480 . DAO Governance: Voting Power, Participation, and Controversy - A Review and an Empirical Analysis MARKUSJUNGNICKEL,ImperialCollegeofLondon,DepartmentofComputing,UK FERDAÖZDEMIRSÖNMEZ,ImperialCollegeofLondon,DepartmentofComputing,UKandUniversity ofWestLondon,SchoolofComputingandEngineering,UK CATHYMULLIGAN,ImperialCollegeofLondon,DepartmentofComputing,UK WILLIAMJ.KNOTTENBELT,ImperialCollegeofLondon,DepartmentofComputing,UK DAOshaveemergedasanovelorganizationalstructure,attractinggrowinginterestduetotheirdecentralized,transparent governance,whichreplacestraditionalhierarchieswithstakeholder-managedrulescodifiedassmartcontracts.Althoughvar- iousgovernancemodelsexist,comparativeresearchacrossdimensionsremainslimited,leavingtheliteraturefragmentedand offeringlittlepracticalguidanceforselectingsuitablemodels.Thispapercriticallyanalysesexistinggovernancemechanisms andtheirimplementationtosupportthedevelopmentofmoreeffectiveDAOmodels.Toaddresscurrentgaps,wereviewprior quantitativestudiesandconductexploratorydataanalysisoncentralization,participation,anddecisioncontroversy.The findingsshowthatreputationandshare-basedmodelscanmitigatethecentralizationseenintoken-basedsystems,thoughall modelssufferfromlowmemberengagement,suggestinganoverrelianceondirectdemocracy.Ouranalysiscanbereplicated acrossplatformsandtimeframestorefineandvalidatetheseinsights. CCSConcepts:•Appliedcomputing→Electronicfundstransfer;•Computingmethodologies→Distributed computingmethodologies;•Computersystemsorganization→Peer-to-peerarchitectures. AdditionalKeyWordsandPhrases:smart-contract,blockchain,infrastructureprojects,governance,voting,dao,decentralized autonomousorganization 1 INTRODUCTION TheDAOecosystemhasexperiencedstronggrowthoverthelastthreeyears[8],andthereisanincreasing interestinusingDAOstomanagenotjustblockchainsolutionsbutalsofinancialservices,sustainabilitygoals, andnearlyeveryaspectofsocietyhashadaDAOproposedforthem.Atthesametime,DAOsarestilldeveloping appropriategovernancemodelsthatsuitevermorecomplexusecases[8],whichcanbeaccommodatedbythe decentralizedandimmutablenatureofblockchains[23]. Developingarobustgovernancemodelhasalwaysbeencriticaltothesuccessandlongevityoforganizations andhasbeenstudiedforaverylongtime;infact,itshistorypredatesthatofblockchaintechnologyandallother Authors’addresses:MarkusJungnickel,markusjungnickel@gmail.com,ImperialCollegeofLondon,DepartmentofComputing,South KensingtonCampus,London,England,UK,SW72AZ;FerdaÖzdemirSönmez,f.ozdemir-sonmez@imperial.ac.uk,ImperialCollegeof London,DepartmentofComputing,SouthKensingtonCampus,London,England,UK,SW72AZandUniversityofWestLondon,School ofComputingandEngineering,StMary’sRoad,Ealing,London,England,UK,W55RF;CathyMulligan,c.mulligan@imperial.ac.uk, ImperialCollegeofLondon,DepartmentofComputing,SouthKensingtonCampus,London,England,UK,SW72AZ;WilliamJ.Knottenbelt, w.knottenbelt@imperial.ac.uk,ImperialCollegeofLondon,DepartmentofComputing,SouthKensingtonCampus,London,England,UK, SW72AZ. Permissiontomakedigitalorhardcopiesofallorpartofthisworkforpersonalorclassroomuseisgrantedwithoutfeeprovidedthat copiesarenotmadeordistributedforprofitorcommercialadvantageandthatcopiesbearthisnoticeandthefullcitationonthefirstpage. Copyrightsforcomponentsofthisworkownedbyothersthantheauthor(s)mustbehonored.Abstractingwithcreditispermitted.Tocopy otherwise,orrepublish,topostonserversortoredistributetolists,requirespriorspecificpermissionand/orafee.Requestpermissions frompermissions@acm.org. ©2025Copyrightheldbytheowner/author(s). ACM2769-6480/2025/11-ART https://doi.org/10.1145/3777416 Distrib.LedgerTechnol. 2 • Jungnickeletal. pertinenttechnologicaladvancements.Aristotle[49]wasanearlyadvocateforgovernancemodels.Theformation ofthe”RuleofLaw”[30]isalsoasignificantaccomplishmentinthehistoryofgoverningparadigms,asitsupports theequalityofallcitizensbeforethelaw,preventsthearbitraryuseofpower,andensuresanon-arbitraryform ofgovernance.Significantamountsofliteraturehavebeendedicatedtostudyingorganizations,organizational theoryandsimilartopicsthatoutlinegovernancemechanismsforthemanagementofsociety,companiesand otherstructureswherehumansneedtocollaborateandcoordinateinvariousforms. Similarideasapplytotoday’sentirelytechnology-dependentfuturisticorganizations,DAOs.However,although theoreticallysettingupanewDAOorganizationisstraightforward,theimplicationsofincorrectgovernance systemswouldbehigh,causingirreversiblesituationsandcausedamagebeyondrepair.Forinstance,in“The DAO”case,anoverlookedpieceofcodeintendedtopromotedecentralizationandencouragethecreationof “childDAOs”ultimatelyledtotheorganization’scollapse”[58].Suchholesingovernancemechanismsarecritical issuesforDAOstoaddresstobeusedeffectivelywithinsociety. Thispaperexaminesexistinggovernancemodels,focusingoncentralization,participation,controversyand theeffectsoftheseongovernancesystems.Theresearchismotivatedbytheneedtoincreaseunderstandingof andimproveexistingDAOgovernancemodels,aswellasenablecomparisonofthemalongseveralvariables. 1.1 ObjectivesandAim ThisstudyaimstocriticallyevaluateexistingDAOgovernancemodelstosupportthedevelopmentandselection ofmoreeffectivemechanisms.Thespecificobjectivesare: • Analyzehowkeygovernancemetrics—participation,controversy,andvotingpower—differacrossmodels. • ReviewandclassifyDAOvotingmechanismsbyweight,type,andmethodtoofferastructuredoverview. • Performexploratorydataanalysistocomparevotingpowerdistribution,memberparticipation,anddecision controversyacrossgovernancemodels. • Providetransparent,repeatableanalysismethodstoaccommodatetheevolvingDAOecosystem. 1.2 Contributions Themostsignificantcontribution,inlinewiththecorrespondingobjectivessetoutinSection1.1,istheliterature reviewwhichidentifiedtheproblemswithexistinggovernancemodelsandhighlightscriticalareasthatremain understudiedbyexistingresearch. Thestudyalsoincludesanexploratorydataanalysisasasecondarycontributiontopresenttheidentified problems and the gaps visually. For this project, the exploratory analysis does not form the basis of further statisticalhypothesistesting.Withoutmorerobustempiricaldata,exploratorydataanalysiswasusedtodetermine theadvantagesanddisadvantagesofvariousdesignapproachesfortheDAOs.Theoutputsoftheexploratory dataanalysiscanbeusedtosetupcomprehensivelyandtesthypothesesforfuturework. Thereviewshowsthatexistingresearchissomewhatpiecemealandfailstocomparedifferentgovernance modelsacrossmultipledimensions.Thedifferencebetweenon-chainandoff-chainvotinghasalsoremained understudied.Thesegapsareproblematicbecausesuchcomparisonsarenecessarytoformulatehypothesesabout whichgovernancefeaturesarecausingissues.Theexploratorydataanalysisconductedforthisresearchattempts tofillsomeofthesegaps.Thisaimstomakeinitialobservations,revealconnectionsanddevelophypothesesabout thedata;theresultsarethentypicallytestedinaconfirmatorydataanalysis[61].Threeofthemostcommon governancesystemsdifferintheirperformanceacrosskeydimensions: • UnequalVotingPower:Theresultsrevealhowreputationandshare-basedgovernancesystemsmanageto avoidthehighdegreesofcentralizationofthevotingpowerintoken-basedgovernance. Distrib.LedgerTechnol. DAOGovernance:VotingPower,Participation,andControversy-AReviewandanEmpiricalAnalysis • 3 • LowEngagement:Theanalysisshowslowvoterparticipationacrossgovernancemodelsandsupportsthe hypothesisthatthisisnotduetothecostandcomplexityofon-chainvotingbutanexcessiveuseofdirect democracy. • Absence of Controversy: The results demonstrate that both on and off-chain decision-making in DAOs is mostlyuncontroversial;itisnotthecasethatafewpowerfulmembersareusingtheirvotingweighttooutvote apowerlessmajority.However,thisispartiallytheresultofextremelylowparticipationamongthemajority. Therestofthepaperisorganisedasfollows.Section2summarisesthestateoftheart.Section3definesthe scopeofthereview.Section4describesthedata’slimitations,followedbyadetailedmethodologydescription. Section5isdedicatedtothereviewresultscontainingboththeliteraturesummarywithgapsandtheresultsof theexploratorydataanalysis.Section6hasthediscussionpart,whichincludesadetaileddiscussionofselected parametersandtheiridentifiedeffectsonthegovernancesystems.Finally,Section7concludes. 2 STATEOFTHEART 2.1 ConceptandDefinitionforDAOs Adecentralizedautonomousorganization(DAO)employsalgorithmicgovernancethroughsmartcontractsona blockchain,withthedegreeofautonomyvaryingamongdifferentDAOs[68][64]. ThedefinitionofDAOshasbeenfurtherrefinedbyButerin,whodistinguishesthemfromDistributedOrgani- zations(DOs)[64].Whilethelatterisgovernedbythedecision-makingofhumanmembers,theformermakes decisionsautonomously.Althoughthedistinctionisoftheoreticalinterest,thepreciseclassificationofaproject asaDOorDAOhaslittlepracticalrelevance.Nonetheless,it’sworthnotingthatsomeDAOsconsideredinthis paperaremeanttobegovernedbymajoritydecisions,whichwouldnotqualifyasfullyautonomousDAOssince theydependonthemajoritywilloftheirmembers. 2.2 DAOGovernanceModels 2.2.1 DemocraticGovernance. Democraticgovernancereferstothedecision-makingprocesseswithinDAOs thatuseavotingsysteminwhichmemberscanfreelyparticipate.Avotingsystemencapsulatesthreeconcepts: VoterRegistration,VotingWeight,andVotingMethod. VoterRegistration: DAOslimitparticipationingovernancevotesbyimplementingmembershipsystems(Voter Registration).Theprimarypurposeofamembershipsystemistorecorddifferencesinvotingweight,ensureonly individualsinterestedintheDAOscanparticipate,andguardagainstSybilattacks.Thethreemostcommonly usedmembershipsystemsare[12]: Token-basedmembership MembershipisgrantedviathepossessionoftheDAO’sfungiblegovernancetokens; theextentofgovernancepowertypicallydependsonthenumberoftokensanaddressholds.Thegovernance token can be freely transferred and purchased on public exchanges. Since the value of a governance token istypicallydependentonthesuccessofitsDAO,ownershipofthetokentheoreticallyalignstheindividual’s economicinterestswiththoseofthecollective. Share-basedmembership Similartoacompany,membersholdsharesthatgrantthem(proportionate )own- ershipoftheDAO’streasuryandvotingrightsinitsgovernance.Unliketoken-basedsystems,sharescannot simplybeacquiredonanopenexchangebutareearnedorgrantedoncertainconditions.Differentsharetypes arepossible,includingadistinctionbetweenpurelyeconomicandvotingshares. Openmembership Inthismodel,anyonecanparticipateingovernance,butgovernancepowerdependson certainattributes.Forexample,someDAOsusereputationsystems,wherebymemberswithahigherreputation scorearefavoredinthegovernanceprocess.Reputationcanonlybeearned(i.e.grantedbyothermembers),not Distrib.LedgerTechnol. 4 • Jungnickeletal. purchasedortransferred.WhileanyonecanparticipateintheDAO,onlythosethathaveearnedareputationwill haveanimpact. DAOs grant membership to addresses, not individuals. Some share-based DAOs seek to vet new member addressesandidentifytheindividualscontrollingthem(e.g.theLAO[60]),butmosttokenandopen-membership DAOsdonot.Sincegovernancetokenscanbetransferredfreely,vettingeachnewholderwouldnotbeviable. Openmembershipbydefinitionimposesnorestrictionsonparticipation;however,attributessuchasreputation can be reserved for vetted members. Most DAOs also allow smart contracts to become members: it is not uncommonthatmembersarethemselvesDAOs(e.g.ConvexDAOinCurveDAO[15]). VoterWeight: AttributingequalweighttothevotesofallmemberswouldexposetheDAOtotheriskofSybil attacks.Unlessnewmemberaddressesarevetted,oneindividualcancontrolmanymemberaddresses,meaning attackerscouldsimplygeneratenewaddressesuntiltheycontrolamajority.SincemostDAOsprefernotto imposerigorousmembershipcontrols,theyinsteaddistributevotingpowerunequally.Thisdecreasestheriskof Sybilattacksbecausevotingpowerisnolongerproportionatetothenumberofaddressesindividualcontrol.Itis alsoviewedasfairthatthosememberswhohaveinvestedorcontributedmosttotheDAOshouldberewarded andtrustedwithalargershareofthevotingpower.Themostcommonapproachesforvote-weightingisdiscussed below. Token-BasedWeighting Intoken-basedvotingsystems,theweightofavotecastbyanaddressisdetermined bythenumberofgovernancetokensitcontrols.Typicallyoneunitofweightisgrantedforeachtokensothat powerisproportionatetotokenownership.Curve[18],Uniswap[63],Maker[37],andothermajorprotocol DAOsusethismechanism. Share-BasedWeighting Proposalsareapprovedbyasimplemajorityofshares;themoresharesamember controls,thegreatertheirvotingweight.Thismeanstheapproachisverysimilartotoken-basedvoting.Itwas famouslypioneeredbytheMolochDAO[55]andisnowusedbyDAOhaus,aDAOcreationwebsite. Reputation-BasedWeighting Inreputation-basedweighting,amember’sreputationbalancedeterminesthe weightofthevotestheycast.Thismeansvotingweightcannotbepurchasedortransferredbutmustbegranted bytheDAO.SeveralmajorDAOtoolingprovidersofferreputation-basedgovernancemechanisms,including Aragon[24],Colony[24],andDAOstack[24].Oneofthemostwell-knownreputation-basedDAOsisdxDAO[39]. QuadraticWeighting Quadraticvotingmechanismstypicallyusegovernancetokensbutimplementanon- linearrelationshipbetweenthenumberoftokensownedandvotingpower[65].Unliketraditionaltoken-based voting,amembercanchoosehowmanyoftheirgovernancetokenstouseonagivenproposal;theweightof theirvoteWisthesquarerootofthenumberoftokensTusedEq.1: √ 𝑊 = 𝑇 (1) Thismeansthecostofaunitofweightincreasesquadratically.Unliketheaboveapproaches,thetokensusedon aproposalarenolongeravailableforvotingonotherproposals.Membersthusneedtochoosehowtoallocate theirsupplyofgovernancetokensacrossdifferentproposals.Theusedtokensarethenre-distributedequally amongvotersafterafixedperiodorburnedpermanently. VotingMethod: AnotherelementofDAOgovernanceisthevotingmethodemployedtoselectawinningou tcome. Although DAOs have access to a wide range of voting methods [43], most primarily adopt one of the three approachesoutlinedbelow.Foramorecomprehensivereview,refertothetaxonomyofDAOvotingmethods createdbyAragon[3],asupplierofDAOgovernanceinfrastructure. PluralityVoting Pluralityvotingmeanstheoptionwiththemajorityofvotingweightwins;inDAOs,the optionsareusuallyabinarychoicebetweentheapprovalorrejectionofagivengovernanceproposal.Sincethis meansonlyasimplemajorityisrequiredtopassaproposal,quorumsareusedtoenforceaminimumamount Distrib.LedgerTechnol. DAOGovernance:VotingPower,Participation,andControversy-AReviewandanEmpiricalAnalysis • 5 of participation. This voting method is very popular and is used by most major token-based DAOs, such as Uniswap[63]andCompound[14]. ContinuousApprovalVoting Incontinuousapprovalvoting,competingproposalsaremutuallyexclusive alternatives.Whateverproposalreceivesthemostvotesgetsimplemented.Onceaproposalhasbeenapproved, thevotesinitsfavorremainrelevantsinceanewalternativeproposalisonlypassedifitreceivesmorevotes thanthelastwinningproposal.Thismeansvotingisalwaysachoicebetweenkeepingthestatusquo(i.e.the lastwinningproposal)orreplacingitwiththenewproposal.Votersseekingtoobjecttothenewproposaldo notvoteagainstitbutsimplyvoteinfavorofthelastwinningproposal,expressingtheirsupportforkeeping thestatusquo.Thisvotingmechanismisused,amongothers,byMakerDAO[37],whichgovernstheprotocol responsiblefortheDaistablecoin[24]. HolographicConsensusVoting Holographic Consensus is a voting mechanism that seeks to resolve the requirementofhighmemberengagement[19][26].Itallowsmemberstobetastakeonwhetherproposalswill beacceptedorrejectedbythecommunity,therebyfilteringtheproposalsforthemostrelevantones[22].The moretokenshavebeenstakedonthesuccessofaproposal,themorelikelyitistocometotheattentionof additionalmembers.ParticipantsintheDAOcanthusbemorepassiveandonlyreviewhighly-backedproposals. Whenenoughtokenshavebeenstakedonthesuccess,thevotingrulesfortheproposalareamendedtoallowfor asimplemajoritywithoutaquorumbeingreached.Thismeansonlyasubsetofmembersneedtoparticipate inthosevotes,forwhichthereisahighchancethattheywouldbevotedforbyalargemajorityanyway[26]. ThisvotingmechanismwaspioneeredbyDAOstack,aDAOcreationwebsite,andhasbeensuccessfullyusedby dxDAO[27]. 2.2.2 ExecutiveGovernance. Executivegovernanceconcernsdecision-makingprocessesinDAOsthatdonot relyonmembervotes.Decisionsareeithermadebysmallgroupsorsingleindividuals. SmallGroupDecisions Small group decision-making in DAOs is typically done through multi-signature wallets, such as Gnosis Safes [53]. In their usage, these wallets are akin to standard wallets (public/private keypairs),butarecontrolledbyseveralowners.Tousethewallet,apreviouslyspecifiedfractionoftheowners mustsignthetransactionwiththeirownwallet. Multi-signaturewalletsaremainlyusedinthreeways:1)astheonlymechanismgoverningtheDAOorasub- sectionoftheDAO,2)asasafeguardthatcanoverrideorvetodemocraticdecisions,or3)implementnon-binding off-chainvotes.Multi-signaturewalletstypicallyuseabasicapprovalsystem,whereanytransactionsignedby MoutofNownersisexecuted.Moresophisticatedmechanismsforsmall-groupdecisionsmakinghavebeen developed(e.g.Colony[13]),butarelesscommonlyused. IndividualDecisions Some DAOs allow certain individuals to make low-impact governance decisions by themselves;inparticular,DAOscreatedusingColony’sgovernancemodelemploythismechanism[2].Such decision-makingistypicallyimplementedviaapermissionregistrythatrecordswhichmemberaddressesare allowedtocallwhichfunctionsintheDAO.Beforeexecutingthefunctioncode,theDAOsmartcontractconsults thepermissionregistry,todeterminewhetherthememberisentitled. 2.3 DAOLandscape TheDAOlandscapecanbroadlybeseparatedintothreesegments:1)DAOsusedforthecollectionanddistribution offunds,2)ProtocolDAOsthatenablecommunitycontrolovertokenprotocols,and3)ServiceDAOs,suchas thoseusedforsocialmediaDApps[20][16].Thisreviewwillpresentacross-sectionofDAOsspanningthese segments.TheDAOswereselectedsothateachofthemaingovernancemodelsdescribedaboveisrepresented. Distrib.LedgerTechnol. 6 • Jungnickeletal. 3 SCOPE 3.1 GovernanceModelsAcrossSelectedDimensions TheanalysisfocusedonunderstandinggovernancebehaviorwithinthemostprominentDAOvotingsystems acrossselecteddimensions.Itshouldbenotedthattheterm’governancemodels’inthiscontextspecifically referstotheaspectsofvotingpower,controversy,andparticipation.Weacknowledgethatgovernancemodels canencompassbroaderintrinsicdetails,suchaschosenvotingmechanisms,hierarchies,ortheabsencethereof. However,ouranalysisislimitedtotheselecteddimensionsmentionedabove,focusingprimarilyonmembership, proposalcreation,andvotingdata.Thislimitationarisesfromthechallengeofcomparingalargenumberof DAOsacrossvariousdetailedparametersthatmaynotbeuniformlyavailable. Inthisstudy,wecategorizeDAOsintothreeprominentgovernancemodels:”share,””reputation,”and”token”- based models, which are identified as the most common approaches. While all these models use tokens for representation,thenatureanduseofthesetokensdiffersignificantly.”Share”DAOsinvolvetokensthatoften representstakedbalances,whichmaynotbeactivelytradable.”Reputation”DAOsusenon-transferabletokens earnedthroughcontributionsandactivitieswithintheDAO.Ontheotherhand,”token”DAOsuseactively tradabletokens,whicharetypicallyliquid(i.e.,transferabletokenswithhighmarketliquidity)andcanbeused fortransactionsoutsidetheDAO.Thisdistinctionhelpstobetteranalyzeandcomparethedifferentgovernance structuresandtheirimplicationsonvotingpowerandparticipationwithinDAOs. EachofthemodelswasrepresentedbyanumberofDAOs;alloftheselectedDAOsaredeployedtoEthereum Mainnetoroneofitssidechains. • Share:DAOhausisanapplicationthroughwhichcreatorscanconfigureanddeploypre-madeDAOtemplates1 , which implements the Moloch DAO’s share-based governance model. DAOhause was used to represent share-basedgovernancebecauseallDAOscreatedthroughthiswebsiteareverysimilar,meaningtheycanbe analyzedintheaggregate.Intotal,503DAOswereincluded;thethreelargestbyAuMwereselectedfordeeper analysis. • Reputation:DAOstackisalsoanapplicationforcreatingandconfiguringDAOs.Itoffersareputation-based governancemodeldescribedinthecasestudyofdxDAO.LikeDAOhaus,DAOstackwasselectedbecauseit hasproducedalargenumberofverysimilarreputation-basedDAOs.Atotalof211DAOswereincluded,and dxDAOwasselectedfordeeperanalysis. • Token:AnumberofprotocolDAOswereselectedtorepresenttoken-basedvoting.TheseDAOsgoverntoken protocols, such as those used for decentralised borrowing and lending [20] [17]. Unfortunately, no DAO creationtoolcouldbeidentifiedthathasproducedalargenumberofsimilartokenDAOsforwhichgovernance dataisavailable2;forthisreason,foursimilarbutindividuallycreatedDAOshadtobeused. Additionally,votingdatafrom9725DAOsonSnapshot[54],apopularoff-chainvotingapp,wasexamined. ArecentstudybyWangetal.[67]providesalarge-scaledescriptiveanalysisofDAOsusingSnapshotdata, focusingonclassificationofDAOtypes,votingmechanisms,andtokenusage.Ourworkiscomplementaryin thatweemployDAO-specifictechnicalmetrics(e.g.,Gini,Lorenz,andNakamotocoefficients)andcontroversy measures,therebyofferingamorefine-grainedperspectiveoncentralizationdynamicsandparticipationtrends thatarenotaddressedinWangetal.’sstudy.TheresultsofvotesonSnapshotarenotimplementedautomatically 1Whiletherearemanycustom-madeDAOswithuniquegovernancemodels,theemergenceofDAOtemplatesandlibrarieshaveledtoa convergenceofgovernancemodels.ThemajorityofDAOsarecreatedviawebsitesthatallowuserstoconfigureanddeployDAOswithout requiringanytechnicalunderstanding.Thesewebsitesdifferentiatethemselvesbythegovernancemodelstheyoffer;acomparisonofthe DAOscreatedfromdifferentwebsitesthusallowsforacomparisonofdifferentgovernancemodels.Themostpopularcreationwebsitesare Aragon,Colony,DAOhausandDAOstack[26][48]. 2WhileAragonisapopularapplicationofferingtoken-basedtemplateDAOs,itwasnotusedbecausegovernancedatacouldnotbeaccessed, andtheDAOsitproducesvarymoreintheirgovernancesystems. Distrib.LedgerTechnol. DAOGovernance:VotingPower,Participation,andControversy-AReviewandanEmpiricalAnalysis • 7 on-chain.DAOsusuallyeitheruseSnapshotasapollingtoolpriortoon-chainvotingtotestmembersentiment (e.g.Uniswap)orimplementsomemechanismtotranslateoff-chaindataintoon-chainexecution(e.g.Sushi DAO).DAOsonSnapshotuseawiderangeofvotingsystems,includingthethreeselectedforanalysishere.This allowsalike-for-likecomparisonofonandoff-chaindecision-making. 3.2 GovernanceDimensions InSection2.2,weintroducedvariouscategorizationsrelatedtogovernancemodels,suchasdifferentmembership methodsandvotingweightmechanisms.Thiscomprehensiveoverviewaimedtoprovidereaderswithafounda- tionalunderstandingofthediverseapproachesinDAOgovernance.However,thecategorizationinthissection focusesspecificallyonthedataandanalysesconductedinthisstudy,includingvotingpower,controversy,and participation.Whileouranalysisincludesmembershipdataandvotingpower,itdoesnotdelveintothespecifics of how membership is determined or the detailed calculations of voting weighting for each vote. Therefore, thesimplificationisalignedwithourstudy’sscopeanddata,allowingforastructuredanalysisofgovernance behavioracrossthemostprominentvotingsystemsusedinDAOs. • VotingPower:Theanalysissoughttoshedlightonthedistributionofpowerandhowitdiffersacrossgovernance models.Thisisakeygovernancedimensionsincehighlyunequaldistributionscanleadtoacentralizationof power,introducingconcernsaboutsecurity,minorityprotection,andtheextenttowhichdecision-makingis democratic. • Participation:Participationconcernstherateatwhichmembersentitledtovoteongovernanceproposals actuallypartakeindecision-making.Thisisakeydimensionofgovernanceasitrevealswhoismakingthe decisionsintheDAO,andhowthisdependsonotherfactors,suchasvotingpower. • Controversy:ThecontroversyofDAOdecision-makingconcernstheextentofdisagreementaboutgovernance proposalsamongmembers.Itisakeydimensionofgovernancesinceitshedslightontheefficiencyofthe consensusmechanismsusedandtheextenttowhichmemberinterestsarealigned. ThesemetricsareofparticularimportanceastheyprovidevaluableinsightsintothehealthofDAOsandhelp identifyareasforenhancingtheirgovernancestructures.Byanalyzingandcomparingthesemetricsacrossa selectedgroupofDAOmodels,weaimtoprovideacomprehensiveevaluationoftheirgovernancehealth.Other potentialgovernancemetricsmayalsoberelevant,butforthepurposeofthisstudy,weprioritizetheanalysisof participation,controversy,andvotingpowermetrics.Thereareseveralreasonsforthis.First,thedistribution ofvotingpowerwithinaDAOcansignificantlyimpactitsgovernanceprocessesandoutcomes.Second,ahigh levelofparticipationfrommembersiscrucialforthesuccessofDAOs.Third,measuringthelevelofcontroversy withinaDAOcanhelpidentifyareasofpotentialconflictandfacilitatetheresolutionofdisputes. WhileourstudypresentsanempiricalanalysisofdecentralizationtrendsacrossleadingDAOs,recenttheoretical worksshouldbeacknowledged.Forinstance,Hanetal.[32]developamodelthatexploreshowtokenlocking mayalleviateconflictsofinterestamongstakeholders,offeringagovernancedesignrationale.Laturnus[36] highlightspersistentlylowparticipationratesinDAOs,complementingourfindingsonfluctuatingmember engagement.Additionally,recentreviewstudiessuchasHanetal.[31]andJiangandLi[34]providebroader theoreticalperspectives.Comparedtothese,ourcontributionisdistinctinofferingplatform-specifictemporal metricsandvisualanalysestodissectevolvinggovernancebehaviorsempirically. 4 DATAANDMETHODOLOGY 4.1 Dataset DataaboutgovernancebehaviorisemittedandstoredbythesmartcontractsconstitutingtheDAO.Theoretically, thisdatacanberetrievedbyqueryingablockchainnodeforpastsmartcontracteventsorcallingmethodsonthe Distrib.LedgerTechnol. 8 • Jungnickeletal. Table1. DataCollectionCodeandCommandSamples (a) (c) { curl −X POST −H ”Content−Type: application/json” avatarContracts(first:5) −d ’{”query”: ”{ proposals(first: 5) { { id proposer description status id votes { id voter support } } }”}’ address https://api.thegraph.com/subgraphs/name/daosubgraph/moloch name Ifauthenticationisrequired,includeanAPIkeyintheheaders: nativeToken } curl −X POST −H ”Content−Type: application/json” −H ”Authorization: Bearer YOUR_API_KEY” } −d ’{”query”: ”{ proposals(first: 5) { id proposer description status votes { id voter support } } }”}’ https://api.thegraph.com/subgraphs/name/daosubgraph/moloch (b) (d) { proposals( first: 1000, {”data”: { ”avatarContracts”: [{ skip: 0, ”address”: ”0x006087d6ac20840c23ba298512db454a05c19b10”, where: { ”id”: ”0x006087d6ac20840c23ba298512db454a05c19b10”, space: ”sushigov.eth” ”name”: ”FitTogether”, }, ”nativeToken”:”0xa3820e0f6be1c306c0a76746af80b60c228d99c2” orderBy: ”created”, }, { orderDirection: desc ”address”: ”0x00e1b6de09e01d5b178ecf68966a34bd1dcd4064”, ) { ”id”: ”0x00e1b6de09e01d5b178ecf68966a34bd1dcd4064”, id ”name”: ”YoyoDAO”, title ”nativeToken”:”0xb9697151c7af8f8a4d2702c8291e3d649525b1d9” choices }]}} state end (e) snapshot scores_total import Web3 from ”web3”; scores import { AbiItem } from ”web3−utils”; state import { LAO_ABI } from ”./ABIs”; author import fs from ”fs”; created space { const provider = ”https://mainnet.infura.io/v3/ id YOUR_INFURA_PROJECT_ID”; name const web3 = new Web3( } new Web3.providers.HttpProvider(provider)); votes const adr = ”0x8F56682a50BECB1df2Fb8136954f2062871bc7fc”; } const contract = new } web3.eth.Contract(LAO_ABI as AbiItem[], adr); async function getPastVotes() { fs.writeFileSync( ”./votes.csv”, ”memberAddress,proposalId ,vote\n”); const events = await contract.getPastEvents(”SubmitVote”,{ fromBlock: 9000000, toBlock: ’latest ’ }); events.forEach(event => { const memberId = event.returnValues.memberAddress; const propId = event.returnValues.proposalId; const vote = event.returnValues.uintVote; fs.appendFileSync(”./votes.csv”, ‘${memberId},${propId}, ${vote}\n‘); }); } Distrib.LedgerTechnol. DAOGovernance:VotingPower,Participation,andControversy-AReviewandanEmpiricalAnalysis • 9 contracttoextractthedataitcontains.SeveralDAOsusesubgraphstomakedataabouttheirgovernancemore readilyaccessible.Subgraphsaremaintainedbyindexers,wholistentoeventsontheblockchain,storetheevent data,andofferaGraphQLinterfacetorunqueries[59]. Theprimarysourceofdatafortheanalysiswassuchsubgraphs,buttheydidnotalwayscontainalltherequired information.Insuchcases,datapublishedontheDAOwebsitewereusedtofillinthegaps.Table2containsa completelistofthesourcesusedfortheanalysis3. 4.1.1 DataCollectionUsingTheGraph. TheGraphisadecentralizedprotocolforindexingandqueryingdata fromblockchains.ItallowsdeveloperstobuildandpublishopenAPIs,calledsubgraphs,thatapplicationscan queryusingGraphQL.Subgraphsarebuiltbydefiningaschemathatspecifiesthedatastructureandthesources ofthedata.Oncedeployed,indexersonTheGraphnetworkcontinuouslyupdatethesubgraphsasnewdatais addedtotheblockchain. TheGraphQueries:QueriestosubgraphsonTheGrapharemadeusingGraphQL,whichallowsyouto requestspecificdatainastructuredformat.InTable1athereisabasicexampleofaGraphQLquerytoretrieve thefirstfiveofthemaincontracts(avatarcontracts)fortheDAOsthatarepartofDAOstackDAOs. RunningTheQueries:Torunthesequeries,youcanuseTheGraph’shostedserviceoradecentralizedGraph node.Table1cshowsacommandtorunaqueryusingthehostedservice. EachsubgraphhasauniqueURL.Forexample,theMolochDAOsubgraphURLmightbe: https :// api . thegraph .com/subgraphs/name/daosubgraph/moloch SomeGraphQLendpointsrequireanAPIkey.YoucanobtainthisbysigninguponTheGraph’shostedservice platform.YoucanthenusethesamecommandwiththeinclusionofAPIkey,Table1c. QueryResultsWhenthequerydescribedinSection4.1.1isexecuted,theresultissimilartothesample showninTable1a.Dependingonthequeryandcorrespondingdatastructure,TheGraphprovidesaJSONformat file.ThisJSONfileincludeskey-valuepairsrepresentingdifferentaspectsofthequerieddata,suchasaddresses, IDs,names,andnativetokens.ThesamplefromTable1dillustratesthetypicalstructureofsucharesultforthe firsttwoDAOsfromDAOstack. 4.1.2 DataCollectionUsingSnapshotAPI. SnapshotisadecentralizedvotingsystemwidelyusedbyDAOsfor off-chaingovernance.SnapshotallowsDAOstocreateproposalsandvotingsystemsthatdonotincurgascosts, makingitapopularchoiceformanydecentralizedorganizations.SnapshotAPIcanbequeriedtogatherproposal andvotingdata.Table1blistsaquerycompatiblewithSnapshotAPIbasedbyGraphQLfortheSushiDAO. Torunthisquery,youcanuseaJavaScriptcodewithAxios,apromise-basedHTTPClientfornode.jsandthe browser.ThiswayyoucanfetchtheproposaldataandsaveittoaCSVfileoranyothersuitableformat. Snapshot data can also be queried using the snapshot.js node package. However, in this study, using the GraphQL-basedqueriesprovidedbytheSnapshothubservicefordatacollectionispreferred. 4.1.3 DataCollectionviaDAOContracts. DAOcontractscanbedirectlyqueriedtocollectgovernancedatasuch asvotingrecords,proposals,andmemberinformation.Byinteractingwiththesmartcontractdirectlythrougha blockchainnode,historicaleventdataemittedbythecontractcanbeextracted.Thismethodwasappliedfor severalDAOstoretrievespecificgovernancedata. Table1ehasasimplifiedexampleofqueryingTheLAODAOcontracttoretrievepastvotingevents.This exampleusesWeb3.jstointeractwiththeblockchain. Usingthesemethods,thestudycollectedcomprehensivedataonDAOgovernancebehavior,ensuringarobust datasetforanalysis. 3Datauptothe4thweekofJune2025wasincludedinthegovernanceanalysis. Distrib.LedgerTechnol.

10 • Jungnickeletal. Table2. SummaryofTheExploratoryAnalysis CrossRefer- AnalysisName DAOs #of DataUsed DataOrigin ence DAOs Figure1a,Figure GiniDistributionAcross AllActiveDAOhaus 503 MembersDataforAllDAOshavingDAOID,member On-chain 1b DAOs,GiniDistribution DAOs ID,MemberShareConvertedtoaGiniDataHaving AcrossDAOMemberSize DAOID,MemberSize,GiniCoefficient Figure2a,Figure GiniDistributionAcross AllActiveDAOstack 211 MembersDataforAllDAOshavingDAOID,member On-chain 2b DAOs,GiniDistribution DAOs ID,MemberShareConvertedtoaGiniDataHaving AcrossDAOMemberSize DAOID,MemberSize,GiniCoefficient Figure3a LorenzCurves DxDao,Lao,Meta, 7 HoldersData* On-chain Moloch,Maker, Uniswap,Curve Figure3b,Figure ChangeofGiniCoeffi- DxDao,Lao,Meta, 6 HistoricalBalances(Shares)Data On-chain 3c cientinTime,Changeof Moloch,Uniswap, NakamotoCoefficientin Curve Time Figure3d VotingPowerPieCharts DxDao,Lao,Meta, 7 HoldersDatawhichisMemberswithShares/Balances> On-chain Moloch,Uniswap, 0 Curve,Maker Figure4a On-chainVotingFrequency DxDao,Curve, 6 MembersData,VotingData On-chain Uniswap,Lao, MetaCartel,Moloch Figure4b On-chainvsOff-Chain Curve,Uniswap, 3 MembersData,VotingData On-chainand VotingFrequency MetaCartel Off-Chain Figure5a,Figure MajorityvsTurnout,Major- AllSnapshotDAOs 9725 ProposalDataHavingProposalID,VoteCounts,Pro- Off-chain 5b,Figure5d ityvsSize,MajoritySize posalResults Figure5c,Figure MajorityvsTurnout(Sushi), Sushi 1 ProposalDataHavingProposalID,VoteCounts,Pro- Off-chain 5e MajoritySize(Sushi) posalResults Figure6 StakeMajorityvsShare AllActiveDAOstack 211 ProposalDataHavingProposalID,VoteCounts,Pro- On-chain Majority(DAOstack) DAOs posalResults Figure7a,Figure ControversyAnalysisVote AllActiveDAOhaus 503 ProposalDataHavingProposalID,VoteCounts,Pro- On-chain 7d,Figure7g vsMajorityforDAOhaus DAOs posalResults Proposals,Controversy AnalysisTurnOutvs MajorityforDAOhaus Proposals,Controversy Analysis%ofProposals vsMajorityforDAOhaus Proposals Figure7c,Figure ControversyAnalysisVote Aave,Uniswap 2 ProposalDataHavingProposalID,VoteCounts,Pro- On-chain 7f,Figure7i vsMajorityforSelected posalResults ProtocolDAOs,Contro- versyAnalysisTurnOutvs MajorityforSelectedPro- tocolDAOs,Controversy Analysis%ofProposals vsMajorityforSelected ProtocolDAOs Figure7b,Figure ControversyAnalysisVote AllActiveDAOstack 211 ProposalDataHavingProposalID,VoteCounts,Pro- On-chain 7e,Figure7h vsMajorityforDAOstack DAOs posalResults Proposals,Controversy AnalysisTurnOutvs MajorityforDAOstack Proposals,Controversy Analysis%ofProposals vsMajorityforDAOstack Proposals Distrib.LedgerTechnol. DAOGovernance:VotingPower,Participation,andControversy-AReviewandanEmpiricalAnalysis • 11 4.1.4 DataLimitations. Theexploratoryanalysisfacedseveraldata-relatedchallenges: • Heterogeneity:Governancedatastructuresvaryacrosscontractsandsubgraphs,complicatingretrievaland limitingdirectcomparability. • Availability:Onlyon-chaindatastoredoremittedbycontractswasaccessible.Duetogascostconcerns,this dataisoftenminimalandinconsistentacrossDAOs. • SelectionBias:DataavailabilityinfluencedDAOselection,potentiallyskewingthesampletowardDAOs withmoretransparentdata. 4.2 Methodology Thevotingpowerofanaddressreferstotheweightattributedtoavotecastfromthataddress.TheGinicoefficient wastheprimarymetricusedtomeasurethedegreeofinequalityinvotingpower.Lorenzcurveswereusedto displaythefractionofthetotalvotingpower(y-axis)heldbyagivenfractionoftheDAO’smembers(x-axis). Sinceneitherofthesemetricsrevealstheimpactofinequalityondecision-making,theNakamotocoefficient wasalsocalculated.Thiscoefficientistheminimalnumberofmembersrequiredforanabsolutevotingmajority withintheDAO.Onlyon-chaindatawasusedsincemembers’off-chainvotingweightistypicallythesame. (cid:40) • VotingPower:𝑉𝑃(𝑚) = share𝑚 ifDAOusesshares balance𝑚 otherwise • GiniCoefficient:𝐺 =1− (cid:205)𝑛 𝑖=1 ( 𝑛 𝑥 − 𝑖+ 1 𝑥𝑖−1 ) ,where𝑥 𝑖 isthecumulativevotingpowerofthe𝑖𝑡ℎ member(sorted), and𝑛isthenumberofmembers. • LorenzCurve:𝐿(𝑝) = ∫ 0 𝑝𝑥(𝐹−1(𝑢))𝑑𝑢 ,representingthecumulativevotingpowerheldbythebottom𝑝%of ∫1𝑥(𝐹−1(𝑢))𝑑𝑢 0 members. (cid:16) (cid:17) • NakamotoCoefficient:𝑁 =min𝑘 (cid:205)𝑘 𝑖=1 𝑆 𝑖 ≥ 51% ,where𝑆 𝑖 isthevotingshareofmember𝑖;𝑁 denotesthe smallestnumberofmemberscontrolling51%ofvotingpower. SeveralDAOsacrossallthreegovernancemodelswereselectedforatemporalanalysis;temporaldatawasnot availableforallDAOsanalyzed.SincetheDAOshaveexistedfordifferentamountsoftime,thetimeintervals werenormalized.Asaresult,therateofchangeofthecoefficientsisnotcomparableacrossDAOs.Theearliest measurementsforeachDAOaretakenapproximatelyonemonthaftercreationandthelatestmeasurements are fromthe 30th of July 2022. Measurementswere taken at quarterly intervals, meaning not all temporary fluctuationsaredisplayed. TwometricswereusedtomeasureparticipationinDAOgovernance;thefrequencywithwhichmembersvote andvoterturnout.DuetothelimitedavailabilityofdataonlyasubsetoftheDAOsiscompared.Bothonand off-chainvoteswereconsideredtoofferadirectcomparison. • VotingFrequency:votingFrequency(𝑚𝑒𝑚𝑏𝑒𝑟 𝑖) =NumberofVotesbymember𝑖 • NormalizedFrequency:normalizedFrequency(𝑚𝑒𝑚𝑏𝑒𝑟 𝑖) = votingFrequency(𝑚𝑒𝑚𝑏𝑒𝑟𝑖) TotalNumberofProposals • ParticipationDistribution(StackedBar):Normalizedvotingfrequenciesaregroupedintointervals(e.g., 0–20%,20–40%,etc.).Foreachbin𝑏,theproportionofmembersis:weight(𝑏) = #MembersinBin𝑏 TotalMembers Theseproportionsarevisualizedasstackedbars,whereeachsegmentreflectsonebin. • VoterTurnout:Foreachproposal𝑝,voterTurnout(𝑝) = #VotedMembers TotalMembers Controversyisdeterminedbytwofactors:thedivergenceofvoteropinionsandtheimportanceattributedtothe decision.Majoritysize(i.e.percentageofvotesforthewinningoption)capturesdivergence,whilevoterturnout reflectsimportance.Highlycontroversialdecisionsareexpectedtoshowlowmajorities(highdisagreement)and highturnout(highimportance).Bothon-andoff-chainvoteswereanalysedforcomparison. Distrib.LedgerTechnol. 12 • Jungnickeletal. • MajoritySize:MajoritySize(𝑝) = #VotesforWinningOption TotalVotes • VoterTurnout:Ratioofvotingmemberstototalmembers:VoterTurnout(𝑝) = #MembersWhoVoted TotalMembers • Controversy Score: Defined as ControversyScore(𝑝) = (1−MajoritySize(𝑝)) ×VoterTurnout(𝑝); higher valuesindicatemorecontroversialproposals. • Controversy Hexbin Plot: Shows vote density based on turnout (x-axis) and majority size (y-axis): HexbinDensity=Density(Turnout,MajoritySize) • NumberofProposalsHistogram:ShowshowfrequentlydifferentproposalcountsoccuracrossDAOs: Histogram(𝑥) =Frequency(NumberofProposals) • Turnout vs Majority: Relationship between turnout and majority size: TurnoutvsMajority = (Turnout,MajoritySize) • TurnoutbyDAOSize:EachpointshowsaDAO’ssizeanditscorrespondingturnout:(DAOSize,Turnout) • ControversybyDAOSize:Plots(DAOSize,MajoritySize)toassesscorrelation. • Turnout by DAO Proposals: Relationship between number of proposals and turnout: TurnoutbyDAOProposals= (NumberofProposals,Turnout) • Controversy by DAO Proposals: Relation between number of proposals and majority size: ControversybyDAOProposals= (NumberofProposals,MajoritySize) 5 REVIEWRESULTS 5.1 SummaryoftheReviewResults SeveralquantitativeanalysesofDAOgovernancebehavioralreadyexist,butmuchoftheresearchispiecemeal, andlackscomparativeanalysesacrossmultiplegovernancemodels,andgenerallyfailstoconsideroff-chain voting.Table3providesanoverviewofthemainacademicpapersthateithercontainquantitativeanalysesof DAOgovernance,proposemethodsforconductingsuchanalyses,orhighlightothergovernanceissues.The subsectionsbelowdescribehowthispaperseekstobuilduponandextendtheexistingresearchacrossthethree governancedimensionsbeinginvestigated. 5.1.1 VotingPower. WhilemuchresearchhasexploredcentralizationinDAOgovernance,ithaslargelyfocused on token-based models. There is broad consensus that such models exhibit high inequality and centralized decision-making[56,35,33,6],thoughthepreciseextentremainsdebated[41].Thisstudyexpandsonexisting workbyincludingDAOstack[22]andDAOhaus[21]models,aimingnottoresolvedisputesovertokengovernance buttorankinequalityacrossmodelsandexamineitsdrivers.Rikkenetal.[47]analyzevotingweightbasedon 6000DAOprojects,findinghigherviabilityinthoseusingweightedvoting.Axelsenetal.[5]proposeabroader framework,TIGER,toassesscentralizationthroughfactorsliketoken-weighting,infrastructure,governance, escalation,andreputation. 5.1.2 Participation. Voterparticipationhasalsobeenexaminedseveraltimes,includingtwocomparativeanalyses ofgovernancemodelsbyRikkenetal.[48]andFaqir-Rhazouietal.[26].Theirresearchshowsdisparitiesin thedegreeofparticipationacrossvotersandamongDAOs.Themajorityofoverallactivityhappensinasmall minorityofDAOsbyasmallminorityofmembers.Typically,memberswithmorevotingpowerparticipatemore actively.ThisstudyseekstoreplicatetheseresultsacrossaslightlydifferentselectionofDAOsandextendsthe researchbyincludingacomparativeanalysisofonandoff-chainvoting.Amongothergovernanceproblems,lack ofparticipationismentionedbyPereiraandGarcia[45]butnoevaluationmethodispresented. 5.1.3 Controversy. Controversyhasbeenstudiedtheleastamongthegovernancedimensionsexaminedinthis research.Whileseveralstudieshavefoundlargemajoritysizes,especiallyintoken-basedprotocolDAOs(e.g. Fritschetal[29]),moreanalysisisneededontherelationshipsbetweenmajoritysizeandotherindicatorsof Distrib.LedgerTechnol. DAOGovernance:VotingPower,Participation,andControversy-AReviewandanEmpiricalAnalysis • 13 controversy,suchasvoterturnoutandthevotermajority4.Theexistingresearchalsodoesnotfocusenoughon comparingdifferentgovernancemodelsandoffvson-chainvoting.Thisstudyseekstoaddressthesegaps. Governanceissuesthatdirectlyorindirectlyimpactvotingpowerdistribution,participation,andcontroversy includelegalambiguities,rigidity,andvotingmisconduct,asnotedbyPereiraandGarcia[45].Similartothis Bellavitisetal.[7]alsomentionedregulatoryuncertaintyasoneoftheshortcomingsofDAOSbygivingsamples frombothinandoutsideofUSA.Beyondoperationalconcerns,thelong-termsustainabilityandadoptionofthese governancemodelsremainpersistentchallenges[46].Bellavitisetal.[7]alsoempiricallystudytheeconomy andevolutionofDAOs,bothofwhicharedirectlytiedtotheirsustainability.Whileempiricalstudieshelpto understandthecurrentsituationandtoplanforthefuture,benefitingfromlong-knownmanagementtheorems suchasagencyandstewardshiptheoremsisalsoveryusefulasusedbyAlawadietal.[1].Someotheruseful theoremsincludetransactionalcosttheory[69],sociomaterialitytheory[50]. 5.2 ExploratoryDataAnalysisResults 5.2.1 VotingPower. Theresultsshowthattoken-basedprotocolDAOsexhibitsignificantlyhigherinequalityin votingpowerdistributioncomparedtoDAOhausandDAOstack.ThemedianGinicoefficientsforthesethree groupsare0.98,0.75,and0.46,respectively.TheGinicoefficientdistributionsforDAOhausandDAOstackare showninFigures1aand1b,andFigures2aand2b,respectively.ThesefiguresillustratethattheDAOhausmodel isassociatedwithgenerallyhighlevelsofinequalityinvotingpower,asnearlyhalfoftheDAOsexhibitaGini coefficientexceeding0.8.Incontrast,theDAOstackmodeldemonstratesamuchbroaderdistribution,withthe majorityofDAOsmaintainingGinivaluesbelow0.6. Several DAOs were selected for a more detailed comparative analysis. Figure 3a presents Lorenz Curves comparingvotingpowerinequalityacrossrepresentativeDAOsfromeachgovernancemodel.Thecurvefor protocolDAOs(dashed-dotted)deviatesmostfromtheequalityline,indicatingthemostunequaldistributionof votingrights.DAOhaus(dashed)andDAOstack(dotted)DAOsfollowbehind,respectively. Tounderstandhowinequalityevolvesovertime,Figure3btracksthechangeinGinicoefficientsoverthe normalized lifetime of selected DAOs. For most DAOs, the Gini value remains relatively stable, with minor fluctuationsobservedintheearlyphasesofactivity.Thepatternconfirmsthatinequalitytendstobepersistent ratherthantransient. Figure3cdisplaystheNakamotocoefficientsovertimeforthesameselectionofDAOs.Althoughnoneofthe DAOsreachaNakamotocoefficientabove16,themetricdoesfluctuate,reflectingchangesinconcentrationlevels amongthemostpowerfulmembers. Inotherwords,asmallbutchangingminorityofmembersisinfullcontroloftheseDAOs.Figure3dprovides abreakdownofvotingpowerbyaddressinselectedDAOs.Thechartshighlightthatevenwithinthesubsetof dominantvoters,powerishighlyconcentrated.Itshouldbenotedthatthisanalysisconsidersvotingaddresses ratherthanindividuals.Asingleindividualmaycontrolmultipleaddresses,oroneaddressmayrepresentasmart contractoperatedcollectively.Forinstance,theCurveDAO’sprimaryvotingaddressbelongstoConvexFinance, meaningtheactualdecision-makingpowerlieswiththemultisigsignersbehindthatprotocol5. time. 4Thevotermajorityreferstothefractionofmemberswhovotedinfavorofaproposal.SinceDAOsgenerallydonotgiveeachmemberthe samevotingweight,alargemajorityofvotingweightcastinfavorofadecisiondoesnotnecessarilymeanamajorityofvotersarealsoin favor. 5WhileConvexholdsvotesonwhethertosupportCurveFinanceproposals,thesevotesarenonbindingandoff-chain.Theyareimplemented atthediscretionofthesignatoriesoftheDAO’smultisigwallet.ThismeansthemajorityofvotingpowerinCurveisnotheldbyone individual,butasmallgroupofindividuals.Theexampledemonstratesthatthedistributionofvotingpoweracrossindividualscanonlybe determinedbyanalyzingeachaddressandreviewingthegovernancesystemcontrollingit;thedifficultyofsuchananalysismeansthatthe precisedistributionremainsunknowninmostmid-to-largesizedDAOs. Distrib.LedgerTechnol. 14 • Jungnickeletal. (a)GiniAcrossDAOs (b)GinibyDAOSize Fig.1. DAOhausPowerDistribution (a)GiniAcrossDAOs (b)GinibyDAOSize Fig.2. DAOstackPowerDistribution 5.2.2 Participation. Figures4aand4billustratememberparticipationdistributionsforaselectionofDAOs, showingthefractionofmembersfallingintofivevotingactivitybrackets.Inthecaseofon-chainparticipation (Figure4a),theoverwhelmingmajorityofmembersinallDAOsparticipatedinfewerthan20%ofproposals.Only asmallfractionofmembersreachedhigherparticipationlevels,andparticipationabove60%israreacrossall cases. Three of the DAOsalso use Snapshot for off-chain voting, which allowsfor a direct comparison between thetwomechanisms.AsshowninFigure4b,off-chainparticipationfollowsasimilarpattern:mostmembers engageinlessthan20%ofproposals,withonlyafewexhibitingconsistentvotingbehavior.However,some Snapshotplatforms,suchasMetaCartelSnap,showslightlymoreactivevotersubsetscomparedtotheiron-chain counterparts. VoterturnoutperproposalisdiscussedseparatelyinSection5.2.3. Distrib.LedgerTechnol. DAOGovernance:VotingPower,Participation,andControversy-AReviewandanEmpiricalAnalysis • 15 (a)LorenzCurves (b)GiniCoefficient (c)NakamotoCoefficient (d)SharesofVotingPower Fig.3. InequalityinselectedDAOs 5.2.3 Controversy. ThissubsectionpresentstheupdatedanalysisofproposalcontroversyinDAOgovernance votes,basedonthenewlyrefinedvisualizationinFigure5.Thefigurecapturesthedistributionofgovernance proposalsfromSnapshot-basedoff-chainvotingsystems,acrossseveralDAOs. InFigure5a,theconcentrationofproposalsinthetop-leftcornerhighlightsacommonpattern:mostproposals receivelowvoterturnoutyetexhibitoverwhelmingmajorities.ThisindicatesthatinmanyDAOs,asmallportion ofmembersdeterminetheoutcomewithlittleopposition.Theincludedtrendlineconfirmsanegativecorrelation between turnout and majority size—higher participation levels are generally associated with closer results. Figure5dfurtherreinforcesthisobservationbydisplayingthedistributionofmajoritysizesacrossallSnapshot proposals,withamedianmajorityof90.49%,suggestinglimitedcontroversyinmostcases. ToexaminetheroleofDAOsizeinthesedynamics,Figure5bmapsmajoritysizeagainstDAOsize.Themajority ofdatapointsareagainclusteredtowardlowparticipationandhighconsensus.Thetrendlineremainsrel atively flatacrossDAOsizes,indicatingthatevenasDAOsgrow,thepatternoflowturnoutandstrongagreement persists—suggestingthatscalealonedoesnotsubstantiallyincreasedisagreementorcontroversy. SushiDAOwasanalyzedseparatelyasitisoneofthefewlargeDAOsthatrelyexclusivelyonoff-chainvoting. Theaimwastoassesswhethervotingbehaviorinpurelyoff-chainsystemsdeviatesfromDAOsthatuseahybrid model.AsshowninFigure5e,thedistributionofproposalcontroversyinSushiDAOcloselyresemblestheoverall Snapshotdistribution(Figure5d),indicatingnomajordeviationincontroversypatterns. Distrib.LedgerTechnol. 16 • Jungnickeletal. (a)On-ChainFrequency (b)On-ChainvsOff-Chain Fig.4. ParticipationviaVotingFrequency (a)Maj.vsTurnout (b)Maj.vsSize (c)Maj.vsTurnout(Sushi) (d)MajoritySize (e)MajoritySize(Sushi) Fig.5. ControversyinOff-ChainVoting Thecontroversiesinon-chainvotingacrossDAOhaus,DAOstack,andtheselectedprotocolDAOsaredepicted inFigure7.Theseresultssuggestthatcontroversyinon-chainvotingmaybeslightlylowerthaninoff-chainvoting, withmedianmajoritysizesof98.28%,97.85%,and96.74%respectively.Inallthreecategories,mostproposals receiveaconsistentlylargemajorityevenwithvaryingparticipationlevels.AlthoughtheprotocolDAOshave Distrib.LedgerTechnol. DAOGovernance:VotingPower,Participation,andControversy-AReviewandanEmpiricalAnalysis • 17 Fig.6. StakeMajorityvsShareMajority(DAOstack) thelowestmedianmajorityamongthethreegroups,proposalsonDAOhausandDAOstackdemonstrateawider spreadinmajoritysizes,indicatingmorevariabilityinvotingoutcomes. Sincetheaboveresultsconsidermajoritysizeintermsofvotingpower(i.e.,weightedvotes),theymayobscure thepresenceofcontroversyamongindividualvoters.Membersoftenholddifferentvotingweights,meaninga majorityofthevotingpowermaysupportthewinningoutcomeevenifmostvotersopposeit.Toassessthis potentialdiscrepancy,Figures7ato7eplotthesharemajority(fractionoftotalvotingpower)againstthevote majority(fractionofindividualvoters).Inallthreediagrams,alargeproportionofobservationsareclustered nearthetop-rightcorner,reflectingnear-unanimousoutcomeswherebothvoteandsharemajoritiesareclose to100%.Asaresult,thedegreeofdivergenceislesspronouncedthanitmightinitiallyappear.Thetrendlines confirmthatvoteandsharemajoritiesgenerallyincreasetogetherinproportion.Themeanvotemajoritiesfor DAOhaus,DAOstack,andtheprotocolDAOsare98.09%,97.38%,and96.74%,respectively.Thepercentageof proposalswheretheoutcomewouldhavedifferedifbasedonvotemajorityratherthansharemajorityis0.91% forDAOhaus,11.01%forDAOstack,and1.58%fortheprotocolDAOs. The results for DAOstack additionally include an analysis of the staking majority relevant for the used exclusivelyinDAOstackproposals.In77.3%ofproposals,thereisastakemajorityinfavourofthewinning outcome;however,Figure6showsthatthesizeofthestakemajoritydoesnottendtoincreaseproportionately withthesharemajority.Thisgraphhelpsassesstheeffectivenessoftheholographicconsensusmechanismin DAOstack.Thelackofahighcorrelationbetweenstakemajorityandsharemajorityindicatesthattheboosting mechanismmaynotbeeffectivelypredictingandinfluencingproposaloutcomes. 6 DISCUSSION TheresultsoftheexploratorydataanalysisleadtothreemainconclusionsaboutDAOgovernance: • Centralization:Traditionaltokengovernanceshouldbeavoidedduetoitstendencytofavourhighlycentral- izationvotingpowerdistributions. • LowParticipation:Participationingovernancedecisionsislow,somembersshouldinsteadelectdecision- makerstoconductgovernanceforthem. • LackingControversy:Governanceproposalsareoftenuncontroversial,somakingthemsubjecttovotingis inefficient;instead,governanceshouldfocusonfewer,morecontroversialdecisions. Thefollowingsubsectionswilladdressthesepointsinturn;eachsubsectioncontainsadiscussionoftheproblem andanevaluationofitsimpact. Distrib.LedgerTechnol. 18 • Jungnickeletal. Table3. SummaryofRelatedWork Authors Topic Findings DAOs Fritschetal.[29] Governancebehavior,inequality, Unequalvotingpower,lowdelegation,infrequent ENS,Uniswap, votedelegation participation Compound Rikkenetal.[48] Governancesystempopularityand Aragonmostpopular,DAOhausandDAOstackfollow; Aragon,DAOhaus, activity mostDAOsinactive,fewhighlyactive DAOstack ElFaqiretal.[24] Viabilityofholographicconsensus BoostingpredictsproposalsuccessinlargeDAOs DAOstack FaqirRhazouiet EffectofgaspricesonDAOactiv- Nosignificantimpactonvotingbehavior DAOhaus,DAOstack al.[28] ity FaqirRhazouietal. PopularityandactivityinDAO Aragonmostpopular,DAOhausandDAOstackfollow; Aragon,DAOhaus, [26] governancesystems lowactivity DAOstack Sunetal.[56] Centralizationandvotingactivity Highinequality,lowparticipation,largeholdersdomi- Maker inMaker nate Stroponiatietal.[35] Centralizationintokengover- Highinequality,decisionsbysmallminority;security Maker,Curve,IDEX, nance risks Compound,Uniswap Jensenetal.[33] Inequalityintoken-basedgover- Highinequality,powerfulminoritydominance Yearn,Compound, nance Uniswap,Balancer NadlerandSchar Wrappedownershipofvoting Lesscentralizationthanappears;largetoken-holders Maker,Compound, [41] power oftencontractscontrolledbysmallerowners Sushi Barbereauetal.[6] Inequalityintoken-basedgover- Highinequalityduetoinitialdistributionandunlimited Uniswap,Aave, nance votingpowerpurchase Maker,Compound Rikkenetal.[47] ParametersaffectingDAOviability DAOswithoutweighteddecision-makingorincentives 6000initialDAO aremoreviable projects Axelsenetal.[5] DAOcentralizationmeasurement Proposesaframeworktomeasurecentralizationin UsesCompoundto DAOgovernance evaluateframework PereiraandGarcia DAOgovernanceissuesandfuture Proposesfuturestudytopics,e.g.,adoptingdecentral- - [45] topics izedgovernance,communityretention Peña-Calvinetal. FrameworktocategorizeDAOs CategorizesDAOsbydomain,purpose,votingprocess, 40DAOsusing [44] crypto-tokenuse;identifiesarchetypesinAragon Aragon Pinioetal.[46] ComparativeanalysisofDAO ProvideskeyinsightsintoDAOgovernancebutfocuses 10leadingDAOs governancemetrics onalimitedsetofmetrics,which,whilevaluable,may (unnamed) notfullycapturethecomprehensivelandscapeofDAO governancedynamics Bellavitisetal.[7] DAOs’challenges,opportunities, HighlightsDAOs’marketpotential,providesempirical 2300activeDAOs empiricalanalysis governanceanalysis fromDAOAnalyzer AAlawadietal.[1] SurveyonDAOgovernancesys- DAOsalignedwithstewardship,butoperationslean Participantsfrom20 tems towardagencyperspectives DAOs 6.1 CentralizationinTokenGovernance AllgovernancemodelsinvestigatedproducedDAOswithhighlyunequalvotingweightdistributions,leadingto centraliseddecision-making,consistentwithexistingresearch[56][35][33][6].Thisanalysisexploresdifferences incentralizationlevelsacrossmodelsandunderlyingcauses.Token-basedprotocolDAOsconsistentlyshowed thehighestGinicoefficients,followedbyDAOhausandDAOstack.Itisarguedthattoken-basedsystemsare morepronetoinequalityduetothetradability,rewardfunction,andunequalinitialallocationoftokens.The relativelylowercentralizationinDAOhausandDAOstackhighlightsgovernancefeaturesthatmayreducepower imbalances. Amajordriverofinequalityintoken-basedgovernanceistheabsenceoflimitsonpurchasingvotingrights. Here,votingpowerderivesfromfreelytradeabletokens,enablingindividualstobuyorborrowlargeamounts foroutsizedinfluence.Incontrast,DAOhausandDAOstackimplementstructuralsafeguards:DAOhausrequires memberapprovaltoissuevotingshares,whileDAOstackgrantsnon-transferable,earnedreputationscores. Thesemechanismsshowhowlimitingtransferabilityandaccumulationcanreducecentralization. Distrib.LedgerTechnol. DAOGovernance:VotingPower,Participation,andControversy-AReviewandanEmpiricalAnalysis • 19 (a)VotevsShare(DAOhaus) (b)VotevsShare(DAOstack) (c)VotevsShare(ProtocolDAOs) (f)MajorityvsTurnout(Protocol (d)MajorityvsTurnout(DAOhaus) (e)MajorityvsTurnout(DAOstack) DAOs) (g)MajoritySize(DAOhaus) (h)MajoritySize(DAOstack) (i)MajoritySize(ProtocolDAOs) Fig.7. ControversyinOn-ChainVoting Intoken-basedDAOs,thetreatmentofgovernancetokensasfinancialassetscreatesstrongeconomicincentives formemberstoaccumulatevotingpower,therebyamplifyingtheriskofcentralization.Governancetokensof majorprotocolDAOs—suchasMKR,CRV,andUNI—havedeliveredsubstantialinvestmentreturns,makingthem attractiveevenbeyondtheirutilityingovernance.AsStroponiatietal.[35]highlight,suchtokensareoften acquirednotonlytoexertinfluencewithintheDAO,butalsoasspeculativeassets.Thesedynamicsareless pronouncedinDAOhausandDAOstack.DAOhausdistinguishesbetweenvotingsharesandeconomicshares, Distrib.LedgerTechnol. 20 • Jungnickeletal. partiallydecouplinggovernanceinfluencefromfinancialinvestment.Thismeansthatindividualsinvestingfor financialgaindonotnecessarilygainequivalentvotingrights.DAOstackfurthermitigatesthisissuebyassigning votingpowerthroughnon-transferablereputationscores,whichcarrynomonetaryvalueandcannotbetraded, therebyavoidingdirectalignmentbetweenfinancialincentivesandgovernanceinfluence. Manytoken-basedDAOsusetheirgovernancetokenstorewardcontributors;theinitialallocationofthe governancetokennormallyincludessignificantgrantstofounders,employeesandventurecapitalinvestors, grantingthemadisproportionatelylargeandpersistentshareofgovernancepower[35][33]Forexample,40%of theinitialsupplyoftheUniswapgovernancetokenwasdistributedinthismanner[62].TypicallyDAOcreators arguethatthecentralizationtheyintroducedthroughtheinitialallocationwillnaturallydeclineovertimeasmore membersjointhecommunity[62],buttheresultsofthedataanalysissuggestthattheGinicoefficientremains stable throughout the DAO’s lifecycle. This is consistent with existing studies on inequality in token-based governance[35][29].ThetemporalanalysisoftheNakamotocoefficientsrevealsnoconsistenttrendtowards decentralizationacrosstoken-basedDAOs.WhileUniswap’scoefficientincreasesmodestlyfrom9to10,this changeisnegligiblegiventheDAO’slargememberbase.Curve,ontheotherhand,initiallyrisesfrom13to 16butsubsequentlydropssharplyto1,indicatingamovetowardsgreatercentralization.dxDAOpresentsa morestructuredpattern,withagradualstepwiseincreaseinitscoefficient,suggestingthatdecentralizationmay beachievableundercertaingovernanceconditions.Thesefindingsimplythatdecentralizationdoesnotoccur naturallyovertimeandmayevendeteriorateinsomecases.Jensenetal.[33]comparethecentralizationof severaltoken-basedDAOs,someofwhichemployedafair-launchstrategyintendedtopreventanunequalinitial distribution[33].Theyobservethatdespitesuchefforts,inequalitylevelsintheseDAOsrapidlyconvergedwith thoseofmoretraditionallylaunchedprojects.Thissuggeststhatalthoughafairinitialallocationmaynotbe sufficientforachievinglong-termdecentralization,itcouldstillbeanecessaryfoundation. SinceDAOhausandDAOstackdonotusegovernancetokens,theyarelessvulnerable—butnotimmune—to unequalinitialdistribution.DAOhausseparateseconomicandvotingshares,allowingfoundersandinvestors toreceivefinancialrewardswithoutgainingexcessivegovernancepower.However,applyingthisseparation dependsonthecommunity.Similarly,DAOstack’sreputationisintendedtobeearnedovertime,yetthisisnot alwayspracticed.Forinstance,dxDAOgrantedinitialreputationbasedontokenholdings,possiblyexplainingits notablyhigh 6.2 ImpactofCentralizationinTokenGovernance ConsideringthatthehighlycentralisedprotocolDAOsincludedintheanalysisareverysuccessful–moreso thananyDAOstackandDAOhausproject–thequestionarises,whydoesdecentralizationmatter?Itwillbe arguedinthissectionthatthecentralizationofvotingpowerintokengovernanceraisesconcernsaboutsecurity andminorityrights.TheseconcernsareespeciallypertinentfornewandsmallerDAOprojectsmeaningthey deserveattentionirrespectiveofthesuccessofmajorprotocolDAOs. 6.2.1 SecurityRisks. Theabsenceofstrongprotectionsagainstcentralizationintoken-basedgovernancesystems resultsinconsiderablesecurityrisks,whichhaveonlybeenpartiallymitigated.Token-basedgovernanceallows individualswithsufficientresourcestopurchasevotingmajorities.Stroponiatietal.notethatthecostofacq uiring amajoritysharemaysometimesbelowerthanthevalueoftokenstowhichthegovernancecontracthasaccess. Atthetimeoftheirwriting,thecostofacquiringavotingmajorityinMakerDAOstoodat$44MandtheDAO treasurywasvaluedat$2B[35]. Thecostofsuchanattackcanbefurtherreducedifgovernancetokensareborrowedinsteadofbought(e.g. theBeanstalkHack[52]).Centralizationalsoincreasestheriskofbriberyandcollusion.Sincefewermembersare requiredtoattainatokenmajority,fewerneedtobeconvincedtoparticipateinthescheme. Distrib.LedgerTechnol. DAOGovernance:VotingPower,Participation,andControversy-AReviewandanEmpiricalAnalysis • 21 DAOshavetriedtomitigatetheseproblems,byincreasingthehurdlesformaliciousactorstoacquiresignificant votingpower.Twomechanismshaveprovenparticularlypopular,butnoneareentirelysatisfactory: TokenLocking Thisapproachrequiresmemberstolocktheirgovernancetokensintoanescrowaccountin ordertoparticipateinvoting;whiletokensarelocked,thememberdoesnothaveaccesstothem.Forexample, inCurveDAOmemberscancommittolockingtheirtokensforuptofouryearsandthelongertheperiod,the largerthevotingpowertheyreceiveinreturn.Thismeansamalicioususerwouldhavetolockthegovernance tokenforapotentiallylongperiodoftime,preventingtheuseofshort-termloans,suchasflashloans[66].While thismechanismcertainlymakesitmoredifficulttoacquireamajority,itdidn’tpreventoneCurveDAOmember fromtemporarilyholding71%ofvotingpower[10].Therearealsosmartcontractsthatcanhelpbypassthese locks,allowingtheiruserstostilleffectivelytransferthelockedtokens[42]. ExecutionDelays TimedelaysafteraproposalhasbeenapprovedallowDAOmemberstoenactemergency procedurestosecurethefundscontainedinthetreasuryifamaliciousactormanagestoacquireamajority.Maker DAO,forexample,implementedthismechanismafterdiscoveringthesecurityriskdescribedabove[38].While thepossibilityofashutdowncandeterattackers,aDAOshouldnotrelyonthethreatofitsowndestructionas aprimarydefencemechanism;forexample,itmaynotwanttounwinditsentireoperationsduetoasmallor medium-sizedattack. Themainreasonmajortoken-basedDAOshavenotexperiencedmoresecuritybreachesislikelytheirsize and popularity. Their size means the investment needed to acquire a majority is large, and their popularity meansthatanyattackerwouldlikelyfaceconsiderablecountermeasuresandretribution.Theproblemsdescribed are thus most pertinent for new and smaller DAOs in which the cost of acquiring a voting majority can be significantlylower.Althoughthesolutionsadoptedbysometoken-basedDAOscertainlymitigatethesecurity risks,agovernancemodelinwhichtheydonotariseinthefirstplacewouldbepreferable. 6.2.2 Minority Rights. Token-based DAOs exhibit similar degrees of voting inequality as traditional public companies, yet offer none of the protections afforded to minority shareholders under corporate law. DAOs have often been described as decentralized alternatives to traditional organizations, that avoid hierarchical andundemocraticdecision-making[4][11][40][57].However,theNakamotocoefficientsofthethreelargest companiesintheS&ParehigherthanthoseofallprotocolDAOsthatwereanalysedhere,meaningtheyaremore decentralised.Whileafairlike-for-likecomparisonbetweenDAOsandtraditionalcompanieswouldrequirea muchlargerdatasetandmoredetailedanalysis,theresultsindicatethattheprevalentassumptionthatDAOs aredecentralisedalternativestotraditionalcompaniesisnotnecessarilytrue.Minorityshareholderrightswere introducedasanantidotetothecentraliseddistributionofvotingpowerinmanypubliccompanies;unlessthey becomemoredecentralised,DAOsshouldbeexpectedtodothesame.Minorityrightsareimportantbecausethe interestsofgovernancetokenholdersarenotnecessarilyaligned[9];thereareclearexampleswithinexisting DAOs,wherelargeshareholdersusedtheirvotingpowertoinfluencedecisionsthatservetheirowninterests[35]. 6.3 LowParticipation 6.3.1 ProblemofLowParticipation. Theresultsoftheexploratorydataanalysisreveallowvoterengagement across all governance models, including in off-chain voting. A minority of members participate actively in decision-making,whilethevastmajorityrarelycastsavote.Severalpotentialreasonsforlowparticipationwill beexploredinthissubsection,resultingintheconclusionthatDAOsshouldmovetowardsmorerepresentative andlessdirectdemocracy. Centralizedvotingpowerreducesengagementamonglessinfluentialmembers.Whilethisstudyfoundno stronglinkbetweendecentralizationandparticipationlevels,priorresearchshowsthatmemberswithhigher votingpowervotemorefrequently[29,56,6].Formemberswithlittleinfluence,thecostandeffortofparticipation mayoutweighthebenefitsofafavourableoutcome.If𝑝 isthechanceavoteisdecisiveand𝐵isthebenefitfrom Distrib.LedgerTechnol.

22 • Jungnickeletal. apositiveresult,theexpectedbenefitis𝑝×𝐵[9].Forsmallstakeholders,both𝑝 and𝐵arelow.Sincethecost𝑐 (e.g.,time,gasfees)isconstant,onlythoseforwhom𝑝×𝐵 >𝑐 willbemotivatedtoparticipate,regardlessofthe collectivebenefit.On-chainvotinginvolvesgasfeesandmaybeburdensomedependingontheinterface,butthere isnostrongevidencethatthisalonereducesvoterengagement.ToolslikeSnapshotwereintroducedtoeliminate gascostsandsimplifyparticipation;however,theanalysisshowsthatSnapshotvotingalsosuffersfromlow turnoutintheDAOsstudied.Thischallengestheassumptionthatcostisthemainbarrier.Supportingstudies[28, 26]foundnoclearlinkbetweengasfeesandparticipation,evenonlow-costchainslikexDAI.Thesefindings shouldbeinterpretedcautiously:Snapshotvotesarenon-binding,possiblylimitingmotivationtovote;gasprice studiesdidn’taccountforfiatexchangerates;andDAOsonmainnet—generallylargerandmoreactive—tendto showhigherparticipationthansmallerxDAI-basedDAOs[48]. While the causes of low voter participation remain somewhat unclear, the data tentatively supports the hypothesis that a combination of centralization and direct democracy—rather than cost—is the main driver. Centralizationcanmakemembersfeeltheirvotehaslittleimpact,whilefrequentproposalsindirectdemocracy increasetheeffortrequiredtostayengagedbeyondtheperceivedbenefit.Thenextsectiondiscussestheimpact oflowvoterengagement. 6.3.2 ImpactofLowParticipation. LowvoterparticipationincreasessecurityrisksinDAOs,underminesminority representation,andchallengesthecorerationalefordecentralization.TheNakamotocoefficientwasusedto illustratehowfewmembersareneededforabsolutemajoritycontrol;however,duetolowquorumthresholdsand relianceonsimplemajorities,farfewerparticipantscandetermineoutcomes.Forexample,inUniswap,just7.5% ofgovernancetokenscoulddecideallproposals[29].Whileexplicitattacksmightmobilisevoters,aconcealedor self-interestedactoroftenneedsfarlessthananabsolutemajority—makinglowparticipationakeyenablerof centralizationrisks. Sincelowparticipationisnotequallydistributedacrossusersbutmoreprevalentinlesspowerfulmembers [29],itdecreasestheextenttowhichtheinterestsofminorityshareholdersarerepresented.Buterinnotesthat alargenumberofsmallshareholderswillfindithardertoorganizethemselvestoadequatelyrepresenttheir interestingovernancethanasmallnumberoflargeshareholders[9].Lowparticipationreducestheprotectionof minorityinterests,whicharealreadylargelyunprotected. Finally, decentralization is intended to democratize decision-making, but if participation is very low, the benefitsofdemocratisationarenotfullyrealised.However,thedisadvantagesofdemocraticdecision-making, suchasthedelayintroducedbyvotingperiodsarepresentirrespectiveofthedegreeofparticipation. 6.4 Controversy 6.4.1 ProblemofLowControversy. Theanalysisshowedthatmostproposalshadlargemajoritiesandlowturnout acrossallgovernancemodels,indicatinglowcontroversy.Snapshotexhibitedslightlyhighercontroversy,possibly becausemajorDAOsuseittogaugesentimentbeforeon-chainvoting.Ifamajorityisclearoff-chain,opponents maychoosenottovoteon-chaintoavoidgascosts.However,thisremainsspeculativeandwarrantsfurther investigation. Theanalysisalsorevealedthatthemajorityofvotesarealmostalwayscastinfavourofthesameou tcome asthemajorityofvotingweight;inotherwords,ifeachvoterhadbeenweightedequallytheoutcomewould haverarelydiffered.Whilethisprimafaciesupportsthehypothesisthatvotingisgenerallyuncontroversial,the existenceofcontroversymaybeconcealedduetothehighdegreeofcentralizationinvotingpower.Ifalarge majorityofvotingweightiscastinfavourofanoutcome,memberswithlittlevotingpowermaynotvoteagainst it,asitwouldhavelittleimpact.Thiscangivethefalseimpressionthatnotjustthemajorityofweightbutalso themajorityofvotersfavouranoutcome.Theextenttowhichtheseconcernsarevalidrequiresfurtheranalysis. Distrib.LedgerTechnol. DAOGovernance:VotingPower,Participation,andControversy-AReviewandanEmpiricalAnalysis • 23 6.4.2 ImpactofLowControversy. Votingonuncontroversialproposalsiscostlyandinefficientbecauseeachvoter paysgasfeesandthefullvotingperiodneedstoelapseuntiltheproposalcanbeexecuted.Thecostanddelay ofcollectivedecision-makingcanreducethescalabilityandisparticularlyproblematicfornewDAOsintheir growthphase,wheredecisionfrequencycanbehigh.Moreover,frequentuncontroversialdecision-makingcan causevoterfatigueandfacilitatelowparticipation;ifvotersareconstantlyconfrontedwithsmall,uncontroversial decisions,theymayhavelesstimeandpatiencewhenmoreimportantproposalsarise. 7 CONCLUSION 7.1 SummaryofAchievements ThispaperaimedtoanalyseDAOgovernancebehaviorsthroughareviewofexistingliteratureandexploratory dataanalysis.Bycomparingdifferentgovernancemodels—bothonandoff-chain—thestudyhighlightedkey issuessuchascentralization,lowengagement,andlimitedcontroversy.Theseinsightsareintendedtoinformthe developmentofimprovedDAOgovernancedesigns. 7.2 Limitations Thischapterpresentsanexploratoryquantitativeanalysiswithoutformalstatisticalvalidation,anditsconclusions shouldbeinterpretedcautiously.Thefindingsofferhypothesesratherthandefinitiveclaimsaboutgovernance behavior.Thescopewaslimitedbydataavailability,andonlyasubsetofmetricsandDAOs—particularlyfor token-basedgovernance—couldbeincluded.Insomecases,onlyafewDAOswereanalyzedpergovernance dimension.Despitetheselimitations,theresultsprovidevaluableinsightstoinformthedesignoffutureDAO prototypes. 7.3 FutureWork Future work should seek to validate the results of the exploratory data analysis through formal statistical hypothesistesting.ItshouldalsoconsideradditionaldimensionsofDAOgovernancethatwereleftunexplored, suchassmall-groupgovernancebehavior.Finally,thedatausedwaslimitedbothinsamplesizeandquality; moreDAOsandvariationsongovernancemodelsneedtobereviewed. Inaddition,complementaryapproachesbeyondtoken-distributionmetricscouldbeconsidered.Forinstance, recent work by Fábrega et al. [25] introduces Voting-Bloc Entropy (VBE), a framework that assesses DAO decentralizationbymodelingvotingblocsandalignedparticipantbehaviors.Likewise,Sharmaetal.[51]illustrate howadvancedvisualizationandclusteringtechniquesappliedtoselectedDAOscanrevealfurtherdynamicsof decentralization.Extendingourworktoincorporatesuchbehavior-orientedmetricsandvisualizationmethods remainsanimportantavenueforfutureresearch. ACKNOWLEDGEMENTS ResearchsupportedbyBrevanHowardCentreforFinancialAnalysis. REFERENCES [1] AsmaAlawadietal.“Decentralizedautonomousorganizations(DAOs):Stewardshiptalksbutagencywalks”.In:Journalof Business Research178(2024),p.114672. [2] AlexRea,DanielKronovet,AronFischer,andJackduRose. Colonywhitepaper.[Online].Availablefrom:https://colony.io/whitepaper. pdf.[Accessed23thAugust2022].2022. [3] Aragon. ATaxonomyofVotingMethods.[Online].Availablefrom:https://blog.aragon.org/a-taxonomy-of-voting-methods.[Accessed 23thAugust2022].2021. [4] Aragon. Whatisadao.[Online].Availablefrom:https://blog.aragon.org/what-is-a-dao/.[Accessed23thAugust2022].2022. [5] HenrikAxelsen,JohannesRudeJensen,andOmriRoss.“WhenisaDAODecentralized?”In:arXivpreprintarXiv:2304.08160(2023). Distrib.LedgerTechnol. 24 • Jungnickeletal. [6] TomBarbereauetal.“DecentralisedFinance’sUnregulatedGovernance:MinorityRuleintheDigitalWildWest”.In:AvailableatSSRN (2022). [7] CristianoBellavitis,ChristianFisch,andPaulPMomtaz.“Theriseofdecentralizedautonomousorganizations(DAOs):afirstempirical glimpse”.In:VentureCapital25.2(2023),pp.187–203. [8] BinanceResearch. TheRiseofDAOs.[Online].Availablefrom:https://research.binance.com/static/pdf/rise-of-daos.pdf.[Accessed 20thAugust2022].2022. [9] VitalikButerin.“Movingbeyondcoinvotinggovernance”.In:Vitalik.ca(2021). [10] CastorAmyandPhillipsDaniel. Curvefounderseizes71%ofcurvedaovotingpower.[Online].Availablefrom:https://decrypt.co/ 39599/curve-founder-seizes-71-of-curve-dao-voting-power.[Accessed23thDecember2022].2022. [11] CathyHackl. Whataredaosandwhyyoushouldpayattention.[Online].Availablefrom:https://www.forbes.com/sites/cathyhackl/ 2021/06/01/what-are-daos-and-why-you-should-pay-attention/?sh=1c6773ec7305.[Accessed23thAugust2022].2021. [12] Kyung-nokChun,Su-YeonPark,andMoon-SooKim.“UnderstandingDecentralizedAutonomousOrganizations(DAOs)asaReaction toCorporateGovernanceProblems”.In:SmatoosBusinessReview(2022),p.33717. [13] Colony. ColonyDapp.[Online].Availablefrom:https://colony.io/features.[Accessed23thAugust2022].2022. [14] Compound. TryCompoundCommunity-builtinterfacesintegratingtheprotocol.[Online].Availablefrom:https://compound.finance/. [Accessed23thDecember2022].2022. [15] ConvexFinance. Convex.[Online].Availablefrom:https://www.convexfinance.com.[Accessed20thAugust2022].2022. [16] Coopha Turley. Dao landscape. Mirror. [Online]. Available from: https : / / coopahtroopa . mirror . xyz / _EDyn4cs9tDoOxNGZLfKL7JjLo5rGkkEfRa_a-6VEWw.[Accessed23thAugust2022].2021. [17] Coopha Turley. Dao landscape. Mirror. [Online]. Available from: https : / / coopahtroopa . mirror . xyz / _EDyn4cs9tDoOxNGZLfKL7JjLo5rGkkEfRa_a-6VEWw.[Accessed23thAugust2022].2021. [18] Curve. Curve.[Online].Availablefrom:https://gov.curve.fi/.[Accessed23thDecember2022].2022. [19] DanielKronovet. Aragon,DAOstack,Colony,Moloch.[Online].Availablefrom:http://kronosapiens.github.io/blog/2019/06/16/aragon- daostack-colony-moloch.html.[Accessed23thDecember2022].2022. [20] DAOHaus. BestPractices.[Online].Availablefrom:https://daohaus.club/docs/handbook/summoners/summoners-best-practices. [Accessed23thAugust2022].2022. [21] DAOHaus. Unlockthenexttierincommunitycooperation.[Online].Availablefrom:https://daohaus.club/.[Accessed04thJanuary 2023].2022. [22] DAOStack. DaoStackAnoperatingsystemforcollectiveintelligence.[Online].Availablefrom:https://daostack.io/wp/DAOstack- White-Paper-en.pdf.[Accessed23thAugust2022].2018. [23] QuinnDuPont.“Experimentsinalgorithmicgovernance:Ahistoryandethnographyof“TheDAO,”afaileddecentralizedautonomous organization”.In:Bitcoinandbeyond.Routledge,2017,pp.157–177. [24] YoussefElFaqir,JavierArroyo,andSamerHassan.“Anoverviewofdecentralizedautonomousorganizationsontheblockchain”.In: Proceedingsofthe16thinternationalsymposiumonopencollaboration.2020,pp.1–8. [25] AndresFabregaetal.“{Voting-Bloc}Entropy:ANewMetricfor{DAO}Decentralization”.In:34thUSENIXSecuritySymposium (USENIXSecurity25).2025,pp.1299–1318. [26] YoussefFaqir-Rhazoui,JavierArroyo,andSamerHassan.“Acomparativeanalysisoftheplatformsfordecentralizedautonomous organizationsintheEthereumblockchain”.In:JournalofInternetServicesandApplications12.1(2021),pp.1–20. [27] YoussefFaqir-Rhazoui,JavierArroyo,andSamerHassan.“Ascalablevotingsystem:Validationofholographicconsensusindaostack”. In:(2021). [28] YoussefFaqir-Rhazouietal.“Effectofthegaspricesurgesonuseractivityinthedaosoftheethereumblockchain”.In:Extended Abstractsofthe2021CHIConferenceonHumanFactorsinComputingSystems.2021,pp.1–7. [29] RobinFritsch,MarinoMüller,andRogerWattenhofer.“AnalyzingVotingPowerinDecentralizedGovernance:WhocontrolsDAOs?” In:arXivpreprintarXiv:2204.01176(2022). [30] StanleyGreenstein.“Preservingtheruleoflawintheeraofartificialintelligence(AI)”.In:ArtificialIntelligenceandLaw30.3(2022), pp.291–323. [31] JungsukHan,JongsubLee,andTaoLi.“AreviewofDAOgovernance:Recentliteratureandemergingtrends”.In:JournalofC orporate Finance(2025),p.102734. [32] JungsukHan,JongsubLee,andTaoLi.“Daogovernance”.In:(2023). [33] JohannesRudeJensen,VictorvonWachter,andOmriRoss.“HowDecentralizedistheGovernanceofBlockchain-basedFinance: EmpiricalEvidencefromfourGovernanceTokenDistributions”.In:arXivpreprintarXiv:2102.10096(2021). [34] WeiJiang,TaoLi,etal.“Corporategovernancemeetsdataandtechnology”.In:FoundationsandTrends®inFinance14.2(2024), pp.61–136. Distrib.LedgerTechnol. DAOGovernance:VotingPower,Participation,andControversy-AReviewandanEmpiricalAnalysis • 25 [35] KaterinaStroponiati,IlyaAbugov,YiannisVarelas,KostasStroponiatis,ModestaJurgeleviciene,andYashodaSavanthRaghunatha Rao. Decentralizedgovernanceindefi:Examplesandpitfalls.[Online].Availablefrom:https://dappradar.com/blog/decentralized- governance-in-defi-examples-and-pitfalls/.[Accessed23thAugust2022].2020. [36] ValerieLaturnus.“Theeconomicsofdecentralizedautonomousorganizations”.In:AvailableatSSRN4320196(2023). [37] Maker. MakerDAOAnUnbiasedGlobalFinancialSystem.[Online].Availablefrom:https://makerdao.com/.[Accessed23thDecember 2022].2022. [38] MakerDAO. Emergencyshutdownforpartners.[Online].Availablefrom:https://docs.makerdao.com/smart-contract-modules/ shutdown/emergency-shutdown-for-partners.[Accessed23thDecember2022].2022. [39] MartinKoppelmann. Abriefdiscussionoftheprotocolgov-¨erningthedxdao.[Online].Availablefrom:https://blog.gnosis.pm/a- brief-discussion-of-the-protocol-governing-the-dxdao-7331407a2555.[Accessed23thDecember2022].2022. [40] MoralisAcademy. Daovstraditionalorganization.[Online].Availablefrom:https://academy.moralis.io/blog/dao-vs-traditional- organization.[Accessed23thAugust2022].2021. [41] MatthiasNadlerandFabianSchär.“Decentralizedfinance,centralizedownership?aniterativemappingprocesstomeasureprotocol tokendistribution”.In:arXivpreprintarXiv:2012.09306(2020). [42] OpenZeppelin. Bypassingsmartcontracttimelocks.[Online].Availablefrom:https://blog.openzeppelin.com/bypassing-smart- contract-timelocks/.[Accessed23thDecember2022].2022. [43] EricPacuit.“Votingmethods”.In:StanfordEncyclopediaofPhilosophy(2019). [44] AndreaPeña-Calvinetal.“Acategorizationofdecentralizedautonomousorganizations:thecaseoftheAragonplatform”.In:IEEE TransactionsonComputationalSocialSystems(2023). [45] JoanaRPereiraandGiselleGarcia.“DAOs:Governanceintheblockchainera”.In:Blockchainapplications-transformingindustries, enhancingsecurity,andaddressingethicalconsiderations.IntechOpen,2023. [46] PawelPinio,RomanBatko,andDagmaraLewicka.“BetweenTheoryandValueTransactions:AMultifacetedExplorationofRelevance andResilienceofDecentralisedAutonomousOrganisations”.In:Proceedingsofthe20247thInternationalConferenceonSoftware EngineeringandInformationManagement.2024,pp.42–48. [47] OlivierRikken,MarijnJanssen,andZenlinKwee.“Governanceimpactsofblockchain-baseddecentralizedautonomousorganizations: anempiricalanalysis”.In:PolicyDesignandPractice6.4(2023),pp.465–487. [48] OlivierRikken,MarijnJanssen,andZenlinKwee.“TheInsandOutsofDecentralizedAutonomousOrganizations(Daos)”.In:Available atSSRN3989559(). [49] LucaSampetrean.“TheRuleofLawintheInternationalContext”.In:AvailableatSSRN3470459(2018). [50] CarlosSantanaandLauraAlbareda.“BlockchainandtheemergenceofDecentralizedAutonomousOrganizations(DAOs):An integrativemodelandresearchagenda”.In:TechnologicalForecastingandSocialChange182(2022),p.121806. [51] TanusreeSharmaetal.“Unpackinghowdecentralizedautonomousorganizations(daos)workinpractice”.In:2024IEEEInternational ConferenceonBlockchainandCryptocurrency(ICBC).IEEE.2024,pp.416–424. [52] ShuklaSidhartha. Defiprojectbeanstalkloses$182millioninflashloanattack.[Online].Availablefrom:https://www.bloomberg. com/news/articles/2022-04-18/defi-project-beanstalk-loses-182-million-in-flash-loan-attack.[Accessed23thDecember2022].2022. [53] SimranJagdev. Mmiandgnosissafe:Apartnershipmadeindaoheaven.[Online].Availablefrom:https://consensys.net/blog/ metamask/metamask-institutional/mmi-and-gnosis-safe-a-partnership-made-in-dao-heaven.[Accessed23thAugust2022].2022. [54] Snapshot. Snapshotdocs.[Online].Availablefrom:https://docs.snapshot.org.[Accessed23thAugust2022].2022. [55] AmeenSoleimanietal.TheMolochDAO.2019. [56] XiaotongSun,CharalamposStasinakis,andGeorigiosSermpinis.“DecentralizationillusioninDeFi:EvidencefromMakerDAO”.In: arXivpreprintarXiv:2203.16612(2022). [57] TomTaulli.“DecentralizedAutonomousOrganizations(DAOs)”.In:HowtoCreateaWeb3Startup.Springer,2022,pp.81–96. [58] TheAttacker. AnOpenLetterTotheDAOandtheEthereumcommunity.[Online].Availablefrom:https://steemit.com/ethereum/ @chris4210/an-open-letter-to-the-dao-and-the-ethereum-community.[Accessed20thDecember2022].2016. [59] TheGraph. Whatthegraphis.[Online].Availablefrom:https://thegraph.com/docs/en/about/.[Accessed23thAugust2022].2022. [60] TheLAO. Laomembershipverification.[Online].Availablefrom:https://www.thelao.io/verify/non-us.[Accessed20thAugust2022]. 2022. [61] JohnWTukeyetal.Exploratorydataanalysis.Vol.2.Reading,MA,1977. [62] Uniswap. Introducinguni.[Online].Availablefrom:https://uniswap.org/blog/uni.[Accessed23thAugust2022].2020. [63] Uniswap. UniswapProtocol.[Online].Availablefrom:https://uniswap.org/.[Accessed23thDecember2022].2022. [64] VitalikButerin. DAOs,DACs,DAsandMore:AnIncompleteTerminologyGuide.[Online].Availablefrom:https://blog.ethereum.org/ 2014/05/06/daos-dacs-das-and-more-an-incomplete-terminology-guide.[Accessed20thAugust2022].2014. [65] VitalikButerin. Quadraticpayments-aprimer.[Online].Availablefrom:https://vitalik.ca/general/2019/12/07/quadratic.html. [Accessed23thDecember2022].2022. Distrib.LedgerTechnol. 26 • Jungnickeletal. [66] DabaoWangetal.“Towardsafirststeptounderstandflashloananditsapplicationsindefiecosystem”.In:ProceedingsoftheNinth InternationalWorkshoponSecurityinBlockchainandCloudComputing.2021,pp.23–28. [67] QinWangetal.“Understandingdaos:Anempiricalstudyongovernancedynamics”.In:IEEETransactionsonComputationalSocial Systems(2025). [68] ShuaiWangetal.“Decentralizedautonomousorganizations:Concept,model,andapplications”.In:IEEETransactionsonComputational SocialSystems6.5(2019),pp.870–878. [69] MarkosZachariadis,GarrickHileman,andSusanVScott.“Governanceandcontrolindistributedledgers:Understandingthechallenges facingblockchaintechnologyinfinancialservices”.In:Informationandorganization29.2(2019),pp.105–117. Received26January2025;revised24September2025;accepted20October2025 Distrib.LedgerTechnol.