Control of Decentralized Autonomous * Organizations IAN APPEL JILLIAN GRENNAN University of Virginia Santa Clara University & UC-Berkeley January 11, 2023 ABSTRACT Blockchain technology has facilitated the emergence of a new type of organizational form, Decentralized Autonomous Organizations (DAOs). In contrast to corporations and other typesofbusinessorganizations, DAOsdonothavemanagersordirectorswithformaldecision rights. Instead, the members of DAOs collectively make decisions by voting on proposals. The governance process is encoded and executed via a smart contract, negating the need for centralized leadership. We examine the control of DAOs by analyzing 10,639 proposals across 151 DAOs. We find that, rather than democratizing decision-making, a small number, often three or fewer, of entities exert control over most decisions. JEL classification: O30, G34, M10, M50 Keywords: Blockchain, Decentralized Autonomous Organizations, Decentralized Finance, Stable- coins, Entrepreneurship, Organizational Economics, Corporate Governance, Political Economy * Appel: UniversityofVirgina,DardenSchoolofBusiness(email: AppelI@darden.virginia.edu)andGrennan: SantaClara University,LeaveySchoolofBusiness(email: jgrennan@scu.edu). WearegratefulforfinancialsupportfromRipple’sUniversity Blockchain Research Initiative (UBRI). We thank Nick Dahlborg, Allison Hart, George Iyalomhe, Rayan Malik, Dominic Parante,andRyanSchmitzerforhelpingtocollectDAOimprovementproposaldata. Electronic copy available at: https://ssrn.com/abstract=4322917 1. Introduction Recently, blockchain has been used to develop a new organizational form: Decentralized Au- tonomous Organizations (DAOs). DAOs are organizations that facilitate coordination between large numbers of individuals. The objectives of DAOs can be either for-profit or nonprofit, with most focusing on activities related to cryptocurrencies or Web3. In contrast to corporations and other types of business entities, DAOs typically do not have managers or directors with formal de- cision rights. Rather, decisions are subject to a vote by the DAO members. This community-based approach to decision-making is a critical aspect of a variety of DAO functions (e.g., product design, treasury management). As their name indicates, DAOs seek to avoid centralized decision-making. The extent to which this is achieved is unclear. If voting rights are concentrated, a DAO may be controlled by a small number of users. Indeed, an important regulatory concern in this space is how to police misleading characterizations, particularly regarding the extent of decentralization (Grennan, 2022). In this study, we examine 10,639 proposals across 151 DAOs to assess the degree to which decision-making in DAOs is decentralized. 2. Background Activities of a DAO are governed by a “smart contract,” a code programmed on a blockchain that automatically executes when pre-defined conditions are met. DAOs use a smart contract to define their governance structures. For example a smart contract can define, among other things, how many tokens must be held to create a proposal, when a quorum is reached for a vote, and the support required for a proposal to pass. Control of a DAO ultimately rests with the holders of its governance token. Tokens allow for members of a DAO to offer proposals and vote for their implementation. In some cases, token holders are entitled to fees/revenue generated by the DAO or to receive other perquisites (e.g., access to events). The initial allocation of tokens is typically split between the founders of a DAO, 1 Electronic copy available at: https://ssrn.com/abstract=4322917 large investors (sometimes VCs or other institutions), the public, and the DAO’s treasury. Tokens for many DAOs trade on secondary markets (e.g., Coinbase). The growth of DAOs as an organizational form has been rapid. The first entity organized as a DAO, an investor-directed venture capital fund (“The DAO”), was created in 2016 (Huberman et al., 2019). As of mid-2022, there are 4,000 active DAOs with aggregate treasury assets of $20 billion(Cointelegraph,2022). TheeconomicactivitiespursuedbyDAOsarevariedinnature. DAOs are most commonly used in decentralized finance (DeFi) (Harvey et al., 2021; Makarov and Schoar, 2022). In the following section, we provide examples of DAOs and the activities they undertake. 3. DAO Examples 3.1. Uniswap DAO The Uniswap DAO controls Uniswap, a decentralized exchange (DEX) based on self-executing (i.e., smart) contracts (Lehar and Parlour, 2022). Governance decisions (e.g., fees charged to users) are made by the DAO based on holdings of the UNI token. As of November 2022, the market capitalization of UNI tokens is approximately $5 billion, down from its all-time-high of over $20 billion the previous year. The largest holders of UNI (by voting power) include the venture capital firm Andreessen Horowitz (a16z). 3.2. BitDAO BitDAO is an investment fund that provides funding to crypto-related projects via grants and seed capital. BitDAO’s governance token, BIT, allows users to approve potential projects. Major investors in BitDAO include the cryptocurrency exchange Bybit as well as venture capitalists such as Founders Fund and Peter Thiel. As of November 2021, BitDAO’s treasury exceeds $2 billion. 2 Electronic copy available at: https://ssrn.com/abstract=4322917 3.3. LinksDAO LinksDAO was created to fund and purchase a country club. The DAO raised $10.4 in January 2022bysellingNFTsthatentileholderstogovernancerights, butnotmembershipinthepurchased course. The members of LinksDAO primarily consist of golf enthusiasts. 4. Characterizing DAO Control Several reasons are typically cited for why DAOs seek decentralization, including better de- cisions, censorship resistance, and fairness.1 While the structure of DAOs mitigates the need for managers and directors with decision-making authority, the extent to which they achieve decentral- ization also inherently depends on their ownership structure. That is, if token ownership is highly concentrated, a small number of members can control and influence decisions. It is helpful to consider the control of DAOs in the context of standard agency theory. A rich literature in financial economics studies the relationship between ownership structure and monitoring. If ownership is dispersed, individuals have weak incentives to monitor as they capture a small portion of the effort they expend. If ownership is concentrated, monitoring incentives are stronger (e.g., Shleifer and Vishny, 1986; Grossman and Hart, 1980). Concentrated ownership may also be costly given that there are incentives to “tunnel” assets to gainprivatebenefits(e.g.,Bertrandetal.,2002). Forexample,DAOsaresusceptibleto“governance attacks,” an extreme form of tunneling. In a governance attack, a bad actor legally acquires (by buying/borrowing tokens) voting power and uses this power to manipulate the DAO’s governance rules for a private benefit. For example, attackers could vote to pay themselves a large dividend from the DAO’s treasury. Given that such attacks stem from manipulation of encoded rules, they cannot be addressed through cryptography or cybersecurity. Appel and Grennan (2022) examine several additional trade-offs associated with decentralized governance. Aspectsofgovernancethatpromotebroadparticipationindecision-makingorenhance 1See Vitalik Buterin, “DAOs are not corporations: where decentralization in autonomous organizations matters.” Available at https://vitalik.ca/general/2022/09/20/daos.html 3 Electronic copy available at: https://ssrn.com/abstract=4322917 securityareassociatedwithpositiveoutcomes. Incontrast,barrierstotheadoptionofimprovement proposals are associated with negative outcomes. The findings are consistent with the notion that underanagencyframeworkconcentrateddecision-makingcanleadtotheexpropriationofminority token holders. In Table 1 we characterize the control of DAOs. The top of the table focuses on three broad categoriesofDAOs: DeFi,infrastructure,andWeb3. WemanuallyclassifyeachDAOinoursample into one of these broad categories. DeFi accounts for approximately 61% of DAOs and proposals in our sample. Infrastructure DAOs, which create/manage tools to help scale the crypto industry (e.g., Ethereum Name Service), account for 7% of DAOs and 3% of proposals in our sample. Web3 DAOs, which engage in a variety of activities related to new internet services (e.g., gaming, social clubs), account for 32% of DAOs and 36% of proposals in our sample. The number of unique blockchain addresses voting on proposals averages 767 for infrastructure DAOs, 148 for DeFi, and 110 for Web3. Our analysis indicates that decision-making in DAOs is highly centralized. The mean HHI across all three DAO categories is high, exceeding 0.29. The average voting power of the largest address is sizable, ranging from 39% (Infrastructure) to 45% (DeFi). For the top three voting addresses, power ranges from 67% to 71%, and the top three voters cast the majority of votes for more than 69% of proposals. The top three token holders also tend to vote in unison; between 61% and 79% of votes cast by top three token holders are unanimous. Our findings indicate that most DAO proposals are controlled by the top three token holders, highlighting the centralized nature of decision-making in these organizations. The remainder of Table 1 examines heterogeneity across specific functionalities of DAOs. In this case, we allow for multiple classifications of DAOs. We limit the sample to functionalities with at least one thousand proposals. We find evidence of significantly different levels of centralization across these types. For example, within Web3, social club DAOs tend to have relatively low levels ofcentralization(HHI=0.198),whilegamingDAOstendtohaverelativelyhighlevels(HHI=0.372). ForDeFi,measuresofcentralizationaresimilaracrossthedifferentfunctionalities(e.g.,stablecoins, borrowing/lending), with HHI ranging from 0.29 to 0.36, and the power of the top voter ranging 4 Electronic copy available at: https://ssrn.com/abstract=4322917 from 40% to 48%. In Appel and Grennan (2022), we highlight factors of the governance and voting process that potentially explain the differing levels of centralization across functionalities. Finally, we analyze the evolution of the control of DAOs over time. Figure 1 plots the con- centration of DAOs between Q3 of 2020 and Q3 of 2022. Over this period, HHI fell by nearly a third, from 0.45 to 0.30. This drop in concentration coincided with a large increase in the number of proposals. In early 2020, there were hardly any DAO proposals. By 2022, DAOs in our sam- ple voted on approximately two thousand proposals per quarter. This finding suggests that the concentration of DAOs may continue to decline as the organizational form matures. 5. Conclusion DAOs are a new type of organizational form that seeks to decentralize decision-making. Our findings indicate that DAO proposals are typically highly-centralized, but the degree of central- ization appears to be decreasing as the organizational form matures. Appel and Grennan (2022) considers the consequences of the unique aspects of DAO governance more broadly. For example, the centralized nature of DAOs may be a consequence of their ability to “fork,” which occurs when enough members vote to make a change to the basic set of rules that results in a DAO splitting and creating a second DAO. The relative ease of forking a DAO potentially limits negative effects of centralization, as dissident members can pursue their objectives independently. 5 Electronic copy available at: https://ssrn.com/abstract=4322917 References Appel, I., Grennan, J., 2022. Decentralized governance and digital asset prices. Working Paper . Appel, I., Grennan, J., 2023. Data for: Control of decentralized autonomous organizations. Amer- ican Economic Association [publisher], Inter-university Consortium for Political and Social Re- search [distributor] Project -183946. Bertrand, M., Mehta, P., Mullainathan, S., 2002. Ferreting out tunneling: An application to indian business groups. The Quarterly Journal of Economics 117, 121–148. Cointelegraph, 2022. Dao: The evolution of organization. White Paper . Grennan, J., 2022. Fintech regulation in the united states: Past, present, and future. Working Paper . Grossman, S. J., Hart, O. D., 1980. Takeover bids, the free-rider problem, and the theory of the corporation. The Bell Journal of Economics pp. 42–64. Harvey, C., Ramachandran, A., Santoro, J., 2021. DeFi and the Future of Finance. Wiley. Huberman, G., Leshno, J. D., Moallemi, C., 2019. An economist’s perspective on the bitcoin payment system. AEA Papers and Proceedings 109, 93–96. Lehar, A., Parlour, C. A., 2022. Decentralized exchanges. Working Paper . Makarov, I., Schoar, A., 2022. Cryptocurrencies and decentralized finance (defi). Brookings Papers on Economic Activity pp. 141–196. Shleifer, A., Vishny, R. W., 1986. Large shareholders and corporate control. Journal of Political Economy 94, 461–488. 6 Electronic copy available at: https://ssrn.com/abstract=4322917 slasoporP fo rebmuN 0002 0051 0001 005 0 54. 4. 53. 3. 52. IHH gnitoV 2020q3 2021q1 2021q3 2022q1 2022q3 Quarter Fig. 1. Trends in DAO Voting HHI and Number of Proposals. This figure plots the number of DAO proposals brought to a vote and the voting HHI associated with those proposals from 2020Q2 to 2022Q3. The bars represent the number of proposals with a starting vote date in the quarter. The dashed line represents the average voting HHI associated with those proposals. Source: Authors’ calculations using data collected from 151 DAOs linktrees and then matched to 10,639 proposals and voting records. See Appel and Grennan (2023) for data and code and Appel and Grennan (2022) for a more comprehensive discussion of DAOs, proposals, and voting. 7 Electronic copy available at: https://ssrn.com/abstract=4322917 Table 1: DAOs and Token Holder Control This table reports summary statistics for proposals brought to a vote for 151 DAOs between 2020Q2 and 2022Q3. The common functionalities of DAOs featured in the second half of the table are not mutually exclusive and recognize some DAOs serve multiple functions (e.g., tooling and liquidity purposes). Source: Authors’ calculations using data collected from 151 DAOs linktrees and then matched to 10,639 proposals and voting records. See Appel and Grennan (2023) for data and code and Appel and Grennan (2022) for a more comprehensive discussion of DAOs, proposals, and voting. Mean Mean Mean Pct. Pct. Pct. #of #of #of Mean Power Power Top1 Top3 Top3 DAOs Proposals Voters HHI Top1 Top3 Control Control Unanimous DeFi 93 6,518 148 0.334 45% 71% 35% 78% 79% Infrastructure 10 267 767 0.298 39% 68% 25% 69% 61% Web3 48 3,854 110 0.310 42% 67% 31% 72% 72% DeFi Borrowing/lending 27 1,189 320 0.337 45% 73% 36% 82% 81% DEXsandAMMs 44 3,797 148 0.338 47% 73% 37% 81% 73% Liquiditystaking 61 3,496 182 0.294 40% 68% 27% 72% 77% Stablecoin 20 1,603 215 0.361 48% 74% 40% 82% 81% Infrastructure Tooling 38 1,912 197 0.389 47% 73% 37% 79% 84% Web3 Assetmanagement 25 1,695 66 0.319 44% 69% 33% 78% 81% Gaming 13 1,902 137 0.372 49% 73% 45% 80% 69% Mediaandcuration 18 2,227 124 0.345 46% 72% 39% 79% 70% Publicgoods 28 1,084 85 0.297 41% 68% 29% 73% 76% Socialclubs 21 1,097 89 0.198 30% 57% 13% 60% 74% Talentandgigwork 29 1,087 91 0.318 41% 65% 29% 65% 81% Web3 55 4,067 115 0.304 41% 66% 31% 70% 73% 8 Electronic copy available at: https://ssrn.com/abstract=4322917