UBC Theses and Dissertations

UBC Theses Logo

UBC Theses and Dissertations

Essays on disclosure Yue, Yang 2020

Your browser doesn't seem to have a PDF viewer, please download the PDF to view this item.

Notice for Google Chrome users:
If you are having trouble viewing or searching the PDF with Google Chrome, please download it here instead.

Item Metadata

Download

Media
24-ubc_2020_november_yue_yang.pdf [ 1.55MB ]
Metadata
JSON: 24-1.0392426.json
JSON-LD: 24-1.0392426-ld.json
RDF/XML (Pretty): 24-1.0392426-rdf.xml
RDF/JSON: 24-1.0392426-rdf.json
Turtle: 24-1.0392426-turtle.txt
N-Triples: 24-1.0392426-rdf-ntriples.txt
Original Record: 24-1.0392426-source.json
Full Text
24-1.0392426-fulltext.txt
Citation
24-1.0392426.ris

Full Text

  ESSAYS ON DISCLOSURE by  Yang Yue  B.A., The Beijing Foreign Studies University, 2014 M.S., The University of Illinois at Urbana-Champaign, 2015  A THESIS SUBMITTED IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE DEGREE OF   DOCTOR OF PHILOSOPHY  in THE FACULTY OF GRADUATE AND POSTDOCTORAL STUDIES (Business Administration)  THE UNIVERSITY OF BRITISH COLUMBIA (Vancouver)   July 2020  © Yang Yue, 2020   ii  The following individuals certify that they have read, and recommend to the Faculty of Graduate and Postdoctoral Studies for acceptance, the dissertation entitled: Essays on Disclosure  submitted by Yang Yue in partial fulfillment of the requirements for the degree of Doctor of Philosophy in Business Administration  Examining Committee: Jenny Li Zhang, Associate Professor, Sauder School of Business, UBC Supervisor  Russell Lundholm, Professor, Sauder School of Business, UBC Supervisory Committee Member  Ralph Winter, Professor, Sauder School of Business, UBC Supervisory Committee Member Kai Li, Professor, Sauder School of Business, UBC University Examiner Vitor Farinha Luz, Assistant Professor, Vancouver School of Economics, UBC University Examiner    iii  Abstract This thesis explores the impacts of blockchain technology on accounting practice in two separate chapters. Blockchain is a system of distributed ledgers that can record information in a verifiable and permanent way. As the underlying technology of Bitcoin, Blockchain has received increased attention since 2008.  Chapter 2 takes an empirical approach to examine how startup firms use blockchain to finance their projects in the market for Initial Coin Offerings (ICOs). The blockchain technology allows entrepreneurs to commit to disclosing their transactions with investors before the transactions take place. Such decisions are coded into computer programs, known as ‘smart contracts,’ which become immutable once deployed on blockchains. I manually collected and analyzed the ‘smart contract’ code of 2085 ICO projects. I find that ICOs that make more disclosure commitments with blockchains are more likely to succeed, as measured by the likelihood of reaching fundraising goals and delivering preliminary products. I also find that transaction volumes disclosed on blockchains predict ICO outcomes and that investors punish ICOs with suspicious volumes, e.g., volumes that show signs of automated trading. These findings indicate that blockchains can function as a self-commitment device, and firms in the ICO market use blockchain to signal project quality. Chapter 3 takes an analytical approach to study how blockchain differs from traditional commitment mechanisms, e.g., regulations, and how firms can benefit from the additional features. When firms make commitments through disclosure regulations, they are choosing a  iv  regulation ‘combo,’ a set of predetermined disclosure requirements that apply to many firms. However, when firms make commitments on blockchains, they can customize a set of disclosure requirements that best suit them. I develop a model to study firms’ endogenous commitment decisions. A manager can commit to disclosing a value relevant signal before it is realized, or he can defer the disclosure decision until after he observes the signal. My analyses demonstrate that the commitment decision can credibly convey information that otherwise could not be disclosed, suggesting that blockchain enhances firms’ ability to communicate private information to the market.    v  Lay Summary This thesis explores the impacts of blockchain technology on accounting practice. It proposes blockchain as a commitment device that complements traditional commitment mechanisms. Specifically, firms can write computer programs that specify future activities, such as sending out a message at some future dates. After firms deploy the programs on blockchains, the programs become immutable, and firms can not go back on their ‘code.’ Compared to traditional commitment mechanisms, such as regulations, blockchain offers more flexibility. In capital markets, firms are subject to a regulation ‘combo,’ a set of predetermined disclosure requirements such as IFRS or US GAAP that apply to many firms. With blockchain technology, firms can customize a set of disclosure requirements that best suit them. I argue that blockchain enhances firms’ ability to communicate information to the market and provide supporting evidence from the market for Initial Coin Offerings (ICOs).  vi  Preface My committee members provided extensive support and guidance in various research fields. Both Chapter 2 and Chapter 3 are original and independent work by the author, Yang Yue.  vii  Table of Contents  Abstract ......................................................................................................................................... iii Lay Summary .................................................................................................................................v Preface ........................................................................................................................................... vi Table of Contents ........................................................................................................................ vii List of Tables ..................................................................................................................................x List of Figures ............................................................................................................................... xi Acknowledgments ....................................................................................................................... xii Dedication ................................................................................................................................... xiii Chapter 1: Introduction ................................................................................................................1 Chapter 2: Building Trust from Code: Disclosure Commitments on Blockchains .................3 2.1 Introduction ..................................................................................................................... 3 2.2 Institutional Background about ICOs and Smart Contracts .......................................... 10 2.2.1 Examples of Successful and Failed ICOs ................................................................. 11 2.2.2 Typical ICO Processes and Terminology ................................................................. 12 2.2.3 Smart Contract and Ethereum Platform .................................................................... 13 2.3 Literature Review and Hypotheses Development ......................................................... 15 2.3.1 The Empirical Literature on ICOs ............................................................................ 15 2.3.2 The Theoretical Literature on ICOs .......................................................................... 17 2.3.3 ICO vs. IPO ............................................................................................................... 19 2.3.4 Hypotheses ................................................................................................................ 20  viii  2.4 Research Design............................................................................................................ 23 2.4.1 Data and Sample ....................................................................................................... 23 2.4.2 Measures of Blockchain Disclosures ........................................................................ 24 2.4.3 Measures of ICO Outcomes ...................................................................................... 27 2.4.4 Control Variables ...................................................................................................... 28 2.4.5 Empirical Models ...................................................................................................... 29 2.5 Empirical Results .......................................................................................................... 31 2.5.1 Summary Statistics.................................................................................................... 31 2.5.2 Determinants of Blockchain Disclosures .................................................................. 33 2.5.3 The Test for Hypothesis One .................................................................................... 34 2.5.4 The Test for Hypothesis Two ................................................................................... 34 2.5.5 The Test for Hypothesis Three ................................................................................. 35 2.5.6 Robustness Tests ....................................................................................................... 36 2.6 Conclusions ................................................................................................................... 37 Chapter 3: A Model for Disclosure Commitments ...................................................................57 3.1 Introduction ................................................................................................................... 57 3.2 Literature Review.......................................................................................................... 62 3.3 The Model ..................................................................................................................... 65 3.3.1 Players and Information ............................................................................................ 65 3.3.2 Timeline .................................................................................................................... 65 3.3.3 Objectives and Payoffs ............................................................................................. 66 3.3.4 Some Discussions about the Model .......................................................................... 67  ix  3.3.5 A Simplified Case: Commitments are Impossible .................................................... 69 3.4 Equilibrium Definition .................................................................................................. 70 3.5 Equilibrium Analysis .................................................................................................... 72 3.6 Simulation Results ........................................................................................................ 73 3.7 Conclusions ................................................................................................................... 76 Chapter 4: Conclusion .................................................................................................................79 Bibliography .................................................................................................................................81 Appendices ....................................................................................................................................86 Appendix A Appendix for Chapter 2 ........................................................................................ 86 A.1 Variable Definitions .................................................................................................. 86 A.2 An Example of Ethereum Address ........................................................................... 88 A.3 Examples of Blockchain Information ....................................................................... 89 A.4 Examples of Smart Contracts Code .......................................................................... 90 A.5 Examples of ICOs That Make Partial or Full Disclosure Commitments.................. 91 A.6 Sample Code of Smart Contract ............................................................................... 92 A.7 Blockchain Information Under Each Commitment Decision ................................... 93 Appendix B Appendix for Chapter 3 ........................................................................................ 94 B.1 Proof of Theorem 1 ................................................................................................... 94   x  List of Tables Table 2-1: Sample Compositions .................................................................................................. 40 Table 2-2: Summary Statistics ...................................................................................................... 41 Table 2-3: Correlation Matrix ....................................................................................................... 44 Table 2-4: Determinants of Blockchain Disclosures .................................................................... 45 Table 2-5: Blockchain Disclosures and ICO Outcomes (H1) ....................................................... 46 Table 2-6: Predicting ICO Outcomes with Blockchain Information (H2) ................................... 48 Table 2-7: Suspicious Volumes and ICO Outcomes (H3) ............................................................ 51 Table 2-8: Robustness Tests ......................................................................................................... 52   xi  List of Figures Figure 2-1: ICO Statistics in 2018 ................................................................................................ 38 Figure 2-2: Sample Decomposition .............................................................................................. 38 Figure 2-3: Transitions from Public ICOs to Private ICOs .......................................................... 39 Figure 3-1: Illustrations of Equilibrium ........................................................................................ 77 Figure 3-2: Illustrations of Equilibrium with Different Disclosure Costs .................................... 78    xii  Acknowledgments My survival in the Ph.D. program would not have been possible without the support of many individuals. In particular, I would like to express my sincere gratitude to my supervisor Jenny Li Zhang for her invaluable guidance and continuous support. She is very open-minded and encourages me to explore the blockchain technology, an unconventional topic in accounting research. I am also grateful for Russell Lundholm for his advice and encouragement. His great personality and life attitude are always inspiring. I would also like to thank Ralph Winter for joining my committee and providing economic insights. I thank other accounting faculty members for their helpful feedbacks throughout my Ph.D. study, particularly Sandra Chamberlain, Dan Simunic, Alexander Bleck, and Ira Yeung. I also thank Jenna D’Adduzio, Rajesh Vijayaraghavan, and Xin Zheng for their guidance on the job market. My Ph.D. life would not have been as smooth without the help from our divisional assistant Debra Harris and Ph.D. administrator Elaine Cho. I am also grateful for the friendship and support from previous and current Sauder Ph.D. students. I particularly enjoy our weekend food trips to Richmond.  I own special thanks to my parents, who have been supportive all the time, although I can never thank them enough.  xiii  Dedication  To my parents.   1  Chapter 1: Introduction This thesis explores the impacts of blockchain technology on accounting practice. Blockchain can be broadly defined as a system of distributed ledgers that can record information in a verifiable and permanent way. As the underlying technology of Bitcoin, blockchain has received increased attention and mixed reviews since 2008. On the one hand, many companies (JP Morgan and Facebook) and governments (the Chinese and the Singaporean governments) are rapidly developing this new technology. On the other hand, it has been criticized as overhyped, and many people question the actual benefits of blockchain. As an information system or a database, blockchain is inherently linked to accounting. This thesis sheds light on what firms can achieve with blockchain and how firms can benefit from it in two separate chapters. Chapter Two takes an empirical approach to examine how startup firms use blockchain to finance their projects in the market for Initial Coin Offerings (ICOs). The blockchain technology allows entrepreneurs to commit to disclosing their transactions with investors before the transactions take place. Such decisions are coded into computer programs, known as ‘smart contracts,’ which become immutable once deployed on blockchains. I manually collected and analyzed the ‘smart contract’ code of 2085 ICO projects. I find that ICOs that make more disclosure commitments with blockchains are more likely to succeed, as measured by the likelihood of reaching fundraising goals and delivering preliminary products. I also find that transaction volumes disclosed on blockchains predict ICO outcomes and that investors punish ICOs with suspicious volumes, e.g., volumes that show signs of automated trading. These  2  findings indicate that blockchains can function as a self-commitment device by ICO firms to signal project quality. Chapter Three takes an analytical approach to study how blockchain differs from traditional commitment mechanisms, e.g., regulations, and how firms can benefit from the additional features. When firms make commitments through disclosure regulations, they are choosing a regulation ‘combo,’ a set of predetermined disclosure requirements that apply to many firms. However, when firms make commitments on blockchains, they can customize a set of disclosure requirements that best suit them. I develop a model to study firms’ endogenous commitment decisions. A manager can commit to disclosing a value relevant signal before it is realized, or he can defer the disclosure decision until after he observes the signal. My analyses demonstrate that the commitment decision can credibly convey information that otherwise could not be disclosed, suggesting that blockchain enhances firms’ ability to communicate private information to the market. Overall, this thesis proposes blockchain as a commitment device that complements traditional commitment mechanisms. From an accounting perspective, it represents a new channel through which firms can communicate with investors. Each chapter is designed to be self-contained, and the introduction section of each chapter includes more detailed discussion of the research question and contribution.  3  Chapter 2: Building Trust from Code: Disclosure Commitments on Blockchains 2.1 Introduction In recent years, several cryptocurrencies, such as Bitcoin and Ether, have become very popular. These cryptocurrencies, also known as ‘coins,’ are powered by blockchain technology, which is a system of open and distributed ledgers that can record information in a verifiable and permanent way. People have made fortunes from these cryptocurrencies. On May 22, 2010, now known as Bitcoin Pizza Day, an early adopter paid 10,000 Bitcoins for two Papa John’s pizzas. In December 2018, Bitcoin reached an all-time high of $19,891 per coin, and it was still being traded at above $8,000 in May 2020. This success has sparked much excitement. People have begun creating new coins or tokens with additional features, hoping that their coins will become the next Bitcoin.1 Such coins represent access to future products or services or serve as a medium of exchange on future platforms. For example, the Filecoin ICO planned to develop a peer-to-peer file storage platform. Individuals earn Filecoin tokens, known as  FIL, by providing unused storage on their devices to clients, who   1 There are no formal definitions of the terms ‘coin’ and ‘token’. Generally, coins are for general purposes and are based on their own blockchain, whereas tokens are for a specific use, e.g. to access a specific service, and are built on existing blockchains. In this study,the terms are used interchangeably.  4  must pay for storage in FIL. The market for initial coin offerings (ICOs), where entrepreneurs issue coins to raise funds from the crowd, has emerged and expanded rapidly. According to Cointelegraph.com, ICOs raised about $10 billion in 2017 and about $11.3 billion in 2018.  In addition to the inherent risks of a startup market, the ICO market faces regulatory uncertainty and offers little investor protection. Since coins represent access to products or services rather than profits, it is not clear whether they are securities and subject to security laws. In November 2018, Judge Gonzalo P. Curiel ruled against the Securities and Exchange Commission (SEC) for its failure to demonstrate that a coin issued by the Blockvest ICO is a security. In a statement on cryptocurrencies and ICOs, SEC chairman Jay Clayton highlighted the risks of fraud and manipulation in the ICO market and stated that the SEC might not be able to pursue bad actors or recover funds effectively. In a market with regulatory uncertainty, entrepreneurs lack mechanisms to communicate with investors credibly. The information asymmetry is severe, and adverse selection leads to market breakdowns. In the ICO market, how can investors assess the legitimacy and popularity of ICO projects? Promoted as ‘trustless, decentralized and transparent,’ blockchains are the solution proposed by crypto communities. A typical ICO project creates tokens on blockchains, conducts token sales on blockchains, and develops blockchain-related products. A fundamental question behind the ICO phenomenon is whether blockchain technology can mitigate the adverse selection problem in the ICO market.  To answer this question, I study the role of smart contracts in ICO processes. Smart contracts are computer programs that run on blockchains and are standard tools to conduct token sales. Such  5  programs contain a set of ‘if-then’ conditions and execute automatically when the conditions are met. Specifically, developers can preset a list of ICO terms, such as start date, end date, token supply, token price, etc. Once a smart contract is deployed on blockchains, it becomes immutable and self-enforcing. A well-designed smart contract receives investments after the preset start date, issues tokens based on the preset token price, and finishes the ICO process if the number of tokens issued reaches the preset token supply or after the preset end date.  While performing those transactions, smart contracts can record them on blockchains and disclose them to the public in real time. Smart contracts can allow investors to access two types of information: (1) token distribution histories, which are comparable to inventory ledgers if tokens are considered inventories of ICO projects,2 and (2) cryptocurrency investments from investors, which are comparable to cash flow statements. Throughout this paper, I refer to transaction information on blockchain as inventory information, cash flow information, or blockchain information in general. Blockchain information allows investors to monitor fundraising processes and to assess aggregate demands for ICO projects.3    2 As of July 2019, there were no accounting standards for cryptographic assets. In practice, crypto assets are often recorded as either inventories or intangible assets. 3 It is worth noting the distinction between inventory information and cash flow information, as a reader might expect that the amount of cryptocurrency raised equals token prices multiplied by the quantities of token distributed. Generally, cash flow information is a more credible and complete measure of funding status than inventory information, because it is possible for smart contracts to distribute tokens without receiving cryptocurrencies, and it is also possible for a smart contract to have a dynamic pricing feature.  6  Interestingly, using smart contracts to disclose blockchain information is a disclosure commitment rather than a disclosure. The distinction between a commitment and a disclosure is that the former is a decision a firm makes about what it will disclose before it knows the content of the information (i.e., ex-ante), whereas the latter is a decision a firm makes after it observes the content (i.e., ex-post). The decision to disclose blockchain information is made at the time of writing smart contracts, before actual ICO processes. At that time, entrepreneurs do not know with certainty how many people will invest in their ICO projects. In other words, although entrepreneurs may have private information about their projects, they face demand uncertainty from product markets, and they can decide in advance whether to disclose the information about demand. By virtue of the blockchain technology, the decision is coded into smart contracts and is irreversible. I argue that disclosure commitments made by smart contracts serve as costly signals of project quality. If startups face demand uncertainty and there are frictions for nondisclosure,4 the high-type firms will find it less costly to make disclosure commitments, i.e., to forgo the option of nondisclosure. Specifically, entrepreneurs with high-type projects are more confident that they will receive favorable market reactions and have stronger incentives to disclose blockchain information. However, entrepreneurs with low-type projects value the option of nondisclosure   4 The disclosure frictions in this setting mainly relate to some undesirable features of smart contracts, such as potential security risks. A detailed discussion of this issue appears in the hypothesis development section.  7  more, because they want to disclose blockchain information only when market reactions turn out to be good, which, based on their private information, is less likely. Essentially, the option value of nondisclosure is higher for low type startups, and the decisions of whether to use smart contracts reveal entrepreneurs’ private information about their types. This argument is consistent with theoretical predictions in Titman and Trueman (1986). In practice, many ICOs do not allow investors to access blockchain information. In this study, I collect a sample of ICOs in 2018 on Ethereum, the largest blockchain platform for ICOs. I search for the smart contract of each ICO on Etherscan, the most popular blockchain explorer for Ethereum, and analyze its computer codes if they are available. Approximately 40 percent of ICOs did not use smart contracts, i.e., they did not make any disclosure commitments. Another 40 percent of ICOs only disclose inventory information, i.e., they made partial disclosures commitments. Only about 20 percent of ICOs disclose both inventory information and cash flow information, i.e., they made full disclosure commitments. I first test for the signaling role of smart contracts. I find that ICOs that commit to more blockchain disclosures through smart contracts are associated with a higher probability of success. Specifically, compared to ICOs without any disclosure commitments, ICOs that make partial disclosure commitments are more likely to succeed. Compared to ICOs that make partial disclosure commitments, ICOs that make full disclosure commitments are more likely to succeed. I measure the success of an ICO project with a variety of ICO outcomes, including whether an ICO reaches its fundraising goals, lists its tokens on crypto exchanges, keeps updating its websites, and delivers preliminary products or applications. The results hold after  8  controlling for entrepreneurs’ reputation, expertise, and various ICO characteristics, and they are consistent with the signaling role of disclosure commitment. I then provide evidence that blockchain information is relevant for decision making; e.g., it allows investors to assess the aggregate market demands or the potential of an ICO project. This is a necessary condition for the signaling argument. For the subsample of ICOs that disclose blockchain information, the number of transactions on blockchains in early ICO periods is indicative of both short-term fundraising outcomes and long-term performances. These findings support ‘the wisdom of the crowd’ hypothesis in the crowdfunding market. Lastly, I show that blockchains can offer some governance features in the ICO market. Although information on blockchains becomes immutable once it is recorded, an ICO team may manipulate transaction records in real time, pretending to be a hot project when there are in fact no economic transactions. For example, an ICO can distribute a large number of tokens to random accounts, or it can recycle the raised cryptocurrencies to create constant investment inflows.5 I used a simple measure to capture suspicious transaction volumes: an ICO is suspicious if it transfers the same number of tokens in many transactions. Since people naturally   5 The EOS ICO was accused by the crypto community of re-investing raised funds to its Crowdsale smart contract.  9  invest different amounts, I consider the above patterns as signs of automated trading, and I show that investors punish ICOs with this pattern of transactions on blockchains. This study makes several contributions to the literature. First, it adds to the emerging literature on the ICO market. It provides the first set of evidence that blockchain disclosures are associated with favorable market reactions. The evidence complements a series of papers that focus on traditional disclosure channels in the ICO market, such as social media and ‘whitepapers,’ unregulated documents in which ICO teams describe their business. (Florysiak and Schandlbauer 2018; Lee, Li, and Shin 2018; Fisch 2018; Bourveau et al. 2018). While documenting the benefits of blockchain technology, this paper also highlights that only a small and declining portion of ICOs are making the most of this technology, which could explain the presence of scams and justify pessimistic views about ICO market. These findings can help regulators and the crypto community develop future industry standards for the market and the technology. Second, the paper contributes to our understandings of the benefits of early feedbacks from investors in crowdfunding. Recent evidence from the crowdfunding literature suggests that entrepreneurs face demand uncertainty and learn from early investments about their projects’ potential. When such information is present, investments become more efficient (Chemla and Tinn 2019; Strausz 2017). This paper provides evidence that investors also value this information and favor startups that provide it voluntarily. Indeed, early investments predict ICO outcomes, and the prediction power is higher when more information is disclosed. The prediction power supports the ‘wisdom of the crowd’ hypothesis, suggesting that the crowd collectively has information about a project’s potential that is unknown to the entrepreneurs.   10  Lastly, this paper introduces smart contracts as a new self-commitment device, while previous studies mainly focus on disclosure regulations and auditors as credible commitment mechanisms (Leuz and Verrecchia 2000; Cheng, Liao, and Zhang 2013; Bailey, Andrew Karolyi, and Salva 2006; Titman and Trueman 1986). With blockchains, a firm can make credible commitments at low costs by writing a computer program that they can not modify in the future. The paper suggests that the commitment function of blockchains is effective in mitigating adverse selections in the ICO market and that computer code contains information relevant for investment decisions. This chapter proceeds as follows. Section 2.2 provides institutional background on ICOs and smart contracts. Section 2.3 summarizes the related literature and develops hypotheses. Section 2.4 describes the research design. Section 2.5 presents empirical results, and Section 2.6 concludes. 2.2 Institutional Background about ICOs and Smart Contracts ICOs, or the initial coin offerings, is a fundraising process where startups raise capital by issuing digital assets. The digital assets are cryptographically protected digital records implemented on a blockchain and are referred to as ‘coins’ or ‘tokens.’ The tokens usually represent access to future products or services that startups plan to deliver, or they serve as a medium of exchange on future platforms that startups intend to build. The tokens are thus considered as ‘utility’ tokens, as they do not represent cash flow right to startups’ future profits. However, most ICOs sell their tokens at substantial discounts and plan to list their tokens on crypto exchanges where token prices are subject to supplies and demands. ICO investors can profit from appreciations in  11  token value. By using blockchain technology, ICOs embrace a broad range of investors, low transaction costs, and do not require financial intermediaries.  2.2.1 Examples of Successful and Failed ICOs The FileCoin is a widely cited ICO that is successful in raising over 200 million dollars. The project attempts to utilize unused storage in data centers and hard drives by developing a decentralized peer-to-peer file storage platform. Storage providers will earn Filecoin’s tokens, the FIL, by storing digital files for clients, who must use FIL to pay for storage. In contrast to major cloud storage providers such as Dropbox, Filecoin distributes and stores clients’ files across different storage providers. Filecoin argues that decentralized platforms are robust to cyber-attacks and can avoid centralized control of user data.  The Filecoin ICO sells 200 million FIL tokens, consisting of 10 percent of total token supplies. In the pre-ICO-sale in August 2017, Filecoin raised approximately $52 million from 150 selected investors at a discounted price of $0.75 per token. The main public sale raised $153.8 million from more than 2,100 investors at a weighted-average price of $4.61 per token. After the ICO process, the Filecoin team followed its roadmap in developing products. As of July 2019, they have not launched their first public TestNet yet. The Opacity project offers a similar storage solution with censorship-resistant protocols. However, the project is previously known as the Oyster Protocol, a notorious ICO exit scam. After completing the ICO process, Bruno Block (a fake name), the founder and former CEO of the Oyster Protocol, was alleged to have created 3 million additional PRL tokens, which he  12  immediately sold for about $300,000, roughly 5 percent of project’s market value. Block achieved this through a function in the Oyster smart contract, a function which he insisted must remain in the code, which gives himself administrative privilege to create new tokens. None of the team members know the real identity of Bruno Block, and the value of existing tokens is diluted by new tokens. The team move forward and transform the Oyster Protocol to the Opacity project. 2.2.2 Typical ICO Processes and Terminology A typical ICO project starts with a whitepaper in which entrepreneurs explain their business. The whitepaper is arguably the most important information source in ICOs. The contents of whitepapers vary, but most include a description of the current market or the problem they try to solve, business vision, technological implementation, team members, and token sale information. ICOs rely heavily on social media. In addition to traditional social media, such as Facebook, Twitter, and Reddit, entrepreneurs post their ICO information on specialized ICO data providers, such as ICObench and trackICO. ICO founders often hire employees to interact with the public on crypto-related platforms such as BitcoinTalk, Medium, and Telegram. Many ICOs provide LinkedIn profiles for their team members. Almost all ICOs build their website to conduct token sales. The actual token sale process requires several decisions on ICO designs. An ICO can have many stages of token sales; each has a different price and targets different investors. ICO founders need to decide on the number of tokens to be sold and to be vested for stakeholders. They also  13  need to determine the length of token sales, the hard cap (the maximum amount of funds to be raised), and the soft cap (the minimum amount of funds to be raised to continue the project). Tokens issued during ICOs are stored in crypto-accounts, or wallets, of investors, and are transferable. Successful projects can list their token on one or more cryptocurrency exchanges, including centralized exchanges, such as Binance and Coinbase, and decentralized exchanges, such as Etherdelta and Bancor. Exchanges usually charge a decent amount of listing fees. Once listed, token prices are subject to market demands and supplies, and investors may realize their gains and losses on their investments. The secondary market for tokens can be very liquid. According to Howell, Niessner, and Yermack (2018), average daily trading volumes are 12.3 million US dollars. 2.2.3 Smart Contract and Ethereum Platform The unique feature of ICO is that the procedures related to token sales can be hardcoded in smart contracts. Smart contracts are computer programs that run on blockchains. They usually contain a set of ‘if-then’ conditions and execute automatically when the conditions are met. If a smart contract is deployed on blockchains, it becomes immutable and can be verified by all parties. It is also enforced by computer protocols, thus removing the needs of third parties.  To conduct a token sale, ICO teams can preset a list of ICO terms, such as start date, end date, token supply, and token price. Accordingly, the smart contract will receive investments after the preset start date, issue tokens based on the preset token price, and finish the ICO process if the number of tokens issued reaches the preset token supply or after the preset end date.   14  After receiving investments, the ICO team can store raised funds in ‘multi-signature wallets.’ A multi-signature wallet is a smart contract implemented as a wallet to store cryptocurrencies that belong to multiple owners. Key members of ICO teams can preset a daily limit to withdraw raised funds. A withdrawal must be approved by multiple owners before it can be executed. More complex smart contracts can facilitate Decentralized Autonomous Organizations (DAOs). Members of such organizations make proposals and vote on important management decisions. For example, the MakerDao issues a ‘stable’ coin, whose value is tied to US dollars and backed by collateralized cryptocurrency loans. Smart contracts became popular following the launch of the Ethereum platform in 2015. Ethereum is the world’s leading programmable blockchain that features a Turing-complete programing language, the Solidity, and a native cryptocurrency, the Ether (ETH). Ether can be transferred between accounts and be used to compensate mining nodes for computations performed. The Ethereum Virtual Machine (EVM) can execute scripts written in Solidity, a programing language designed for the Ethereum platform, using an international network of public nodes. Ethereum is the largest platform for ICO, accounting for more than 80 percent of market shares. There are also over 2000 decentralized applications powered by smart contracts on EVM.   15  2.3 Literature Review and Hypotheses Development 2.3.1 The Empirical Literature on ICOs In recent years, many attempts have been made, both empirically and theoretically, to understand the ICO phenomenon. Most empirical studies try to identify factors associated with ICO success, and they focus more on traditional disclosure channels, such as whitepapers and social media, but less on the use of blockchain technology. Whitepapers are unregulated documents in which ICO teams describe their businesses, plans, team members, and so on, and they are arguably the most important information source in the ICO market. Several studies argue that whitepapers are effective in reducing information asymmetry between firms and investors. Feng et al. (2019) analyze whether services or products of an ICO are directly tied to its blockchain platform and assess the technicality of whitepapers. They find that providing technical details can be an effective way for high-quality ICO projects to signal their quality credibly. Similarly, Fisch (2018) finds that ICOs with long and technical whitepapers generally raise more money, and the author and two experienced experts assess the technicality. Bourveau et al. (2018) measure ‘whitepaper opacity’ using the Fog index and find that opacity is negatively related to the amount of raised funds.  In contrast, some studies question the credibility of whitepapers. Momtaz (2019) argue that ICO founders systematically exaggerate information in whitepapers, e.g., use more amplification words. Exaggerating entrepreneurs raise more funds, but the crowd eventually learns about the exaggeration through trading with other investors. Cohsey et al. (2018) study whether  16  whitepapers and smart contracts are consistent in terms of token supply, insider vesting, and contract upgradability, and conclude that smart contracts generally do not implement promises made in whitepapers. It is worth noting that the focus of this paper is different from that of Cohsey et al. (2018). While they examine the operating function of smart contracts, such as token supply and vesting, I focus on the information function in this paper, e.g., whether transaction information is publicly observable. Several papers study factors other than whitepapers. Lee, Li, and Shin (2018) provide evidence that analysts’ ratings on various ICO websites predict ICO outcomes and argue that the wisdom of crowds could substitute underwriters in the ICO market. Howell, Niessner, and Yermack (2018) find that VC backgrounds, a proxy for credibility, and Twitter followers are associated with higher post-listing liquidity, measured in Amihud price impact, dollar volumes, and turnovers. Benedetti and Kostovetsky (2018) show significant returns of 179% from the ICO price to the first day’s opening market price for a sample of ICOs that are eventually listed, consistent with high compensation for high risks. They also identify twitter activities as significant determinants for returns. Perhaps most relatedly, Adhami, Giudici, and Martinazzi (2018) study an early sample of 253 ICOs and conclude that project source code on GitHub, a website where developers share their codes, are significant predictors of ICO success. However, the code in this study is very different from theirs. Source code that firms disclose on GitHub is not regulated, not enforced by computers, and often updated. GitHub is designed to facilitate cooperation among programmers to develop and test computer code. In contrast, the smart contracts in this study are deployed on  17  the Ethereum blockchains. They are immutable and executed by computer protocols. The results in this study are robust to the inclusion of GitHub disclosures. While most of the papers study the information contents from various sources, we know less about the credibility of such information, and the credibility is important in the ICO market, which features information asymmetry and regulatory uncertainty. In this study, I focus on the use of blockchain in the ICO process and assess the effectiveness of smart contracts as ways to build trust. The paper thus complements previous research and deepens our understanding of this emerging market.  2.3.2 The Theoretical Literature on ICOs The theoretical literature is still in debate about the comparative advantages of ICOs over traditional equity financing. Li and Mann (2018) model tokens as the medium of transactions on a platform. They argue that ICOs can solve coordination failure during platform operation. Coordination failure refers to the no-trade equilibrium where each potential participant believes that others will not participate, and it is rational for him/her not to participate, even though participation may be welfare improving. Purchasing platform-specific tokens at a cost is a credible commitment to using the platform in the future and thus solves the coordination failure. Similarly, Bakos and Halaburda (2018) argue that issuing token represents entrepreneurs’ trade-off between current and future revenues. Potential appreciations in token value attract early adopters and help solve the coordination failure. However, they argue that when a platform is not facing capital constraints, the traditional strategy to subsidize adoption is more profitable. In contrast, Michael and Xiong (2018) model tokens as both membership to platforms and as fees to  18  ‘miners,’ the computing power providers of a blockchain. They argue that, in the presence of network effects, clearing two markets with one token price leads to either no equilibrium or two equilibriums. Moreover, the possibility of no equilibrium discourages full disclosure about platforms.  Another potential benefit of ICOs stems from the ability to harness ‘wisdom of the crowd.’ The argument follows the crowdfunding literature and states that early adopters provide information about the potential of a project. Such information helps entrepreneurs to decide whether to carry on their projects and helps investors to make investment decisions. Strausz (2017) model investors as non-strategic, e.g., they either like or dislike a platform regardless of other investors. The author argues that, in the presence of demand uncertainty and the absence of moral hazard, an all-or-nothing crowdfunding scheme reveals aggregate demand uncertainty and allows entrepreneurs to achieve first-best profits. Cong and Xiao (2019) study the effect of an all-or-nothing fundraising scheme on information cascade, a situation where people make decisions sequentially based on previous people’s actions while ignoring their private information. They argue that agents may rationally ignore negative signals, but never ignore positive signals, and effectively delegate their decisions to later ‘gate-keepers,’ leading to uni-directional cascades. Information aggregation approaches full efficiency as crowd size grows.  Michael and Xiong (2018) model that investors, or ‘household’ in their terms, use their endowments and cryptocurrency prices and volumes observed on blockchain to infer aggregate demand of a platform. They show that investors’ decisions to join a platform depend on their shared belief based on public signals of households’ aggregate endowments.    19  Theoretical literature usually abstracts away the classic information asymmetry problem and the associated adverse selection so that they can focus on the role of tokens as a financing device. In many cases, demands for platforms are assumed to be exogenous and built into agents’ utility functions. These assumptions are in sharp contrast to the unregulated nature of the ICO market, where information asymmetry is severe. Relatedly, many papers ignore the endogeneity of the decision to use blockchain. I study the role blockchain as a disclosure commitment device, focus on how it overcomes adverse selection, and highlight the endogenous nature of disclosure commitments.   2.3.3 ICO vs. IPO There is abundant research on the signaling in the IPO market that is relevant to the ICO market. Carpenter and Strawser (1971) document that IPO firms often switch from a local auditor to a national one known for higher quality and receive higher share prices. Based on this observation, Titman and Trueman (1986) argue that an entrepreneur with more favorable private information about his firm’s value will commit to higher levels of disclosure by choosing a higher-quality auditor or underwriter than will an entrepreneur with less favorable private information. Subsequent papers provide consistent evidence. For example, Megginson and Weiss (1991) and Barry et al. (1990) both document the role of venture capital in signaling firms’ qualities. Pittman and Fortin (2004) provide evidence that new public firms with ‘Big Six’ auditors have lower borrowing costs. ICOs differ from IPOs in the following aspects. First, the ICO market is unregulated, while the IPO market is heavily regulated. Without clear regulations and legal consequences, the  20  traditional signaling methods should be less effective, making ICO the ideal setting to study the blockchain as a new commitment tool. Second, due to the lack of regulations, the ICO market embraces earlier-stage startups, while most IPO firms are already well established. The information asymmetry problems are particularly severe for ICO firms, so the marginal benefits of disclosure commitments are larger and easier to capture in the ICO market. Lastly, levels of disclosure commitments can be measured accurately and objectively based on code analyses of smart contracts, which reduces measurement errors and makes it easier to interpret empirical results. 2.3.4 Hypotheses Building on the previous literature, I argue that smart contracts provide investors decision-relevant information by allowing them to monitor the fundraising process. Some smart contracts allow investors to access inventory information, while others allow investors to monitor cash flows. Blockchain information is useful for investors to assess project popularity and to overcome coordination failures. The disclosure of blockchain information is costly because the blockchain technology has certain undesirable features. On the one hand, using smart contracts introduces security risks. Like any other computer program, a smart contract may contain bugs and can be hacked. In the high-profiled DAO attack, 150 million dollars’ worth of cryptocurrencies raised during the ICO processes were stolen by hackers. Relatedly, the founder of the Arcblock ICO made a YouTube video explaining that they would not use a public smart contract in the ICO because of security concerns. On the other hand, because blockchains eliminate the role of third parties, a smart  21  contract can not handle the cases where investors send their money to wrong addresses, or they forget passwords of their accounts. These are common mistakes people make in the emerging industry. These undesirable features are the often-cited reasons why many ICO team do not use smart contracts in their fundraising processes. I argue that entrepreneurs in the ICO market have private but imperfect information about project quality, as they face demand uncertainty about their products or services. Lacking credible ways to communicate with investors, entrepreneurs signal their project quality by committing to disclosing transaction information through smart contracts. Entrepreneurs with high type projects are more confident in favorable market reactions and have stronger incentives to disclose blockchain information. In contrast, entrepreneurs with low type projects expect that favorable market reactions are less likely and thus have incentives to hide blockchain information. The costs of forgoing the option to withhold information are greater for projects with lower quality. Therefore, commitments to disclosure through smart contracts are costly signals of project quality. This argument is consistent with theoretical predictions in Titman and Trueman (1986), and I state my first hypothesis as follows: H1: ICOs that commit to more blockchains disclosures are more likely to succeed. A necessary condition for hypothesis one to hold is that blockchain information is value relevant. If participants in the ICO market are entirely irrational and make investments randomly, then disclosures of transaction information should not make any difference. As contraposition, if disclosure commitments matter, then transaction information on blockchains must have informational value and can update investors’ beliefs. The crowdfunding literature provides  22  evidence for ‘the wisdom of the crowd,’ which states that the crowd collectively has information about the potential of a project that is unknown to entrepreneurs. Following this argument, I expect that ICO investors have information about aggregate demand for projects. I predict that, for ICOs that disclose blockchain information, the number of transactions in early fundraising periods can predict future ICO outcomes. I develop my second hypothesis: H2: Blockchain information predicts future ICO outcomes. My third hypothesis also relates to ‘the wisdom of the crowd’ hypothesis. Not all projects in this market are legit projects, and scams are quite common. Although the information on blockchains becomes immutable once recorded, an ICO team may manipulate transaction records in real-time, pretending to be a hot project when there are no economic transactions. For example, an ICO can distribute a large number of tokens to random accounts, or it can recycle the raised cryptocurrencies to create constant investment inflows. There are many posts about suspicious ICO volume in the crypto community. However, it is an empirical question whether the market in aggregate can systematically identify suspicious blockchain transaction records and punish fraudulent ICO teams. I state my third hypothesis in the null form: H3: Investors can not identify fraudulent ICOs based on suspicious blockchain information.  23  2.4 Research Design 2.4.1 Data and Sample I collect a sample of ICOs that were launched on the Ethereum platform in 2018 from the ICObench, a website that provides ratings for ICO projects. This step yields a total of 2085 ICOs. ICObench contains basic ICO information, such as a description of their business, team members, ICO terms, funding status, and analyst ratings. Figure 2-1 provides information about the number of ICOs and the total funds raised in each month in 2018. The ICO became popular in 2016, reached its peak in late 2017. In 2018, the number of ICO projects and their ability to attract funds both show downward trends. I manually merge the sample with the Coinmarketcap, a website that contains price and volumes data for tokens listed on large crypto exchanges. I also manually search for LinkedIn profiles of ICO founders. I obtain blockchain data from Etherscan, a popular blockchain explorer. A major concern in the ICO market is the high failure rate and survivorship biases. Specifically, most ICO data providers frequently delete ICOs that fail to reach fundraising goals, and a non-random sample could bias the estimations in any directions. To mitigate survivorship bias, I choose ICObench as my primary data source. Unlike most ICO data providers that regularly delete finished ICOs, ICObench positions itself as a rating platform with information on ‘past, present, and future ICOs.’ They rarely remove ICOs and, particularly, would not remove ICOs because of bad ratings. Another way to mitigate survivorship bias is to focus on a recent sample, since information of an ICO that failed a long time ago may no longer be available. For example,  24  firms may abandon their websites and delete their whitepapers. When conducting the analyses in early 2019, I focus on a sample of ICOs that ends in 2018. It is worth noting that this recent sample is still sufficiently large and representative of all the ICOs because the market only became popular for more than two years at the time of analyses. The sample includes more ICOs than many other studies do, such as Adhami, Giudici, and Martinazzi (2018). Another concern is data quality. Due to the lack of regulations and the voluntary nature of the information disclosed on ICO data providers, there might be measurement errors. Some studies drop missing or suspicious observations, a strategy that is not implemented here because of concerns about survivorship biases. Instead, I supplement data from ICObench with data from ICOdata, another ICO data provider. Specifically, if both websites cover an ICO but provide inconsistent data, values from ICOdata are used in the analyses. According to Benedetti and Kostovetsky (2018), ICOdata provides more accurate data than ICObench. It is worth noting that survivorship biases and measurement errors do not affect variables related to blockchain information. I provide a detailed discussion about them in the next section. 2.4.2 Measures of Blockchain Disclosures Before I define measures for blockchain disclosures, I provide some institutional background on how investors interact with smart contracts when investing in ICOs. After reading the information in whitepapers and talking to team members on social media, investors will be provided with an Ethereum address to deposit their cryptocurrency investments. Appendix A.1 provides an example. If the ICO is transparent, this address is the smart contract address. Most  25  of the time, however, this is a one-time disposable account address, which is used to collect investments for a single investor. A smart contract address may be provided somewhere else. To access blockchain information, investors only need to copy and paste the smart contract address in a blockchain explorer, such as Etherscan. For ICOs without smart contracts, investors cannot get any information from the blockchain, i.e., no disclosure commitment. For ICOs with smart contracts, Etherscan will return a page with basic ICO information, such as name, website, token supplies, token holders, and token transfer records, i.e., partial commitment. Appendix A.3 shows examples of ICOs with and without smart contracts. Information on blockchains cannot be trusted unless the smart contract code is verified. Although all smart contracts are immutable once deployed, they are stored as byte code and are not human-readable. Developers can voluntarily disclose the human-readable source code on Etherscan, which is verified with the byte code on the Ethereum Virtual Machine. Appendix A.4 demonstrates examples of verified and unverified smart contracts. Based on the above discussion, I create my first dummy variable, partial_commitment, to identify ICOs that use smart contracts and verify the source code of their smart contract. These ICOs disclose credible inventory information, including token distribution records, while others do not.  Within the group of ICOs that at least make partial disclosure commitment, some allow investors to access cash flow information, i.e., inflows of Ethers to ICO projects. Recall that some ICOs use different one-time account addresses to collect cryptocurrencies from different investors. This fundraising method effectively hides cash flow information. In contrast, when a startup  26  launches an ICO through a public crowdsale contract, investors can observe cash flow information. In this case, investors can monitor startups’ accounts that store raised funds and observe withdrawals by team members. Appendix A.5 provides examples of ICOs that disclose or hide cash flow information. In Appendix A.6, I include a piece of computer code and pseudocode that governs the disclosure of this information.  Based on the analyses of smart contract code, I create the second dummy variable, full_commitment, to identify ICOs that also disclose cash flow information. Compared to the ICOs that only make partial disclosure commitments, these ICOs provide additional information about the inflow of cryptocurrencies. For each commitment decision, I summarize the information available to investors on blockchain in Appendix A.7. Both partial_commitment and full_commitment are measures of disclosure commitment on blockchains and are used to test hypothesis one. To measure the content of blockchain disclosures, I calculate the natural logarithm of the number of token transfers in early fundraising periods, early_token_transfers, for ICOs with at least partial commitments. This variable measures ICO funding progress and is used to test hypothesis two. Although the information on blockchains becomes immutable once recorded, an ICO team may manipulate transaction records in real-time, a concept similar to real-earnings management. For example, an ICO can distribute tokens to random accounts, pretending to be a hot project when there are no economic transactions.   27  I include a dummy variable, suspicious_volume, that identifies suspicious blockchain transactions. The intuition behind this variable is that an ICO is suspicious if it transfers the same number of tokens in many transactions because different people naturally make different amounts of investments. Specifically, for each ICO with blockchain information, I calculate the ratio of the number transactions that sell a unique amount of tokens to the number of all transactions, and I label an ICO suspicious if the ratio is below a threshold. For example, if an ICO sells one token each to three investors, the ratio is one-third. If an ICO sells one, two, three tokens to the three investors, respectively, the ratio is one. While there are more complicated ways to manipulate blockchain information and complicated ways to analyze such information, I choose this simple measure because one cannot expect average investors to conduct a complete analysis of blockchain information for each ICO. Whether there are many transactions with the same amount of tokens being transferred is the first impression when investors examine blockchain information and is often discussed in blockchain forums, such as Bitcointalk. 2.4.3 Measures of ICO Outcomes I employ both short-term and long-term measures for ICO outcomes. The short-term objective of ICOs is to raise enough funds. I construct three variables to measure short-term outcomes. First, I create a dummy variable, raised_dummy, to capture whether an ICO raised any funds. Second, following Lee, Li, and Shin (2018), I create a dummy variable, success, that equals to one if an ICO raised more than its soft cap, or more than $500,000 if it does not have a soft cap. Third, I  28  include the natural logarithm of the raised funds in US dollars, log_raised, as an outcome variable. The long-term objectives of ICOs are to list their tokens on crypto exchanges and to continue their business plans or product development. Note that the ‘long-term’ is relative to fundraising goals. According to Benedetti and Kostovetsky (2018), the average time between the end of ICOs and the first day of trading is 31 days. According to a Bloomberg report, fewer than half of ICOs survive four months. I use a dummy variable, listing, to capture whether an ICO is listed and the trading information is available in the Coinmarketcap. I use another dummy variable, web_active, to capture if an ICO website is still working and updating after the ICO ends. Finally, I use a dummy variable, products, to capture if an ICO has delivered preliminary products, such as a cell phone application or a web interface linked to its own blockchain. I measured all the variables in July 2019. 2.4.4 Control Variables To account for other determinants of ICOs success, I follow previous literature and include a battery of control variables, including ICO characteristics and ICO founder characteristics.  I use the number of team members as a proxy for the size of a project. I include alternative disclosure channels such as social media, GitHub, and whitepapers because previous literature suggests that disclosures in those channels can effectively reduce information asymmetry between startups and investors. I include other ICO characteristics, such as hard and soft cap requirements, private ICO round, and participation restrictions. I proxy for founders’ reputation using prior startup  29  experiences, founders’ expertise using prior blockchain experiences, and general capability using the founders’ educational background. A complete list of variable definitions is in Appendix A.1. 2.4.5 Empirical Models To account for factors that simultaneously affect ICOs’ blockchain disclosure decisions and ICO outcomes, I first examine the determinants of disclosure decisions by estimating the following logit model:  commitment decision = ∂ +∑𝛽𝑖 ∗ 𝑑𝑒𝑡𝑒𝑟𝑚𝑖𝑛𝑎𝑛𝑡𝑖𝑛𝑖=1+ 𝜖, (2.1) where the dependent variable is either partial_commitment or full_commitment. The determinants include both ICO characteristics, such as log_team, social, pre_sale, log_ico_len, soft_req, hard_req, and github, and founders’ characteristics, such as num_prior_starups, prior_blockchain_exp, and high_edu. To test H1 that ICOs that commit to more blockchains disclosures are more likely to succeed, I estimate the following logit model of ICO outcomes on disclosure decision, controlling for other factors of ICO outcomes:  ICO outcomes = ∂ + β1 ∗ 𝑐𝑜𝑚𝑚𝑖𝑡𝑚𝑒𝑛𝑡 𝑑𝑒𝑐𝑖𝑠𝑖𝑜𝑛 +∑𝛽𝑖 ∗ 𝑐𝑜𝑛𝑡𝑟𝑜𝑙𝑠𝑖𝑛𝑖=2+ 𝜖, (2.2)  30  where the dependent variable is one of raised_dummy, log_raised, success, web_active, listing, and product.  H1 predicts that β1 is positive, suggesting that investors respond favorably to ICOs that voluntarily disclose more blockchain information. To test H2 about the informativeness of blockchain information, I regress the ICO outcomes on the number of token transfers in early ICO periods, early_token_transfers. Specifically, I estimate the following logit model for the subsamples of ICO with at least partial disclosures:  ICO outcomes = ∂ + β1 ∗ 𝑒𝑎𝑟𝑙𝑦_𝑡𝑜𝑘𝑒𝑛_𝑡𝑟𝑎𝑛𝑠𝑓𝑒𝑟𝑠 +∑𝛽𝑖 ∗ 𝑐𝑜𝑛𝑡𝑟𝑜𝑙𝑠𝑖𝑛𝑖=2+ 𝜖. (2.3) H2 predicts that β1 is positive, suggesting that the information on blockchains is indicative of future ICO outcomes and helps investors to update their beliefs about ICOs potentials. To test H3 about whether investors can identify and punish ICO projects with suspicious volume, I add the dummy variable, suspicious_volume, to the model (3),   ICO outcomes = ∂ + β1 ∗ 𝑠𝑢𝑠𝑝𝑖𝑐𝑖𝑜𝑢𝑠_𝑣𝑜𝑙𝑢𝑚𝑒 +                                        +β2 ∗ 𝑒𝑎𝑟𝑙𝑦_𝑡𝑜𝑘𝑒𝑛_𝑡𝑟𝑎𝑛𝑠𝑓𝑒𝑟𝑠                           +∑𝛽𝑖 ∗ 𝑐𝑜𝑛𝑡𝑟𝑜𝑙𝑠𝑖𝑛𝑖=3+ 𝜖. (2.4) If investors can identify and punish suspicious ICOs, I expect that  β1 is negative.  31  2.5 Empirical Results 2.5.1 Summary Statistics Table 2-1 provides industry classifications and geographic distributions of ICOs in my sample. There are no standard industry definitions in the crypto market. ICOs label themselves several tags when submitting information to ICO data providers. Thus, a given ICO can have multiple tags and is in many industries. The most common category is platforms. Therefore, people often refer to ICO as a form of platform building. The second common category is cryptocurrencies. ICOs also self-report their office locations and the market of their interests. Countries that are more open to the blockchain technology, such as the US, Singapore, and the UK, have the most ICOs. Table 2-2 provides summary statistics about the variables used in this study. All continuous variables are winsorized at one percent each tail. Panel A provides statistics for the full sample. About 30 to 40 percent of ICOs succeeded in reaching their fundraising goals, and only 17 percent of tokens are listed on major exchanges. These statistics highlight the high failure rate and the risky nature of the ICO market and are mostly consistent with previous papers, e.g., Benedetti and Kostovetsky (2018).  The whitepaper variable does not have much variation: almost all ICOs provided a whitepaper. At the time of the analyses, many whitepapers are no longer available, so it is impossible to follow other studies to measure disclosure qualities in whitepapers. To mitigate the concern of omitted variables, I include the variable of analyst ratings in the robustness tests. To the extent  32  that analysts rely on voluntary disclosures from ICO firms to make recommendations, the rating variable could capture information quality from different disclosure channels, including whitepapers. Figure 2-1 provides the decomposition of ICOs based on their blockchain disclosures. Figure 2-2: Sample DecompositionSurprisingly, about 40 percent of ICOs do not use smart contracts and make no disclosure commitments in their fundraising processes. According to ICObench, however, all ICOs in my sample claimed that they used smart contracts. Among ICOs that have verified smart contracts and make partial disclosure commitment, about one-third use public crowdsale contracts and make full disclosure commitment. The statistics document significant variations in the transparency of smart contracts and ICO projects. Although smart contracts provide options for ICOs to make commitments and be transparent, many ICOs do not adopt those options. Table 2-2 partitions my sample based on the commitment decisions and compares summary statistics between different groups. Panel B compares ICOs with no disclosure commitment with those who make a partial commitment. These two groups are very different in almost all aspects, except that they all had whitepapers. Based on univariate comparisons, ICOs with smart contracts have better profiles in terms of social media and GitHub information. Founders in those ICOs have more reputation and experiences. These ICOs raised more money and were more likely to list their tokens on exchanges. For the subsample of ICOs that make at least partial commitments, Panel C compares ICOs that disclose the flows of cryptocurrencies to those that do not. While the profiles between the two  33  groups are not overwhelmingly different, except for social media and GitHub, ICO outcomes do differ significantly. ICOs with transparent smart contracts, on average, are more successful. 2.5.2 Determinants of Blockchain Disclosures Table 2-4 reports results for the determinants of blockchain disclosure decisions. I find that the uses of GitHub and social media are positively related to more extensive disclosure on blockchains, suggesting complements among different disclosure channels. Prior blockchain experiences are a proxy for founders’ expertise and costs to implement blockchain technology. The positive relation between prior blockchain experiences and blockchain disclosure is consistent with the argument that people who have comparative advantages in blockchain will use it more extensively.  One interesting observation is the negative signs for the coefficient of soft cap requirement, soft_req, in the model where the dependent variable is full_commitment. Recall that the soft cap stems from crowdfunding and requires that invested funds be returned to investors in the event of not reaching the predetermined minimum amount. This requirement can be hardcoded in smart contracts. However, very few ICOs implement such codes. Most ICOs either transfer raised funds to team accounts or hide them completely. In other words, soft cap requirements are the promises not kept in most cases. The results suggest that these promises are less likely to be made by ICOs that make more disclosure commitments.  34  2.5.3 The Test for Hypothesis One Table 2-5 reports the main results. The hypothesis one predicts that ICOs that make partial disclosures are more successful than those without any disclosure commitment. It also predicts that, within the group of ICOs with partial commitment, ICOs that are more transparent about cash flow information, i.e., make full commitments, are more successful than those that are not. Those predictions are tested in multivariate tests. For example, in the model in Panel A where success is the dependent variable, the coefficient of partial_commitment is 0. 84 with a t-stat of 7.39. The marginal effect is 16 percent. In the model in Panel B where success is the dependent variable, the coefficient of full_commitment is 0.33 with a t-stat of 2.67. The marginal effect is 6 percent. These results are both statistically and economically significant. ICOs that disclose more information on blockchains are 22 percent more likely to reach their fundraising goals. The results are consistent across the different measures of ICO outcomes after controlling for other ICO and founders’ characteristics. 2.5.4 The Test for Hypothesis Two Table 2-6 reports the results for hypothesis two. Hypothesis two explores the mechanism that investors benefit from information on blockchains. If the information on blockchains correctly updates investors’ beliefs about ICO outcomes, the early investments must be informative of future outcomes. Table 2-6, Panel A reports regression results for the subsample that only make partial disclosure commitments. I find some evidence, i.e., three out of six specifications, that the number of the token distribution in early ICO stages predicts ICO outcomes. Table 2-6, Panel B reports the regression results for the subsample that make full disclosure commitments, and  35  results are consistently significant only in this subsample. The prediction powers of early investments are also greater in this subsample, as indicated by the F tests. The evidence is consistent with the argument that when ICOs commit to more disclosures, the contents of the disclosure becomes more informative. The evidence helps to explain why investors respond favorably to ICOs that voluntarily make full disclosure commitments. To further understand the relationship between blockchain information and the ICO outcomes, I split each subsample based on the quintiles of blockchain information in Table 2-6, Panel C. The quintiles are calculated based on the full commitment sample. I calculate the mean for the ICO outcomes for each quintile. Panel C does not show a perfect linear relation between the signals and the ICO outcomes. Instead, the ICO outcomes are generally better in the first quintile than in the second quintile, which is particularly the case in the partial commitment sample. The evidence is consistent with the argument that when a firm commits to an increased level of disclosures, the signal becomes more precise, and it is more difficult to hide bad news. 2.5.5 The Test for Hypothesis Three Table 2-7 reports the results for hypothesis three. In the tests, I regress ICO outcomes on blockchain information and add the variable, suspicious_volume, that captures potential manipulations of blockchain information. The coefficients on the dummy variable are significantly negative for all ICO outcomes, consistent with the argument that investors on aggregate can identify fraudulent ICOs, and those ICOs, on average, raise fewer funds. Also, these ICO projects are less likely to survive or to deliver their products or services.   36  2.5.6 Robustness Tests I include several robustness checks. First, it might be a concern that the number of team members is a poor proxy for size, and that the amount of raised funds is not properly scaled. A common way to address this concern in the literature is to scale raised amount by the hard cap. It is challenging to implement the procedures for two reasons. On the one hand, some ICOs do not have hard caps. On the other hand, some ICOs express their hard caps in terms of Ethers, and the exchange rate of Ethers to the US dollars fluctuates, making it difficult to measure the hard cap accurately. I select a sample of observations whose hard cap can be measured reliably. I run the main tests for this sample and replace the dependent variable with the amount of raised funds scaled by hard caps. The results are presented in Table 2-8, Panel A, and are consistent with the main results. Then I show that the information value and the prediction power of blockchain disclosure are robust to alternative definitions of blockchain information. I replace token distributions in early ICO periods with token distribution on the first day since ICO starts and rerun the test for hypothesis two. In Table 2-8, Panel B, and C, I still find that blockchain information predicts future ICO outcomes, and the prediction power is greater for the subsample that makes full disclosure commitments. In addition, I show that the dummy variables that identify suspicious ICOs are robust to the choice of threshold. When raising the threshold to 10 percent, I potentially introduce more noise to the measure of suspicious ICO volumes. In Table 2-8 panel D, I still find that the coefficients  37  of suspicious_10% are significantly negative in four out of six specifications at the ten percent level. Finally, previous studies document that ICO analysts can collect relevant information, and their ratings are informative of ICO outcomes. ICO analysts rely on voluntary disclosure from ICO firms to make recommendations. Therefore, the rating variable should correlate with some other control variables and can mitigate the concerns about measurement errors in the whitepaper variable. In Table 2-8, Panel E, I include analysts’ ratings in the main tests. The variable absorbs some explanatory power of social media and team size, suggesting that analysts provide ratings based on information on social media and from ICO firms. The disclosure commitments variables are almost not affected by the inclusion of the analyst rating variables. 2.6 Conclusions In this study, I examine how blockchain technology and smart contracts are used in the ICO market. I argue that smart contracts serve as a tool for disclosure commitments. They can allow investors to observe the token distributions and cryptocurrency investments and to assess the popularity and potential of an ICO project. Startups that voluntarily commit to disclosing such information on blockchains are more likely to succeed. The disclosed information is indicative of ICO outcomes, and the market on aggregate can identify and punish suspicious ICOs. The findings are consistent with the argument that disclosure commitments signal for project qualities and support the ‘wisdom of the crowd’ hypothesis in the ICO market.    38  Figure 2-1: ICO Statistics in 2018    Figure 2-2: Sample Decomposition     39  Figure 2-3: Transitions from Public ICOs to Private ICOs         40  Table 2-1: Sample Compositions Industry Number of ICOs Country/Region Number of ICOs Platform 1140 USA 230 Cryptocurrency 889 Singapore 227 Business services 537 UK 192 Investment 434 Estonia 121 Smart Contract 371 Switzerland 113 Software 332 Russia 106 Internet 270 Hong Kong 76 Entertainment 228 Germany 60 Banking 224 Cayman Islands 55 Infrastructure 218 Australia 40 Artificial Intelligence 199 Canada 38 Big Data 195 Gibraltar 34 Communication 181 India 32 Media 161 Netherlands 32 Retail 124 Malta 32 Other 116 United Arab Emirates 29 Health 104 British Virgin Islands 27 Real estate 101 Cyprus 24 Education 91 France 23 Tourism 73 Slovenia 23 Energy 63 Seychelles 22 Manufacturing 61 Ukraine 20 Sports 59 Czech Republic 20 Casino & Gambling 57 Belize 20 Charity 52 Japan 18 Virtual Reality 51 Bulgaria 18 Legal 51 Romania 18 Electronics 44 China 17 Art 35 South Africa 16     Latvia 16     41  Table 2-2: Summary Statistics Panel A: Full sample                 Variables N mean std min 25% 50% 75% max ICO outcomes         raised_dummy 2085 0.4 0.49 0 0 0 1 1 success 2085 0.31 0.46 0 0 0 1 1 log_raised 2085 5.94 7.44 0 0 0 14.73 17.73 listing 2085 0.17 0.38 0 0 0 0 1 web_active 2085 0.47 0.5 0 0 0 1 1 products 2085 0.14 0.34 0 0 0 0 1          Blockchain Variables         partial_commitment 2085 0.61 0.49 0 0 1 1 1 full_commitment 2085 0.23 0.42 0 0 0 0 1 early_token_transfers 1282 2.26 2.53 0 0 1.39 3.78 9.18 suspicious_volume 1282 0.04 0.19 0 0 0 0 1 suspicious_10% 1282 0.08 0.27 0 0 0 0 1          ICO Characteristics         log_team 2085 2.05 0.7 0 1.79 2.2 2.48 3.37 social 2085 0.8 0.24 0 0.67 0.83 1 1 pre_sale 2085 0.53 0.5 0 0 1 1 1 log_ico_len 2085 3.76 0.84 0 3.4 3.81 4.29 5.37 soft_req 2085 0.6 0.49 0 0 1 1 1 hard_req 2085 0.79 0.4 0 1 1 1 1 github 2085 0.55 0.5 0 0 1 1 1 restrict_us 2085 0.34 0.47 0 0 0 1 1 whitepaper 2085 0.98 0.13 0 1 1 1 1          Founder Characteristics         num_prior_startups 2085 0.17 0.39 0 0 0 0 1.61 prior_blockchain_exp 2085 0.06 0.24 0 0 0 0 1 high_edu 2085 0.19 0.39 0 0 0 0 1     42  Panel B: Comparison between the samples without commitments and with commitments  Variables No Commitment Commitment Difference  T-stats  raised_dummy 0.26 0.48  0.23***      10.81  success 0.18 0.38  0.20***      10.42  log_raised 3.71 7.34  3.63***      11.62  listing 0.05 0.25  0.20***           13.86  web_active 0.34 0.55  0.22***             9.91  products 0.06 0.18  0.12***             8.86  log_team 1.99 2.09  0.11***        3.45  social 0.74 0.83  0.09***        8.60  pre_sale 0.5 0.55  0.05**         2.05  log_ico_len 3.86 3.7 -0.16***        4.60  soft_req 0.55 0.63  0.08***        3.58  hard_req 0.75 0.82  0.07***        3.67  github 0.45 0.62  0.17***        7.71  restrict_us 0.3 0.36  0.05***        2.60  whitepaper 0.98 0.99  0.01           1.39  num_prior_startups 0.12 0.19  0.07***        4.20  prior_blockchain_exp 0.03 0.08  0.05***        5.01  high_edu 0.17 0.2  0.03*          1.93  N 803 1282         43  Panel C: Comparison between the samples with partial commitments and with full commitments Variables Partial Commitment Full Commitment Difference  T-stats  raised_dummy 0.44 0.55  0.11***        3.80  success 0.35 0.45  0.10***        3.63  log_raised 6.7 8.43  1.73***        3.89  listing 0.21 0.3  0.09***             3.57  web_active 0.52 0.61  0.10***             3.41  products 0.15 0.23  0.08***             3.36  log_team 2.08 2.12  0.04           1.07  social 0.82 0.85  0.03**         2.10  pre_sale 0.57 0.52 -0.05*          1.84  log_ico_len 3.72 3.66 -0.07           1.26  soft_req 0.66 0.58 -0.08***        2.82  hard_req 0.81 0.84  0.04*          1.73  github 0.58 0.67  0.09***        3.25  restrict_us 0.37 0.33 -0.04           1.30  whitepaper 0.98 0.99  0.01           1.04  num_prior_startups 0.18 0.22  0.04*          1.66  prior_blockchain_exp 0.07 0.1  0.03*          1.85  high_edu 0.2 0.21  0.01           0.24  N 807 475      44   Table 2-3: Correlation Matrix 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 201 raised_dummy  1 0.82* 0.96* 0.32* 0.29* 0.17* 0.22* 0.18* 0.18* 0.21* 0.01 -0.07* 0.08* 0.16* 0.11* 0.04 0.06* 0.14* 0.08* 0.10*2 success 0.82*  1 0.87* 0.36* 0.28* 0.20* 0.21* 0.17* 0.17* 0.19* -0.02 -0.10* 0.02 0.11* 0.11* 0.04 0.06* 0.15* 0.10* 0.08*3 log_raised 0.99* 0.86*  1 0.40* 0.32* 0.22* 0.24* 0.19* 0.19* 0.21* -0.00 -0.11* 0.05 0.15* 0.10* 0.04 0.05 0.17* 0.09* 0.10*4 listing 0.32* 0.36* 0.36*  1 0.33* 0.33* 0.26* 0.19* 0.11* 0.16* 0.00 -0.19* -0.04 0.08* 0.11* 0.04 0.02 0.17* 0.10* 0.09*5 web_active 0.29* 0.28* 0.31* 0.33*  1 0.42* 0.21* 0.16* 0.20* 0.20* 0.01 -0.07* 0.02 0.10* 0.13* 0.07* 0.07* 0.23* 0.09* 0.11*6 products 0.17* 0.20* 0.20* 0.33* 0.42*  1 0.17* 0.15* 0.11* 0.14* -0.02 -0.10* -0.02 0.04 0.10* 0.03 0.01 0.11* 0.04 0.08*7 partial_commitment 0.22* 0.21* 0.24* 0.26* 0.21* 0.17*  1 0.43* 0.08* 0.18* 0.04 -0.06* 0.08* 0.08* 0.17* 0.03 0.06 0.10* 0.10* 0.048 full_commitment 0.18* 0.17* 0.18* 0.19* 0.16* 0.15* 0.43*  1 0.06* 0.11* -0.02 -0.05 -0.02 0.07* 0.13* 0.03 -0.00 0.08* 0.09* 0.029 log_team 0.19* 0.16* 0.19* 0.12* 0.20* 0.11* 0.08* 0.05  1 0.26* 0.11* -0.00 0.11* 0.15* 0.16* 0.02 0.15* 0.19* 0.07* 0.09*10 social 0.21* 0.18* 0.21* 0.16* 0.20* 0.15* 0.19* 0.12* 0.24*  1 0.21* -0.01 0.25* 0.28* 0.43* 0.13* 0.23* 0.19* 0.10* 0.15*11 pre_sale 0.01 -0.02 0.01 0.00 0.01 -0.02 0.04 -0.02 0.10* 0.21*  1 -0.09* 0.16* 0.13* 0.11* 0.07* 0.08* 0.05 0.02 0.06*12 log_ico_len -0.11* -0.15* -0.14* -0.24* -0.11* -0.15* -0.09* -0.07* -0.02 -0.03 -0.06*  1 0.07* 0.03 -0.02 -0.04 -0.07* -0.10* -0.04 -0.0313 soft_req 0.08* 0.02 0.07* -0.04 0.02 -0.02 0.08* -0.02 0.10* 0.25* 0.16* 0.06*  1 0.51* 0.17* 0.06* 0.12* 0.03 -0.03 0.06*14 hard_req 0.16* 0.11* 0.16* 0.08* 0.10* 0.04 0.08* 0.07* 0.16* 0.28* 0.13* 0.01 0.51*  1 0.20* 0.09* 0.16* 0.11* 0.02 0.10*15 github 0.11* 0.11* 0.11* 0.11* 0.13* 0.10* 0.17* 0.13* 0.13* 0.42* 0.11* -0.02 0.17* 0.20*  1 0.09* 0.11* 0.12* 0.05 0.09*16 whitepaper 0.04 0.04 0.04 0.04 0.07* 0.03 0.03 0.03 0.04 0.13* 0.07* -0.04 0.06* 0.09* 0.09*  1 0.06* 0.04 0.02 0.0517 restrict_us 0.06* 0.06* 0.06* 0.02 0.07* 0.01 0.06 -0.00 0.15* 0.22* 0.08* -0.06* 0.12* 0.16* 0.11* 0.06*  1 0.09* 0.07* 0.09*18 num_prior_startups 0.14* 0.15* 0.15* 0.15* 0.22* 0.09* 0.09* 0.07* 0.17* 0.17* 0.04 -0.10* 0.03 0.11* 0.11* 0.04 0.08*  1 0.26* 0.14*19 prior_blockchain_exp 0.08* 0.10* 0.08* 0.10* 0.09* 0.04 0.10* 0.09* 0.08* 0.10* 0.02 -0.02 -0.03 0.02 0.05 0.02 0.07* 0.26*  1 0.11*20 high_edu 0.10* 0.08* 0.10* 0.09* 0.11* 0.08* 0.04 0.02 0.11* 0.15* 0.06* -0.04 0.06* 0.10* 0.09* 0.05 0.09* 0.13* 0.11*  1* significant at the 1 percent, 45  Table 2-4: Determinants of Blockchain Disclosures   partial_commitment full_commitment Intercept  0.03      -2.08***   (0.06)     (-3.58)    log_team  0.06       0.07      (0.79)     (0.80)     social  1.11***    0.94***   (4.72)     (3.19)     pre_sale -0.04      -0.22**    (-0.38)    (-1.98)    log_ico_len -0.24***   -0.17***   (-4.41)    (-2.64)    soft_req  0.16      -0.41***   (1.44)     (-3.29)    hard_req  0.02       0.47***   (0.17)     (2.86)     github  0.42***    0.48***   (4.14)     (4.02)     restrict_us -0.01      -0.21*     (-0.09)    (-1.80)    whitepaper -0.03       0.29      (-0.08)    (0.58)     num_prior_startups  0.15       0.12      (1.13)     (0.86)     prior_blockchain_exp  0.80***    0.56***   (3.27)     (2.73)     high_edu -0.03      -0.06      (-0.22)    (-0.44)    N 2085 2085 pseudo-R2 0.05 0.04 ***, **, * significant at the 1 percent, 5 percent, and 10 percent levels, respectively. T-stats in parentheses.    46  Table 2-5: Blockchain Disclosures and ICO Outcomes (H1) Panel A: Partial disclosure commitments and ICO outcomes        raised_ dummy success log_ raised listing web_ active products Intercept  -2.60***   -2.45***    0.82      -3.46***   -2.53***   -3.98***    (-5.31)    (-4.50)    (0.63)     (-4.20)    (-4.70)    (-4.84)    partial_commitment +  0.82***    0.84***    2.80***    1.65***    0.72***    1.02***    (7.97)     (7.39)     (8.92)     (9.02)     (7.20)     (6.11)     log_team +  0.44***    0.41***    1.36***    0.29***    0.44***    0.39***    (5.71)     (4.91)     (6.46)     (2.69)     (5.80)     (3.64)     social +  1.30***    1.27***    3.82***    1.69***    1.07***    1.90***    (4.94)     (4.33)     (5.11)     (4.18)     (4.34)     (4.75)     pre_sale ? -0.26***   -0.36***   -0.89***   -0.20      -0.23**    -0.35***    (-2.66)    (-3.43)    (-2.84)    (-1.50)    (-2.31)    (-2.59)    log_ico_len - -0.24***   -0.31***   -0.99***   -0.60***   -0.20***   -0.36***    (-4.10)    (-5.17)    (-5.32)    (-7.99)    (-3.59)    (-5.07)    soft_req ? -0.08      -0.29**    -0.40      -0.69***   -0.29**    -0.40**     (-0.71)    (-2.38)    (-1.05)    (-4.65)    (-2.49)    (-2.54)    hard_req +  0.67***    0.55***    1.92***    0.59***    0.30**     0.08       (4.41)     (3.43)     (4.35)     (2.90)     (2.11)     (0.38)     github + -0.06       0.02      -0.19       0.14       0.08       0.16       (-0.60)    (0.16)     (-0.55)    (0.96)     (0.80)     (1.05)     whitepaper +  0.18       0.12       0.29       0.64       0.78*      0.53      (0.46)     (0.28)     (0.28)     (0.90)     (1.73)     (0.73)     restrict_us - -0.11      -0.03      -0.32      -0.29**    -0.01      -0.23       (-1.11)    (-0.23)    (-0.94)    (-2.11)    (-0.07)    (-1.61)    num_prior_startups +  0.28**     0.37***    1.30***    0.40***    0.83***    0.19       (2.26)     (2.91)     (2.91)     (2.62)     (5.90)     (1.20)     prior_blockchain_exp +  0.18       0.35*      0.63       0.34       0.09      -0.10       (0.88)     (1.71)     (0.89)     (1.45)     (0.44)     (-0.39)    high_edu +  0.24**     0.13       0.83**     0.32**     0.31**     0.30*      (2.01)     (1.07)     (2.01)     (2.09)     (2.51)     (1.94)     N   2085  2085  2085  2085  2085  2085 pseudo-R2/adj_R2   0.10  0.10  0.13  0.18  0.10  0.10 ***, **, * significant at the 1 percent, 5 percent, and 10 percent levels, respectively. T-stats in parentheses.    47  Panel B: Full disclosure commitments and ICO outcomes         raised_ dummy success log_ raised listing web_ active products Intercept  -2.58***   -2.43***    0.88      -3.39***   -2.52***   -3.92***    (-5.31)    (-4.45)    (0.68)     (-4.16)    (-4.70)    (-4.77)    partial_commitment +  0.69***    0.71***    2.31***    1.52***    0.61***    0.87***    (6.10)     (5.76)     (6.64)     (7.93)     (5.61)     (4.83)     full_commitment +  0.37***    0.33***    1.39***    0.30**     0.31**     0.37**     (3.05)     (2.67)     (3.23)     (2.10)     (2.48)     (2.43)     log_team +  0.44***    0.41***    1.35***    0.28***    0.44***    0.38***    (5.71)     (4.88)     (6.46)     (2.67)     (5.76)     (3.59)     social +  1.28***    1.25***    3.75***    1.66***    1.05***    1.87***    (4.86)     (4.26)     (5.03)     (4.11)     (4.26)     (4.66)     pre_sale ? -0.25**    -0.35***   -0.84***   -0.19      -0.21**    -0.34**     (-2.52)    (-3.29)    (-2.68)    (-1.39)    (-2.19)    (-2.46)    log_ico_len - -0.23***   -0.31***   -0.97***   -0.59***   -0.20***   -0.35***    (-4.05)    (-5.13)    (-5.27)    (-8.03)    (-3.53)    (-5.04)    soft_req ? -0.05      -0.26**    -0.28      -0.65***   -0.26**    -0.36**     (-0.41)    (-2.10)    (-0.74)    (-4.39)    (-2.24)    (-2.27)    hard_req +  0.64***    0.52***    1.81***    0.56***    0.27*      0.04       (4.19)     (3.24)     (4.11)     (2.74)     (1.93)     (0.20)     github + -0.08       0.00      -0.26       0.12       0.07       0.14       (-0.78)    (0.01)     (-0.75)    (0.84)     (0.66)     (0.93)     whitepaper +  0.17       0.11       0.23       0.61       0.77*      0.50        (0.44)     (0.25)     (0.23)     (0.86)     (1.72)     (0.68)     restrict_us - -0.10      -0.01      -0.27      -0.27**     0.00      -0.21       (-0.96)    (-0.09)    (-0.80)    (-1.99)    (0.04)     (-1.47)    num_prior_startups +  0.28**     0.37***    1.29***    0.39***    0.83***    0.18       (2.22)     (2.87)     (2.87)     (2.59)     (5.88)     (1.19)     prior_blockchain_exp +  0.16       0.33       0.55       0.32       0.08      -0.14       (0.77)     (1.62)     (0.78)     (1.36)     (0.38)     (-0.52)    high_edu +  0.25**     0.14       0.84**     0.32**     0.31**     0.31*       (2.05)     (1.09)     (2.05)     (2.09)     (2.54)     (1.95)     N   2085  2085  2085  2085  2085  2085 pseudo-R2/adj_R2    0.10  0.10  0.14  0.18  0.11  0.11 ***, **, * significant at the 1 percent, 5 percent, and 10 percent levels, respectively. T-stats in parentheses.    48  Table 2-6: Predicting ICO Outcomes with Blockchain Information (H2) Panel A: Predictive power for the sample that only makes partial disclosure commitments.    raised_ dummy success log_ raised listing web_ active products Intercept  -2.35***   -1.81**     0.87      -2.72***   -2.22***   -3.19***    (-3.07)    (-2.29)    (0.37)     (-2.92)    (-2.79)    (-3.28)    early_token_transfers +  0.02       0.05       0.09       0.09**     0.11***    0.12***    (0.67)     (1.36)     (0.73)     (2.13)     (3.05)     (2.78)     log_team +  0.37***    0.32**     1.29***    0.26*      0.44***    0.28*      (3.18)     (2.56)     (3.53)     (1.65)     (3.61)     (1.78)     social +  1.88***    1.91***    6.16***    1.56***    1.07***    1.58**     (4.41)     (4.09)     (4.64)     (2.71)     (2.58)     (2.40)     pre_sale ? -0.30*     -0.34**    -0.99*     -0.01      -0.08      -0.03       (-1.91)    (-2.08)    (-1.84)    (-0.04)    (-0.49)    (-0.14)    log_ico_len - -0.17**    -0.27***   -0.76**    -0.49***   -0.30***   -0.39***    (-2.04)    (-3.16)    (-2.55)    (-5.12)    (-3.37)    (-3.78)    soft_req ? -0.21      -0.44**    -0.85      -0.75***   -0.41**    -0.72***    (-1.09)    (-2.13)    (-1.24)    (-3.45)    (-2.06)    (-2.99)    hard_req +  0.73***    0.69***    2.45***    0.56*      0.70***    0.53       (3.02)     (2.68)     (3.14)     (1.92)     (2.97)     (1.62)     github + -0.14      -0.17      -0.53       0.17       0.03       0.28       (-0.85)    (-1.01)    (-0.92)    (0.81)     (0.16)     (1.15)     whitepaper +  0.12      -0.32      -0.10       1.00       0.99       0.55        (0.19)     (-0.51)    (-0.05)    (1.28)     (1.55)     (0.67)     restrict_us - -0.10       0.10      -0.35      -0.24      -0.13      -0.11       (-0.65)    (0.58)     (-0.61)    (-1.20)    (-0.82)    (-0.52)    num_prior_startups +  0.37*      0.46**     1.76**     0.59***    0.80***    0.36       (1.85)     (2.34)     (2.36)     (2.72)     (3.31)     (1.57)     prior_blockchain_exp + -0.00       0.10      -0.08       0.40       0.06      -0.12       (-0.00)    (0.33)     (-0.07)    (1.20)     (0.19)     (-0.31)    high_edu +  0.20       0.10       0.67       0.18       0.49**     0.21       (1.05)     (0.51)     (0.97)     (0.77)     (2.48)     (0.83)     N   807  807  807  807  807  807 pseudo-R2/adj_R2    0.07  0.07  0.08  0.11  0.10  0.08 ***, **, * significant at the 1 percent, 5 percent, and 10 percent levels, respectively. T-stats in parentheses.    49  Panel B: Predictive power for the sample that makes full disclosure commitment.    raised_ dummy success log_ raised listing web_ active products Intercept  -0.17      -1.53       6.84      -1.31      -2.08      -2.86*      (-0.16)    (-1.13)    (1.58)     (-0.93)    (-1.56)    (-1.72)    early_token_transfers +  0.27***    0.27***    0.85***    0.39***    0.24***    0.13***    (5.96)     (6.17)     (6.77)     (7.50)     (5.20)     (2.81)     log_team +  0.48***    0.49***    1.61***    0.31*      0.47***    0.41**     (3.05)     (2.87)     (3.59)     (1.72)     (2.87)     (2.29)     social +  0.78       0.62       2.55       1.68*      1.76***    2.87***    (1.29)     (0.95)     (1.36)     (1.91)     (2.94)     (3.17)     pre_sale ?  0.18      -0.09       0.40      -0.18      -0.17      -0.32       (0.86)     (-0.43)    (0.61)     (-0.74)    (-0.78)    (-1.34)    log_ico_len - -0.53***   -0.55***   -1.63***   -0.75***   -0.27**    -0.39***    (-4.38)    (-4.40)    (-5.21)    (-4.76)    (-2.47)    (-3.22)    soft_req ?  0.12      -0.04       0.17      -0.49*      0.03      -0.04       (0.49)     (-0.15)    (0.22)     (-1.82)    (0.13)     (-0.15)    hard_req +  0.71**     0.38       1.84*      0.38      -0.23      -0.54       (2.20)     (1.22)     (1.76)     (1.03)     (-0.70)    (-1.50)    github +  0.02       0.23      -0.06       0.26       0.18       0.26       (0.08)     (0.92)     (-0.08)    (0.92)     (0.72)     (0.96)     whitepaper + -0.99       0.27      -3.06      -0.78       0.24      -0.52        (-1.15)    (0.22)     (-0.80)    (-0.70)    (0.21)     (-0.36)    restrict_us - -0.41*     -0.04      -0.97      -0.10      -0.15      -0.30       (-1.78)    (-0.19)    (-1.38)    (-0.37)    (-0.66)    (-1.14)    num_prior_startups + -0.07      -0.00       0.30       0.15       0.90***    0.20       (-0.25)    (-0.01)    (0.36)     (0.52)     (3.00)     (0.72)     prior_blockchain_exp +  0.71*      0.87**     1.86*      0.07       0.40      -0.19       (1.76)     (2.24)     (1.72)     (0.15)     (0.91)     (-0.46)    high_edu +  0.47*      0.31       1.44*      0.73**     0.36       0.71***    (1.78)     (1.23)     (1.82)     (2.51)     (1.35)     (2.68)     N   475  475  475  475  475  475 pseudo-R2/adj_R2    0.16  0.17  0.20  0.26  0.16  0.13         F test of the equality of coefficients of early_token_transfers between two samples F stats   (3.57)     (2.94)     (3.60)     (2.46)     (1.74)     (0.66)     ***, **, * significant at the 1 percent, 5 percent, and 10 percent levels, respectively. T-stats in parentheses.    50  Panel C: Summary statistics based on the quintiles of blockchain information Sample Partial commitment   Full commitment early_token_transfers quintiles 1 2 3 4 5  1 2 3 4 5 Number of ICOs 478 123 91 53 56   94 96 94 96 88             raised_dummy 0.45 0.4 0.43 0.49 0.46  0.43 0.42 0.41 0.67 0.82 success 0.34 0.32 0.34 0.34 0.43  0.31 0.32 0.28 0.57 0.74 log_raised 6.79 6 6.23 7.26 7.23  6.48 6.08 6.06 10.12 13.1 listing 0.22 0.15 0.14 0.21 0.34  0.19 0.08 0.2 0.34 0.68 web_active 0.51 0.41 0.54 0.62 0.62  0.5 0.45 0.52 0.74 0.85 products 0.14 0.15 0.12 0.25 0.25  0.19 0.12 0.15 0.33 0.36             pre_sale 0.6 0.51 0.49 0.57 0.54  0.54 0.5 0.6 0.5 0.44 num_prior_startups 0.21 0.15 0.1 0.13 0.12   0.24 0.15 0.16 0.3 0.25 The quintiles are calculated from the sample that makes full disclosure commitments    51  Table 2-7: Suspicious Volumes and ICO Outcomes (H3)    raised_ dummy success log_ raised listing web_ active products Intercept  -1.59***   -1.62**     3.14      -2.04**    -2.15***   -3.00***    (-2.67)    (-2.45)    (1.61)     (-2.40)    (-3.16)    (-3.45)    suspicious_volume - -0.73**    -0.92***   -2.46**    -1.01**    -0.61*     -0.84*     (-2.25)    (-2.59)    (-2.19)    (-2.40)    (-1.86)    (-1.73)    early_token_transfers +  0.14***    0.16***    0.50***    0.22***    0.17***    0.14***    (5.58)     (6.22)     (6.18)     (7.64)     (6.51)     (4.83)     log_team +  0.38***    0.36***    1.36***    0.24**     0.45***    0.32***    (4.09)     (3.60)     (4.79)     (2.06)     (4.58)     (2.71)     social +  1.56***    1.50***    5.11***    1.67***    1.31***    2.12***    (4.53)     (3.93)     (4.75)     (3.46)     (3.81)     (3.99)     pre_sale ? -0.15      -0.26**    -0.54      -0.08      -0.11      -0.18       (-1.18)    (-1.99)    (-1.29)    (-0.54)    (-0.89)    (-1.11)    log_ico_len - -0.32***   -0.40***   -1.26***   -0.62***   -0.30***   -0.40***    (-4.74)    (-5.82)    (-6.00)    (-7.74)    (-4.43)    (-5.17)    soft_req ? -0.11      -0.28*     -0.47      -0.62***   -0.22      -0.40**     (-0.75)    (-1.81)    (-0.94)    (-3.71)    (-1.48)    (-2.20)    hard_req +  0.74***    0.59***    2.30***    0.49**     0.37**     0.04       (3.82)     (2.96)     (3.72)     (2.15)     (2.01)     (0.18)     github + -0.11      -0.06      -0.41       0.13       0.06       0.26       (-0.78)    (-0.45)    (-0.90)    (0.77)     (0.40)     (1.43)     whitepaper + -0.14      -0.04      -0.63       0.48       0.78       0.13        (-0.31)    (-0.07)    (-0.39)    (0.65)     (1.37)     (0.18)     restrict_us - -0.20       0.06      -0.59      -0.17      -0.13      -0.18       (-1.54)    (0.45)     (-1.33)    (-1.11)    (-0.95)    (-1.08)    num_prior_startups +  0.25       0.32**     1.23**     0.42**     0.85***    0.29       (1.56)     (2.05)     (2.22)     (2.49)     (4.54)     (1.62)     prior_blockchain_exp +  0.25       0.41*      0.78       0.29       0.21      -0.11       (1.07)     (1.76)     (0.98)     (1.06)     (0.86)     (-0.40)    high_edu +  0.29*      0.17       0.98*      0.39**     0.45***    0.43**     (1.91)     (1.16)     (1.88)     (2.25)     (2.83)     (2.41)     N   1282  1282  1282  1282  1282  1282 pseudo-R2/adj_R2    0.09  0.11  0.13  0.16  0.12  0.11 ***, **, * significant at the 1 percent, 5 percent, and 10 percent levels, respectively. T-stats in parentheses.     52  Table 2-8: Robustness Tests Panel A: Amount of raised funds scaled by hard caps        scaled _raised scaled _raised scaled _raised scaled _raised Intercept   0.66***    0.65***    0.82**     0.74       (4.20)     (4.09)     (2.12)     (0.72)     partial_commitment +  0.16***    0.13***      (7.06)     (5.06)       full_commitment +   0.10***       (2.83)       early_token_transfers +    0.01       0.04***      (0.70)     (3.83)     log_team +  0.04**     0.04**     0.03       0.06       (2.07)     (1.96)     (0.70)     (1.16)     social +  0.02       0.02       0.05      -0.07       (0.33)     (0.27)     (0.36)     (-0.34)    pre_sale ? -0.10***   -0.10***   -0.09**    -0.07       (-4.39)    (-4.22)    (-2.44)    (-1.24)    log_ico_len - -0.13***   -0.12***   -0.11***   -0.16***    (-9.21)    (-9.06)    (-4.58)    (-6.76)    soft_req ? -0.08***   -0.08***   -0.11*     -0.13*      (-2.86)    (-2.70)    (-1.87)    (-1.82)    github +  0.04*      0.04       0.02       0.06       (1.72)     (1.52)     (0.58)     (0.95)     whitepaper + -0.08      -0.08      -0.12       0.08        (-0.58)    (-0.59)    (-0.37)    (0.08)     restrict_us - -0.05**    -0.05**    -0.08**    -0.08       (-2.46)    (-2.37)    (-2.20)    (-1.33)    num_prior_startups +  0.05       0.05       0.11**     0.00       (1.56)     (1.54)     (2.23)     (0.00)     prior_blockchain_exp + -0.05      -0.06      -0.13*     -0.01       (-1.26)    (-1.42)    (-1.79)    (-0.09)    high_edu +  0.02       0.02       0.00       0.04        (0.74)     (0.90)     (0.07)     (0.65)     N   906  906  378  192 pseudo-R2/adj_R2    0.19  0.20  0.11  0.31      53  Panel B: Alternative definition of blockchain information    raised_ dummy success log_ raised listing web_ active products Intercept  -2.31***   -1.78**     1.01      -2.64***   -2.22***   -3.15***    (-3.00)    (-2.23)    (0.42)     (-2.74)    (-2.91)    (-3.08)    num_transfer_day1 + -0.05      -0.01      -0.19       0.09       0.09*      0.18***    (-0.81)    (-0.17)    (-0.86)    (1.35)     (1.67)     (2.69)     log_team +  0.37***    0.33***    1.29***    0.27*      0.45***    0.31*      (3.20)     (2.61)     (3.53)     (1.75)     (3.68)     (1.92)     social +  1.89***    1.90***    6.24***    1.52***    1.05**     1.52**     (4.46)     (4.11)     (4.69)     (2.67)     (2.54)     (2.33)     pre_sale ? -0.31**    -0.35**    -1.02*     -0.03      -0.09      -0.06       (-1.97)    (-2.15)    (-1.90)    (-0.14)    (-0.59)    (-0.26)    log_ico_len - -0.17**    -0.26***   -0.75**    -0.45***   -0.25***   -0.34***    (-2.04)    (-3.03)    (-2.55)    (-4.71)    (-2.84)    (-3.21)    soft_req ? -0.22      -0.44**    -0.87      -0.75***   -0.41**    -0.71***    (-1.11)    (-2.15)    (-1.28)    (-3.44)    (-2.07)    (-2.90)    hard_req +  0.75***    0.70***    2.50***    0.56*      0.69***    0.51       (3.06)     (2.71)     (3.19)     (1.91)     (2.94)     (1.54)     github + -0.13      -0.16      -0.50       0.20       0.05       0.31       (-0.79)    (-0.91)    (-0.86)    (0.93)     (0.30)     (1.27)     whitepaper +  0.12      -0.33      -0.09       0.88       0.93       0.38        (0.20)     (-0.52)    (-0.05)    (1.06)     (1.56)     (0.43)     restrict_us - -0.11       0.09      -0.39      -0.25      -0.14      -0.11       (-0.70)    (0.53)     (-0.67)    (-1.24)    (-0.86)    (-0.52)    num_prior_startups +  0.36*      0.44**     1.70**     0.57***    0.77***    0.35       (1.79)     (2.27)     (2.29)     (2.65)     (3.21)     (1.53)     prior_blockchain_exp +  0.00       0.10      -0.08       0.39       0.05      -0.14       (0.00)     (0.33)     (-0.07)    (1.18)     (0.16)     (-0.37)    high_edu +  0.20       0.10       0.68       0.17       0.49**     0.20       (1.08)     (0.53)     (0.99)     (0.76)     (2.50)     (0.81)     N   807  807  807  807  807  807 pseudo-R2/adj_R2    0.07  0.07  0.08  0.10  0.09  0.08 ***, **, * significant at the 1 percent, 5 percent, and 10 percent levels, respectively. T-stats in parentheses.      54  Panel C: Alternative definition of blockchain information (Continued)    raised_ dummy success log_ raised listing web_ active products Intercept   0.06      -1.36       8.05      -1.02      -2.03      -2.73       (0.05)     (-0.90)    (1.63)     (-0.65)    (-1.36)    (-1.54)    num_transfer_day1 +  0.18***    0.17***    0.55***    0.24***    0.22***    0.09*      (3.68)     (3.80)     (3.82)     (4.95)     (4.10)     (1.76)     log_team +  0.49***    0.50***    1.72***    0.31*      0.49***    0.43**     (3.18)     (2.97)     (3.67)     (1.80)     (2.98)     (2.38)     social +  1.02*      0.89       3.52*      1.97**     1.83***    2.98***    (1.75)     (1.43)     (1.83)     (2.44)     (3.12)     (3.32)     pre_sale ?  0.10      -0.17       0.12      -0.29      -0.21      -0.37       (0.46)     (-0.83)    (0.18)     (-1.27)    (-0.99)    (-1.57)    log_ico_len - -0.42***   -0.45***   -1.48***   -0.59***   -0.15      -0.37***    (-3.51)    (-3.66)    (-4.36)    (-4.15)    (-1.27)    (-2.90)    soft_req ?  0.11      -0.03       0.18      -0.43*      0.07      -0.03       (0.49)     (-0.13)    (0.24)     (-1.74)    (0.29)     (-0.11)    hard_req +  0.53*      0.23       1.44       0.18      -0.38      -0.60       (1.66)     (0.73)     (1.30)     (0.51)     (-1.21)    (-1.61)    github + -0.04       0.14      -0.22       0.07       0.15       0.21       (-0.16)    (0.59)     (-0.27)    (0.26)     (0.59)     (0.78)     whitepaper + -1.05       0.33      -3.35      -0.58       0.19      -0.46        (-1.01)    (0.24)     (-0.75)    (-0.43)    (0.14)     (-0.30)    restrict_us - -0.48**    -0.15      -1.32*     -0.24      -0.25      -0.35       (-2.17)    (-0.67)    (-1.83)    (-0.94)    (-1.08)    (-1.34)    num_prior_startups +  0.05       0.11       0.61       0.31       0.97***    0.25       (0.19)     (0.44)     (0.71)     (1.21)     (3.18)     (0.93)     prior_blockchain_exp +  0.65       0.81**     1.84       0.03       0.42      -0.21       (1.64)     (2.13)     (1.60)     (0.07)     (0.94)     (-0.50)    high_edu +  0.40       0.25       1.30       0.61**     0.31       0.68***    (1.56)     (1.00)     (1.59)     (2.24)     (1.15)     (2.58)     N   475  475  475  475  475  475 pseudo-R2/adj_R2    0.12  0.12  0.15  0.19  0.15  0.12 ***, **, * significant at the 1 percent, 5 percent, and 10 percent levels, respectively. T-stats in parentheses.     55  Panel D: Alternative definition of suspicious volumes    raised_ dummy success log_ raised listing web_ active products Intercept  -1.53**    -1.56**     3.34*     -1.94**    -2.10***   -2.93***    (-2.55)    (-2.34)    (1.70)     (-2.28)    (-3.06)    (-3.35)    suspicious_10% - -0.46*     -0.40      -1.61**    -0.72***   -0.39      -0.68**    (-1.94)    (-1.64)    (-1.99)    (-2.59)    (-1.64)    (-2.03)    early_token_transfers +  0.14***    0.15***    0.51***    0.22***    0.17***    0.15***    (5.54)     (5.98)     (6.14)     (7.71)     (6.44)     (4.90)     log_team +  0.38***    0.36***    1.37***    0.24**     0.45***    0.32***    (4.11)     (3.65)     (4.81)     (2.07)     (4.59)     (2.73)     social +  1.58***    1.53***    5.18***    1.70***    1.32***    2.14***    (4.59)     (4.01)     (4.81)     (3.52)     (3.87)     (4.02)     pre_sale ? -0.15      -0.26**    -0.53      -0.08      -0.11      -0.18       (-1.18)    (-2.03)    (-1.28)    (-0.54)    (-0.88)    (-1.10)    log_ico_len - -0.32***   -0.40***   -1.28***   -0.63***   -0.30***   -0.41***    (-4.81)    (-5.89)    (-6.08)    (-7.81)    (-4.48)    (-5.28)    soft_req ? -0.12      -0.28*     -0.48      -0.62***   -0.23      -0.40**     (-0.78)    (-1.84)    (-0.96)    (-3.74)    (-1.50)    (-2.21)    hard_req +  0.74***    0.58***    2.31***    0.49**     0.37**     0.04       (3.83)     (2.96)     (3.73)     (2.14)     (1.99)     (0.18)     github + -0.10      -0.06      -0.38       0.14       0.06       0.27       (-0.71)    (-0.40)    (-0.83)    (0.84)     (0.47)     (1.49)     whitepaper  -0.20      -0.10      -0.80       0.39       0.74       0.07        (-0.42)    (-0.18)    (-0.49)    (0.54)     (1.28)     (0.10)     restrict_us - -0.20       0.06      -0.58      -0.16      -0.12      -0.16       (-1.52)    (0.45)     (-1.30)    (-1.04)    (-0.94)    (-0.99)    num_prior_startups +  0.24       0.31**     1.19**     0.40**     0.84***    0.27       (1.49)     (2.00)     (2.14)     (2.36)     (4.48)     (1.49)     prior_blockchain_exp +  0.24       0.39*      0.73       0.28       0.20      -0.13       (1.02)     (1.70)     (0.92)     (1.01)     (0.80)     (-0.46)    high_edu +  0.28*      0.16       0.95*      0.37**     0.45***    0.41**     (1.84)     (1.06)     (1.81)     (2.14)     (2.77)     (2.32)     N   1282  1282  1282  1282  1282  1282 pseudo-R2/adj_R2    0.09  0.10  0.13  0.16  0.12  0.11 ***, **, * significant at the 1 percent, 5 percent, and 10 percent levels, respectively. T-stats in parentheses.     56  Panel E: Predicting ICO outcomes with analyst rating    raised_ dummy success log_ raised listing web_ active products Intercept  -3.03***   -2.97***   -0.56      -4.07***   -3.11***   -4.72***    (-6.19)    (-5.43)    (-0.44)    (-5.07)    (-5.69)    (-6.13)    partial_commitment +  0.62***    0.63***    2.00***    1.45***    0.52***    0.76***   (5.38)     (5.00)     (5.81)     (7.49)     (4.66)     (4.17)     full_commitment +  0.38***    0.34***    1.35***    0.30**     0.32**     0.38**     (3.06)     (2.68)     (3.21)     (2.08)     (2.50)     (2.40)     rating +  3.87***    4.63***    13.68***   4.64***    5.42***    5.45***     (7.23)     (8.03)     (8.33)     (6.43)     (9.95)     (7.25)     log_team +  0.26***    0.20**     0.63***    0.07       0.19**     0.14       (3.12)     (2.18)     (2.80)     (0.57)     (2.24)     (1.18)     social +  0.23      -0.02       0.07       0.36      -0.38       0.34       (0.78)     (-0.06)    (0.09)     (0.77)     (-1.31)    (0.72)     pre_sale ? -0.31***   -0.42***   -1.01***   -0.25*     -0.30***   -0.42***    (-3.04)    (-3.91)    (-3.26)    (-1.86)    (-2.94)    (-2.96)    log_ico_len - -0.22***   -0.29***   -0.89***   -0.57***   -0.19***   -0.31***    (-3.65)    (-4.61)    (-4.87)    (-7.41)    (-3.11)    (-4.14)    soft_req ? -0.03      -0.25**    -0.22      -0.66***   -0.24**    -0.36**     (-0.25)    (-1.98)    (-0.58)    (-4.33)    (-2.07)    (-2.20)    hard_req +  0.53***    0.39**     1.42***    0.42**     0.12      -0.15       (3.45)     (2.37)     (3.27)     (2.02)     (0.81)     (-0.68)    github + -0.29***   -0.25**    -0.95***   -0.11      -0.21*     -0.13       (-2.60)    (-2.05)    (-2.76)    (-0.73)    (-1.91)    (-0.78)    whitepaper + -0.32      -0.49      -1.53       0.13       0.06      -0.11        (-0.84)    (-1.11)    (-1.59)    (0.19)     (0.12)     (-0.17)    restrict_us - -0.15      -0.07      -0.44      -0.34**    -0.06      -0.29*      (-1.42)    (-0.59)    (-1.31)    (-2.40)    (-0.58)    (-1.95)    num_prior_startups +  0.15       0.22*      0.77*      0.25       0.66***    0.01       (1.15)     (1.68)     (1.74)     (1.64)     (4.53)     (0.06)     prior_blockchain_exp +  0.13       0.31       0.47       0.31       0.03      -0.17       (0.66)     (1.53)     (0.68)     (1.28)     (0.14)     (-0.63)    high_edu +  0.16       0.02       0.48       0.21       0.19       0.17       (1.28)     (0.20)     (1.17)     (1.31)     (1.49)     (1.04)     N   2085  2085  2085  2085  2085  2085 pseudo-R2/adj_R2    0.12  0.13  0.17  0.21  0.14  0.14 ***, **, * significant at the 1 percent, 5 percent, and 10 percent levels, respectively. T-stats in parentheses.   57  Chapter 3: A Model for Disclosure Commitments 3.1 Introduction Accounting disclosure is critical for the functioning of an efficient capital market. One prominent argument is that a firm’s commitment to increased levels of disclosure should reduce information asymmetries between managers and shareholders and decrease the costs of capital. How can firms make credible disclosure commitments in practice? Previous studies focus on the following two mechanisms.  On the one hand, mandatory disclosure provides a credible commitment mechanism. Firms choose accounting standards, e.g., GAAP or IFRS, which provide specific requirements on the disclosure format and quantity. Firms are also subject to disclosure regulations, e.g., SEC regulations, and it can be costly to violate them. On the other hand, firms make implicit commitments by following voluntary disclosure policies consistently. Such commitments can be enforced by the market. For example, firms can provide earnings forecasts on a regular basis, and Chen, Matsumoto, and Rajgopal (2011) document negative market reactions when firms discontinue this practice. Blockchain technology provides a new way to make commitments. Conceptually, a blockchain is a system of open and distributed ledgers that record information in a verifiable and permanent way. One can make a commitment by writing a computer program that specifies some future actions, e.g., sending a message at a future date. After the program is deployed on blockchains, it becomes immutable and is enforced by the blockchain protocols. As long as the protocols are transparent and well designed, the commitment is credible.  58  To illustrate the commitment function of blockchains, I briefly describe its role in the market for initial coin offerings (ICOs). ICOs became a popular way to finance startups in 2016. In an ICO, investors invest with a well-established cryptocurrency, such as Bitcoins, and in return they receive a new cryptocurrency, a new coin. The new coin may have values or entitle its holders to future services. This process is governed by a computer program, called the ‘smart contract.’ A smart contract works the same way as a vending machine. It can perform certain transactions automatically, such as collecting Bitcoins investments and distributing new coins. The ‘vending machine’ also has a disclosure function: it can record the transactions on blockchains and disclose them to the public. The disclosure decision, e.g., whether to disclose inflows of Bitcoins or outflows of new coins to their investors, is made by the entrepreneurs who write their own smart contracts. Importantly, they write and deploy smart contracts before actual transactions take place. That is, the disclosure decision is made before the entrepreneurs know the outcome of the fundraising, i.e., they make a disclosure commitment. If they commit to disclosing transactions on blockchains, such disclosure decisions become irreversible. Meanwhile, entrepreneurs have the option not to commit but rather to make an ex-post disclosure decision, i.e., to defer the disclosure decisions until after transactions take place. Blockchains differ from traditional commitment mechanisms in the following ways. First, commitments on blockchains are enforced by computer programs, which, if well-designed, prevent potential violations. For example, one can write a program that will send out a message at a future date without a ‘cancel’ function. Once the program is deployed on a blockchain, no  59  one can stop it from sending out the message. In contrast, commitments through regulations are enforced by a third party who punish actual violations. For example, the SEC can charge a firm with misreporting in court. The credibility of commitments through regulations depends on the effectiveness of enforcement. In that sense, commitments on blockchains are more credible. Second, making commitments is less costly on blockchains than in capital markets. Entrepreneurs incur costs to design and test smart contracts and costs to deploy smart contracts and perform transactions on blockchains. In a typical ICO, developers can save on development costs by simply copying the ERC-20 standard contract. As an illustration of transaction costs, a transfer of 40,000 dollars’ worth of Ethers could cost only three cents on blockchains.  More importantly, blockchain offers more flexibility than traditional commitment mechanisms. When firms go public and subject themselves to disclosure requirements and enforcement, they choose a regulation ‘combo,’ a set of predetermined disclosure requirements such as the US GAAP or IFRS. Because regulators have their own concerns regarding costs and benefits, they often devise a single set of rules governing many firms. However, when firms make commitments on blockchains, they can customize a set of disclosure requirements that best suit them. In other words, they can choose every single ‘dish’ à la carte because they have full control over the items to be disclosed. The distinctions between blockchains and traditional commitment mechanisms, particularly the flexibility that blockchains offer, call for an understanding of the endogenous commitment decisions and the tradeoff between ex-ante and ex-post disclosures. Prior studies often assume that firms will commit to disclosing some signals (Diamond and Verrecchia 1991) or study  60  mandatory disclosure mechanisms imposed by regulators (Stocken 2000; Lundholm 2003). These analyses do not specify the costs of disclosure commitments and do not explain why some firms do not commit to the highest possible level of disclosures. A few auditing papers endogenize levels of disclosure commitments as levels of audit quality (Titman and Trueman 1986; Datar, Feltham, and Hughes 1991). These studies usually remain silent on the tradeoff between disclosure commitments and ex-post voluntary disclosures. In this chapter, I develop a model where commitment decisions are endogenous. In this model, a manager has two value-relevant (statistically dependent) signals: the first one can not be disclosed credibly, and the second one can be disclosed at a cost. After the manager observes the first signal, and before he observes the second one, he can commit to disclosing the second signal, using blockchains. Meanwhile, the manager has the option not to make the disclosure commitment, i.e., to defer the disclosure decision until after he observes the realization of the second signal. This set-up captures the essential features of the ICO market. The first signal corresponds to managers’ private information about project qualities. Because the ICO market lacks regulations, firms’ disclosures are not credible and often contain false claims. The second signal corresponds to the investment transactions that take place on blockchains. The investments collectively represent ‘the wisdom of the crowd’ and can be disclosed credibly. Recall that in Chapter Two, I provide empirical evidence that these transactions are informative of ICO outcomes. Before managers observe investment transactions, they develop their smart contracts, through which they can make credible commitments to disclosing those transactions as they occur.  61  This set-up also has general implications for the capital markets. The first signal corresponds to ‘soft’ information, such as forward-looking statements, valuations of intangible assets, or anything that can not be faithfully represented in financial statements. The second signal corresponds to ‘hard’ information, such as inventory counts, which can easily be verified by a third party. While it is generally difficult to convey soft information credibly, firms can commit to disclosing hard information through accounting systems. I show that there exists an equilibrium where committing to disclosing the second signal reveals the first signal, the signal that otherwise can not be credibly disclosed. Specifically, the manager with a good first signal wants to commit to disclosing the second signal, while the manager with a bad first signal does not. It is intuitive that, even though the first signal cannot be credibly disclosed on its own, it updates the manager’s beliefs about the distribution of the second signal. Consequently, depending on the realizations of the first signal, managers’ incentives to commit to disclosing the second signal differ, which makes the commitment decision informative.  This chapter proceeds as follows. Section 3.2 provides a literature review on disclosure commitments. In Section 3.3, I provide a formal description of the model and a detailed discussion about its link to the ICO market. In Section 3.4 and 3.5, I define the equilibrium and identify conditions where a unique equilibrium exists. Section 3.6 presents some simulation results. Section 3.7 concludes.  62  3.2 Literature Review  In this section, I provide a review of the literature related to disclosure commitments. A disclosure commitment is a decision a firm makes about what it will disclose before it knows the content of the information (i.e., ex-ante). One stream of disclosure literature argues that a commitment a firm makes to increased levels of disclosure decreases the costs of capital.  This literature models accounting information as a noisy signal of cash flows, and models the level of disclosure as the precision of the accounting signal. This literature assumes that the agents in those models will disclose the signals, regardless of their realizations. The intuition is that increased levels of disclosures, namely more precise signals, will reduce the uncertainty about terminal cash flows and decrease market risk discount. A few empirical papers provide evidence in support of the above argument. They measure disclosure commitment using mandatory disclosure regulations and examine the capital market consequences of exogenous changes in disclosure regulations, controlling for the fact that firms self-select into different regulations. For example, Leuz and Verrecchia (2000) focus on firms that switched from German GAAP to US GAAP. They argue that the disclosure levels in German are relatively low and document that firms’ costs of capital decrease after the switch. Cheng et al. (2013) study firms that maintained their disclosure levels after deregulations in segment reporting, and document that their cost of capital increases after the deregulation. The limitation of this literature is that it treats disclosure commitments as exogenous. For example, in discussing Leuz and Verrecchia (2000), Joos (2000) points out that their research design allows assessment on the benefit of the decision to switch reporting standards, but not on the costs.  63  Therefore, the paper fails to explain why only some firms choose to commit to a higher level of disclosure, but not all of them. Another stream of disclosure literature studies disclosure commitments in the context of interactions between mandatory disclosure and voluntary disclosures. The disclosure commitments correspond to the mandatory disclosures. A common argument is that committing to mandatory disclosure requirements enhances the credibility of voluntary disclosures. For example, Stoken (2000) and Lundholm (2003) both establish that mandatory disclosure serves as a verification tool, which reduces managers’ incentive to lie about their private signals. There are also studies on how mandatory disclosures affect firms’ voluntary disclosures (Einhorn 2005). The limitation of this literature is again that it treats disclosure commitments as exogenous. The mandatory disclosure mechanisms are often exogenously imposed on firms. The literature also focuses on how voluntary disclosure strategies change in response to exogenous changes in mandatory disclosure regulations.  To some extent, it is reasonable to treat disclosure commitments as exogenous because firms have limited choices in terms of the levels of disclosure commitments in the capital market. Mandatory disclosure regulations are set by regulators and are generally beyond the control of individual firms. However, blockchains allow firms to customize a set of disclosure requirements. This additional feature calls for an understanding of firms’ endogenous commitment decisions. Several papers in auditing endogenize the disclosure commitments in the context of audit quality. In their models, committing to a higher level of disclosures means choosing an auditor that  64  provides a more accurate financial report. Titman and Trueman (1986) argue that an entrepreneur with more favorable private information about his firm’s value will choose a higher-quality auditor than will an entrepreneur with less favorable private information. Datar, Feltham, and Hughes (1991) extend the model by making the choice of audit quality partially revealing. In their model, the auditor choice and the retained ownership jointly signal for project qualities. The limitation of this literature is that they do not consider the tradeoffs between ex-ante disclosure commitment and ex-post disclosure, i.e., firms can voluntarily disclose the information even if they did not make commitment ex-ante. One exception is the confirmation hypothesis proposed by Ball, Jayaraman, and Shivakumar (2012). They argue that mandatory disclosures and voluntary disclosures are complements. They use audit fees to proxy for levels of verifications, i.e., commitments, and document positive relations between audit fees and management forecast. Our understanding of how firms jointly determine their ex-ante disclosure commitment and ex-post disclosure is limited. Some papers argue that making disclosure commitment is costly. According to a survey in Graham, Harvey, and Rajgopal (2005), two-thirds of the top executives avoid setting a disclosure precedent, because it will be difficult to maintain in the future. Einhorn and Amir (2008) argue that by voluntarily disclosing private information, firms make an implicit commitment to provide similar disclosures in the future. Because disclosure commitments are costly, firms are less willing to disclose information in the first place. In the following sections, I build on these arguments and develop a model in which a firm determines its ex-ante disclosure commitments and ex-post voluntary disclosure decisions jointly. The commitment costs are used  65  to establish a separating equilibrium such that firms of different types differ in their commitment decisions. 3.3 The Model 3.3.1 Players and Information An entrepreneur or manager has a project that requires K amount of financing. The project has two possible outcomes: a gross value of 1 (“success”) or an outcome of 0 (“failure”). Let θ denote the probability of success. For simplicity, I assume that the common prior of θ is 0.5. The manager is going to raise capital from two groups of investors: “informed investors” and “uninformed investors.” Informed investors, who have a limited amount of capital A<K, have better information about the investment than uninformed investors. The manager also has private information about the investment. Specifically, the manager receives a private signal about his type (T), T ∈ {Low,High}, such that P(T = High|Success) = P(T = Low|Failure) = a > 0.5. The informed investors receive a different signal (S), S ∈ {Bad, Good}, such that P(S = Good|Success) = P(T = Bad|Failure) = b > 0.5. 3.3.2 Timeline The game includes the following steps: 1. The manager learns about his type (T). 2. The manager decides whether to commit to disclosing the transactions with the informed investors to the uninformed investors before the transactions occur.  66  3. The informed investors receive the signal (S) and provide the amount A at a price, P(S). The manager can learn about the signal (S) from the transaction price. 4. In the event that the manager did not make a disclosure commitment at step 2, he decides whether to disclose the transactions with the informed investors to the uninformed investors. Then, if the commitment to disclose was made ex-ante, or the decision to disclose was made ex-post, disclosure takes place, and disclosure costs are incurred. 5. The uninformed investors observe the manager’s commitment and disclosure decisions and provide the amount (K-A) at a price conditional on all available information. 3.3.3 Objectives and Payoffs The manager’s objective is to maximize wealth, W. Let P be the share price of the firm, a share being one hundred percent ownership of the firm. Moreover, let λ be the share of equity that the manager must issue to investors to compensate investors for their investment. The indexes i and u represent informed and uninformed investors. Specifically, the manager wants to maximize his final wealth, which is the value of his equity:  W = (1 − λi − λu) ∗ Pu (3.1) I assume that investors are risk-neutral and have a zero interest rate. Each class of investors is willing to provide the of capital at a price for which the expected return on their investment equals the amount of capital they provide. The prices will depend on their information sets.  A = λi ∗ Pi (3.2)  67  K − A = λu ∗ Pu (3.3) Equation (3.1), (3.2) and (3.3) yield  W = (1 −APi) ∗ Pu − (K − A) (3.4) Hence, in order to maximize wealth, the manager maximizes the prices, 𝑃𝑖and 𝑃𝑢, of shares sold to each class of investors, which minimizes the dilution of equity. I assume that the manager can make the disclosure commitment at no cost, but he incurs a disclosure cost, c, when the transaction information is disclosed to the uninformed investors. Furthermore, I assume that the informed investors will form expectations of θ only based on the signal they receive, i.e., Pi = E(θ|S). In contrast, the uninformed investors will set the price based on all available information, i.e., Pu = E(θ|Commitment, Disclosure, Signal). 3.3.4 Some Discussions about the Model This model captures the following essential features of the market for the initial coin offerings (ICOs) and the functions of blockchains and smart contracts. First, the assumption that the manager can not disclose his type credibly captures the lack of regulations in the ICO market, in the sense that disclosure of information was not mandatory in the way that it would be for conventional securities. In the early days of the cryptocurrencies, there were debates on whether the coins are securities and whether they should be subject to security laws. While most startups in the market provide information in ‘whitepapers,’ these documents are unregulated and often contain false claims.  68  Second, the timeline of the game reflects actual ICO processes and highlights the immutability of blockchains. A typical ICO starts with a marketing campaign before actual sales of coins. This marketing campaign includes communicating with potential investors on social media and providing relevant materials, such as ‘whitepapers’ and smart contracts. A smart contract is a computer program that will be used in future token sales, and it determines whether future transactions will be disclosed. Once a program is deployed on the blockchains, it becomes immutable. Consequently, if a startup commits to disclosing future transactions, they will not be able to hide them. However, if a startup does not make such commitments, they can choose to disclose the transactions to investors ex-post. Third, the assumption that the manager can make commitments at no costs, but may incur a disclosure cost is a faithful representation of actual costs of smart contracts. On the one hand, the costs to develop smart contracts and to deploy them on blockchains, i.e., to make commitments, is trivial. Specifically, it is easy to achieve the disclosure function using smart contracts, e.g., to implement the ERC-20 standard. On the other hand, the disclosure of token sales transactions, particularly the cash (‘Ether’) inflows, is costly. Like any other computer program, a smart contract may contain bugs and can be hacked, and there were cases where hackers stole raised funds. In addition, a smart contract can not handle the cases where investors send their money to wrong addresses. These undesirable features are legitimate reasons why startups do not want to use smart contracts. Finally, the assumption that informed investors will form expectations only based on the signal they receive reflects their actual behaviors. In practice, blockchains experts can assess the  69  feasibility of a project, although not managers’ hidden actions, independently based on their expertise. Some sophisticated investors may have private access to management such that smart contracts are less important signals. One might think that the level of sophistication in smart contracts proxies for startups’ ability to use blockchains. However, this is less of a concern because it is easy to achieve the disclosure functions mentioned above. Note that this assumption implies that the price of shares sold to the informed investors is determined only by their signal (S). In other words, the manager can only maximize his wealth by affecting the price of shares sold to uninformed investors. 3.3.5 A Simplified Case: Commitments are Impossible To understand the economics of blockchains and to highlight the importance of commitments, I first consider a simplified case when commitments are impossible. If the manager can not credibly disclose the information about his type, the signal (T) becomes irrelevant to investors and the model reduces to the standard disclosure model with regard to the signal (S). If the manager observes a good signal (S) from the informed investors, he will disclose it to the uninformed investors if the benefits exceed the cost, and withhold it otherwise. For b-c>1-b, the manager’s equilibrium strategy and payoffs are as follows.   If Signal=Good, Disclose, 𝑃𝑢(𝑠 = 𝐺𝑜𝑜𝑑) − 𝑐 = 𝑏 − c If Signal=Bad, Withhold, 𝑃𝑢(𝑠 = 𝐵𝑎𝑑) = 1 − 𝑏  This is a simplified version of the standard disclosure model. In the standard model, the bad firms will be pooled together, and the market will not be able to distinguish them. In this setting,  70  there is only one type of bad firm, so nondisclosure is also perfectly revealing. What is preserved in the model is that withholding a bad signal dominates disclosing it. Specifically, if the manager discloses a bad signal and incurs the disclosure cost, he will be worse than if he withholds it, e.g., 1-b-c<1-b. As will be established in the following section, this helps to make the disclosure commitment a credible and informative signal of the manager’s type. 3.4 Equilibrium Definition I look for perfect Bayesian equilibrium. In equilibrium, the manager chooses the commitment and disclosure strategy in light of its rationally anticipated impact on the price of shares sold to uninformed investors, which, in turn, is determined by the investors’ rational expectations about the manager’s commitment and disclosure strategies. I represent the manager’s equilibrium commitment and disclosure strategy by the function CD ∶T ∗ S  →  A, where CD(t, s) ∈ A is the manager’s commitment and disclosure choices given his type t ∈ T and the signal s ∈ S. Note that A =  { (c, d) , (nc, d) , (nc, nd)}, where c and nc represent commitments and non-commitment, and d and nd represent disclosure and nondisclosure. This implies that the managers can not withhold information if he commits to disclosing it. The equilibrium pricing rule is represented by the function P ∶ A ∗ S →  ℝ, where P(cd, s) ∈ ℝ is the price of shares sold to the uninformed investors, given that cd ∈ A is the manager’s commitment and disclosure decisions, and 𝑠 ∈ 𝑆 is the realization of the signal from informed investors if disclosed.  71  I use CD̂ ∶ T ∗ S  →  A and P̂ ∶ A ∗ S →  ℝ to represent the market expectations about the manager’s commitment and disclosure strategy CD and the manager’s expectations about the market pricing rule P, respectively. The equilibrium is formally defined as the vector (CD, CD̂ ∶ T ∗ S  →  A, P, P̂ ∶ A ∗ S →  ℝ), which satisfies the following three conditions. The first equilibrium condition pertains to the manager’s strategy, requiring that CD(t, s) ∈  argmax𝑐𝑑∈𝐴 P̂(cd, s), for any t ∈ T and s ∈ S. That is, the manager’s commitment and disclosure strategy always maximize the price of shares sold to uninformed investors.   The second equilibrium condition describes the market pricing rule, P(cd, s) = E(𝜃|cd, s) for any 𝑐𝑑 ∈ A and s ∈ S. That is, the uninformed investors are risk-neutral and have zero interest rate. They set the price to the expectation of θ conditional on all available information, including the manager’s commitment and disclosure decisions and the signal from the informed investors if disclosed. Note that the market interprets the manager’s behavior based on its expectation of the manager’s strategy, 𝐶?̂?. In addition, the price is reduced by the disclosure cost, c, if applicable. Lastly, the third equilibrium condition imposes P(cd, s) = P̂(cd, s) and CD(t, s) = CD̂(t, s) for any t ∈ T, s ∈ S, and 𝑐𝑑 ∈ A, implying that both the manager and the investors have rational expectations regarding each other’s behaviors.  72  3.5 Equilibrium Analysis I now turn to deriving the equilibrium and analyzing its properties. I am particularly interested in a separating equilibrium where the manager’s commitment decision fully reveals the private information about his type. Specifically, I conjecture that there exists an equilibrium such that high type firms want to make disclosure commitment, but low type firms do not want to commit. Instead, they want to make the disclosure decision after observing the signal. The uninformed investors correctly infer the manager’s type from their commitment decisions and set their price at the expectations conditional on all available information minus disclosure costs if applicable.  𝑃𝑢 ={    𝐸(𝜃|𝑐𝑑 = (𝑐, 𝑑), 𝑠 = 𝐺𝑜𝑜𝑑) − 𝑐, 𝑖𝑓 𝑐𝑜𝑚𝑚𝑖𝑡 𝑎𝑛𝑑 𝑑𝑖𝑠𝑐𝑙𝑜𝑠𝑒 𝑠 = 𝐺𝑜𝑜𝑑𝐸(𝜃|𝑐𝑑 = (𝑐, 𝑑), 𝑠 = 𝐵𝑎𝑑) − 𝑐, 𝑖𝑓 𝑐𝑜𝑚𝑚𝑖𝑡 𝑎𝑛𝑑 𝑑𝑖𝑠𝑐𝑙𝑜𝑠𝑒 𝑠 = 𝐵𝑎𝑑𝐸(𝜃|𝑐𝑑 = (𝑛𝑐, 𝑑), 𝑠 = 𝐺𝑜𝑜𝑑) − 𝑐, 𝑖𝑓 𝑛𝑜 𝑐𝑜𝑚𝑚𝑖𝑡 𝑎𝑛𝑑 𝑑𝑖𝑠𝑐𝑙𝑜𝑠𝑒 𝑠 = 𝐺𝑜𝑜𝑑𝐸(𝜃|𝑐𝑑 = (𝑛𝑐, 𝑛𝑑)), 𝑖𝑓 𝑛𝑜 𝑐𝑜𝑚𝑚𝑖𝑡 𝑎𝑛𝑑 𝑤𝑖𝑡ℎℎ𝑜𝑙𝑑 𝑠 (3.5) The following function categorizes the manager’s commitment and disclosure strategy.  𝐶𝐷(𝑡, 𝑠) ={    (𝑐, 𝑑), 𝑖𝑓 𝑡 = 𝐻𝑖𝑔ℎ 𝑎𝑛𝑑 𝑠 = 𝐺𝑜𝑜𝑑(𝑐, 𝑑), 𝑖𝑓 𝑡 = 𝐻𝑖𝑔ℎ 𝑎𝑛𝑑 𝑠 = 𝐵𝑎𝑑(𝑛𝑐, 𝑑), 𝑖𝑓 𝑡 = 𝐿𝑜𝑤 𝑎𝑛𝑑 𝑠 = 𝐺𝑜𝑜𝑑(𝑛𝑐, 𝑛𝑑), 𝑖𝑓 𝑡 = 𝐿𝑜𝑤 𝑎𝑛𝑑 𝑠 = 𝐵𝑎𝑑 (3.6) Theorem 1. There exists a separating equilibrium for a, b, and c that satisfy the following conditions, (3.7), (3.8) and (3.9). In equilibrium, the manager’s optimal commitment and disclosure strategy is characterized by equation (3.6).  [𝑏 ∗ 𝑎 + (1 − 𝑏) ∗ (1 − 𝑎)] ∗ [𝑎 ∗ 𝑏𝑎 ∗ 𝑏 + (1 − 𝑎) ∗ (1 − 𝑏)− 𝑐] +[1 − 𝑏 ∗ 𝑎 − (1 − 𝑏) ∗ (1 − 𝑎)] ∗ [𝑎 ∗ (1 − 𝑏)𝑎 ∗ (1 − 𝑏) + 𝑏 ∗ (1 − 𝑎)− 𝑐] > (3.7)  73  [𝑏 ∗ 𝑎 + (1 − 𝑏) ∗ (1 − 𝑎)] ∗ [(1 − 𝑎) ∗ 𝑏(1 − 𝑎) ∗ 𝑏 + 𝑎 ∗ (1 − 𝑏)− 𝑐] +[1 − 𝑏 ∗ 𝑎 − (1 − 𝑏) ∗ (1 − 𝑎)] ∗(1 − 𝑎) ∗ (1 − 𝑏)(1 − 𝑎) ∗ (1 − 𝑏) + 𝑎 ∗ 𝑏   [𝑏 ∗ (1 − 𝑎) + (1 − 𝑏) ∗ 𝑎] ∗ [𝑎 ∗ 𝑏𝑎 ∗ 𝑏 + (1 − 𝑎) ∗ (1 − 𝑏)− 𝑐] +[1 − 𝑏 ∗ (1 − 𝑎) − (1 − 𝑏) ∗ 𝑎] ∗ [𝑎 ∗ (1 − 𝑏)𝑎 ∗ (1 − 𝑏) + 𝑏 ∗ (1 − 𝑎)− 𝑐] < [𝑏 ∗ (1 − 𝑎) + (1 − 𝑏) ∗ 𝑎] ∗ [(1 − 𝑎) ∗ 𝑏(1 − 𝑎) ∗ 𝑏 + 𝑎 ∗ (1 − 𝑏)− 𝑐] +[1 − 𝑏 ∗ (1 − 𝑎) − (1 − 𝑏) ∗ 𝑎] ∗(1 − 𝑎) ∗ (1 − 𝑏)(1 − 𝑎) ∗ (1 − 𝑏) + 𝑎 ∗ 𝑏 (3.8)   (1 − 𝑎) ∗ 𝑏(1 − 𝑎) ∗ 𝑏 + 𝑎 ∗ (1 − 𝑏)− 𝑐 >(1 − 𝑎) ∗ (1 − 𝑏)(1 − 𝑎) ∗ (1 − 𝑏) + 𝑎 ∗ 𝑏  (3.9) Proof. See Appendix B.1.                  □ 3.6 Simulation Results There are three parameters in the model: the informativeness of the manager’s information, a, the informativeness of the informed investors’ information, b, and the disclosure cost, c. Figure 3-1 shows the partitions of (a, b) that satisfy the equilibrium criteria in the red area when the disclosure costs, c=0.4. The black line corresponds to the equation (3.7) and represents the line where the high type managers are indifferent between commitment or no commitment. The blue line corresponds to the equation (3.8) and represents the line where the low type managers are indifferent between commitment or no commitment. The yellow line corresponds to the equation  74  (3.9) and represents the line where the low type managers are indifferent between disclosure or no disclosure. The figure offers several insights. The first observation corresponds to the relevance of the two signals: the equilibrium exists when the manager’s signal and the informed investors’ signal are relatively close in terms of informativeness. This observation confirms the intuition that if one signal is much more informative than the other one, then the disclosure decision with regard to the less informative signal should be less relevant. For example, when the manager’s signal is much more informative than the signal of the informed investors, i.e., a>b as in the lower right corner of the graph, both types of managers want to make disclosure commitment. Therefore, a separating equilibrium is not sustained. The second observation reflects the concept of ‘costly’ signals: all partitions that satisfy the equilibrium conditions are above the line y=x. In other words, to make the commitment a credible signal of the type (T), the signal that the manager commits to disclosing, the signal (S) from the informed investors, must be more informative. A signal must be ‘costly’ before it can become credible. The cost here refers to the situation where the manager has to disclose a bad signal and incur the disclosure costs, which is worse than withholding the bad signal. A disclosure commitment means that the manager is willing to risk himself in the above situation. Only in this way can he credibly convey that the chances that he ends up in the situation are smaller, and the benefits of the signaling can not exceed the potential or expected sufferings in that situation.  75  The third observation corresponds to the ‘single crossing property.’ While both high type and low type managers prefer to make commitments when their information is more informative than the informed investors’, i.e., in the lower right corner. Their ‘indifferent curves’ regarding the commitment decision are different. Such difference stems from the fact that they receive different information about the type (T), which updates their belief about the probability of a good signal (S). Specifically, high type managers expect that good signals are more likely than the low type managers would expect. For a given level of informativeness of the type information, the high type managers benefit more, so the commitment decision costs more to the low type managers. The equilibrium is right between the two indifference curves, consistent with previous signaling arguments. Lastly, the yellow line specifies the partitions in which the managers want to disclose good signals in the disclosure sub-game. The line is mainly determined by the disclosure costs. In a classic disclosure game, the disclosure threshold is determined by the disclosure costs, and the signal will be disclosed only if the benefit exceeds the cost. Figure 3-2 presents equilibrium distributions under different disclosure costs. When the disclosure costs increase as in the following graphs, from c=0.4 to c=0.6, the equilibrium area shrinks. However, the disclosure costs also represent the benefits, cost-saving, for not making disclosure commitment. In that sense, larger cost-saving can also increase the partitions of a, b that satisfy equilibrium conditions, from c=0.2 to c=0.4.  76  3.7 Conclusions I develop a model where disclosure commitment decision is endogenous. A manager can commit to disclosing a signal before it is realized, or he can defer the disclosure decision until after he observes the signal. My analyses demonstrate that the commitment decision can credibly convey information that otherwise could not be disclosed. Meanwhile, some firms prefer to make disclosure decisions after receiving the signal, considering how the market would react to the non-commitment decision. The results suggest that blockchain enhances firms’ ability to communicate private information to the market.     77  Figure 3-1: Illustrations of Equilibrium     78  Figure 3-2: Illustrations of Equilibrium with Different Disclosure Costs    79  Chapter 4: Conclusion This thesis examines the impacts of blockchain technology on accounting practice in two separate chapters. Chapter two establishes that blockchain functions as a disclosure commitment device in the market for Initial Coin Offerings (ICOs). I find that ICOs that make more disclosure commitments with blockchains are more likely to succeed. I also find that transaction volumes disclosed on blockchains predict ICO outcomes and that investors punish ICOs with suspicious volumes. Chapter three provides a theoretical explanation for the ICO market. I develop a model where a manager can commit to disclosing a signal before it is realized, or he can defer the disclosure decision until after he observes the signal. I show that the commitment decision can credibly convey information that otherwise could not be disclosed. Collectively, these studies suggest that blockchain complements traditional commitment mechanisms and enhances firms’ ability to communicate private information to the market. The findings also highlight the information content of computer programs and the importance of developing relevant analytic skills for CPAs in the future. It is worth noting that both the blockchain technology and the industry are rapidly developing, and this thesis focuses on one feature of blockchain. Specifically, the commitment function stems from immutability, which comes from the use of cryptography in the design of blockchain protocols. There are other exciting features, as well as many implementation challenges. I would like to add a few thoughts about them at the end of this thesis.  80  First, I believe that the mainstream adoption of blockchain is inevitable. This conjecture follows the main argument in this thesis and is similar to the full disclosure theorem. When firms have the option to make cheap and credible commitments on blockchains, failure to implement the technology is rationally interpreted as negative news. Second, blockchain has the potential to revolutionize our current financial reporting systems. Particularly, blockchain can stream financial data in real time, like the transactions in the ICO market. I expect that, in the future, more auditing efforts will be on system design and blockchain integration, but less on substantive tests. Third, one current implementation challenge is the privacy concern. In other words, how firms can benefit from the commitment function while maintaining privacy? Recent developments in the zero-knowledge proof have shown promising futures in this area. Finally, blockchain development differs across countries. In western countries, firms are leading the development, e.g., the Nightfall of Ernst & Young, the Quorum of JP Morgan and the Libra of Facebook, and they compete for market dominance. In countries such as China and Singapore, governments are leading the development. Will the roles of blockchain differ in different countries? This thesis represents one attempt to understand and harness blockchain technology.    81  Bibliography Adhami, Saman, Giancarlo Giudici, and Stefano Martinazzi. 2018. “Why Do Businesses Go Crypto? An Empirical Analysis of Initial Coin Offerings.” Journal of Economics and Business 100: 64–75. Bailey, Warren, G. Andrew Karolyi, and Carolina Salva. 2006. “The Economic Consequences of Increased Disclosure: Evidence from International Cross-Listings.” Journal of Financial Economics 81 (1): 175–213. Bakos, Yannis, and Hanna Halaburda. 2018. “The Role of Cryptographic Tokens and ICOs in Fostering Platform Adoption.” SSRN Electronic Journal. Ball, Ray, Sudarshan Jayaraman, and Lakshmanan Shivakumar. 2012. “Audited Financial Reporting and Voluntary Disclosure as Complements: A Test of the Confirmation Hypothesis.” Journal of Accounting and Economics 53 (1–2): 136–66. Barry, Christopher B., Chris J. Muscarella, John W. Peavy, and Michael R. Vetsuypens. 1990. “The Role of Venture Capital in the Creation of Public Companies.” Journal of Financial Economics 27 (2): 447–71. Benedetti, Hugo E, and Leonard Kostovetsky. 2018. “Digital Tulips? Returns to Investors in Initial Coin Offerings.” SSRN Electronic Journal. Bourveau, Thomas, Emmanuel T. De George, Atif Ellahie, and Daniele Macciocchi. 2018. “Initial Coin Offerings: Early Evidence on the Role of Disclosure in the Unregulated Crypto  82  Market.” SSRN Electronic Journal. Carpenter, Charles G, and Robert H. Strawser. 1971. “Displacement of Auditors When Clients Go Public.” Journal of Accountancy 131 (6): 55–59. Chemla, Gilles, and Katrin Tinn. 2019. “Learning Through Crowdfunding.” Management Science, 1–43. Chen, Shuping, Dawn Matsumoto, and Shiva Rajgopal. 2011. “Is Silence Golden? An Empirical Analysis of Firms That Stop Giving Quarterly Earnings Guidance.” Journal of Accounting and Economics 51 (1–2): 134–50. Cheng, Lin, Scott Liao, and Haiwen Zhang. 2013. “The Commitment Effect versus Information Effect of Disclosure-Evidence from Smaller Reporting Companies.” Accounting Review 88 (4): 1239–63. Cohney, Shaanan, David Hoffman, Jeremy Sklaroff, and David Wishnick. 2019. “Coin-Operated Capitalism.” Columbia Law Review 119 (3): 591–676. Cong, Lin, and Yizhou Xiao. 2019. “Information Cascades and Threshold Implementation.” SSRN Electronic Journal. Datar, Srikant M., Gerald A. Feltham, and John S. Hughes. 1991. “The Role of Audits and Audit Quality in Valuing New Issues.” Journal of Accounting and Economics 14 (1): 3–49. Diamond, Douglas W., and Robert E. Verrecchia. 1991. “Disclosure, Liquidity, and the Cost of  83  Capital.” The Journal of Finance 46 (4): 1325. Einhorn, Eti. 2005. “The Nature of the Interaction between Mandatory and Voluntary Disclosures.” Journal of Accounting Research 43 (4): 593–621. Einhorn, Eti, and Amir Ziv. 2008. “Intertemporal Dynamics of Corporate Voluntary Disclosures.” Journal of Accounting Research 46 (3): 567–89. Feng, Chen, Nan Li, M.H. Franco Wong, and Mingyue Zhang. 2019. “Initial Coin Offerings, Blockchain Technology, and White Paper Disclosures.” SSRN Electronic Journal. Fisch, Christian. 2018. “Initial Coin Offerings (ICOs) to Finance New Ventures: An Exploratory Study.” SSRN Electronic Journal 34 (1): 1–22. Florysiak, David, and Alexander Schandlbauer. 2018. “The Information Content of ICO White Papers.” SSRN Electronic Journal. Graham, John R., Campbell R. Harvey, and Shiva Rajgopal. 2005. “The Economic Implications of Corporate Financial Reporting.” Journal of Accounting and Economics 40 (1–3): 3–73. Howell, Sabrina, Marina Niessner, and David Yermack. 2018. “Initial Coin Offerings: Financing Growth with Cryptocurrency Token Sales.” SSRN Electronic Journal. Joos, Peter. 2000. “Discussion of The Economic Consequences of Increased Disclosure.” Journal of Accounting Research 38 (1): 125. Lee, Jongsub, Tao Li, and Donghwa Shin. 2018. “The Wisdom of Crowds and Information  84  Cascades in FinTech: Evidence from Initial Coin Offerings.” SSRN Electronic Journal 1 (352). Leuz, Christian, and Robert E. Verrecchia. 2000. “The Economic Consequences of Increased Disclosure.” Journal of Accounting Research. Vol. 38. Li, Jiasun, and William Mann. 2018. “Initial Coin Offering and Platform Building.” SSRN Electronic Journal. Lundholm, Russell. 2003. “Historical Accounting and the Endogenous Credibility of Current Disclosures.” Journal of Accounting, Auditing & Finance 18 (1): 207–29. Megginson, William L., and Kathleen A. Weiss. 1991. “Venture Capitalist Certification in Initial Public Offerings.” The Journal of Finance 46 (3): 879–903. Michael, Sockin, and Wei Xiong. 2018. “A Model of Cryptocurrencies.” Working Paper Momtaz, Paul P. 2019. “Entrepreneurial Finance and Moral Hazard: Evidence from Token Offerings.” SSRN Electronic Journal, February. Pittman, Jeffrey A., and Steve Fortin. 2004. “Auditor Choice and the Cost of Debt Capital for Newly Public Firms.” Journal of Accounting and Economics 37 (1): 113–36. Stocken, Phillip C. 2000. “Credibility of Voluntary Disclosure.” The RAND Journal of Economics 31 (2): 359. Strausz, Roland. 2017. “A Theory of Crowdfunding: A Mechanism Design Approach with  85  Demand Uncertainty and Moral Hazard.” American Economic Review 107 (6): 1430–76. Titman, Sheridan, and Brett Trueman. 1986. “Information Quality and the Valuation of New Issues.” Journal of Accounting and Economics 8 (2): 159–72.   86  Appendices Appendix A  Appendix for Chapter 2  A.1 Variable Definitions Source Variable Name Definitions ICO Outcomes ICObench and ICOdata raised_dummy A dummy variable that equals one if an ICO raised money. log_raised Natural logarithm of US dollars raised by an ICO, if the information is missing, it is set to zero. success A dummy variable that equals one if an ICO raised more than its soft cap if it has one, or more than 500,000. Various web_active A dummy variable that equals one if the website of an ICO project is still working as of July 2019. Coinmarketcap listing A dummy variable that equals one if an ICO is listed on an exchange and its trading information is collected by coinmarketcap as of July 2019. Various products A dummy variable that equals one if an ICO has delivered preliminary products, such as a cell phone application, a web interface linked to its blockchain or an exchange platform as of July 2019.    Blockchain Information Etherscan partial_commitment A dummy variable that equals one if ICO founders verified their smart contract code on Etherscan, and allow investors to observe token distributions full_commitment A dummy variable that equals one if a smart contract allows investors to observe cash inflows, e.g., inflows of bitcoins, to an ICO project. early_token_transfers Natural logarithm of one plus the number of token transfers in the first one-fifth of whole ICO periods. day1_token_transfers Natural logarithm of one plus the number of token transfers in the first day of ICO periods.           87  ICO Characteristics ICObench log_team Natural logarithm of one plus the number of team members for an ICO social The average of dummy variables if an ICO uses the following social media: Twitter, Facebook, Reddit, Bitcointalk, Medium, and Telegram. pre_sale A dummy variable that equals one if there is a pre-sale or private fundraising round for an ICO. log_ico_len Natural logarithm of one plus the number of days between ICO start and end dates. soft_req A dummy variable that equals one if there is a soft cap requirement for an ICO. hard_req A dummy variable that equals one if there is a hard cap requirement for an ICO. github A dummy variable that equals one if an ICO has a GitHub repository. whitepaper A dummy variable that equals one if an ICO has a whitepaper.    Founder Characteristics LinkedIn num_prior_startups Natural logarithm of one plus the number of previous projects that are founded by the ICO founder. prior_blockchain_exp A dummy variable that equals one if the founder has previous blockchain/crypto experiences. high_edu A dummy variable that equals one if the founder has a master’s degree or above.      88  A.2 An Example of Ethereum Address      89  A.3 Examples of Blockchain Information Blockchain information for ICOs without smart contracts  Blockchain information for ICOs with smart contracts   90  A.4 Examples of Smart Contracts Code ICOs with smart contracts, but the contract source code is not verified.  ICOs with smart contracts and the contract source code are verified.  Note that an ICOs makes partial disclosure commitments if its smart contract code is verified.  91  A.5 Examples of ICOs That Make Partial or Full Disclosure Commitments ICOs with smart contracts that do not disclose flows of cryptocurrencies (partial commitments).  ICOs with smart contracts that disclose flows of cryptocurrencies (full commitments). Note that there is a transfer of 45 Ethers to the ICO team’s accounts in this case.  92  A.6 Sample Code of Smart Contract Sample code of smart contracts  Pseudocode of smart contracts: When the contract receives Ethers from a buyer: Check if the buyer is valid; Check if the amount sent by the buyer is greater than 0; Check if the ICO is not in an emergency stop; Check if the ICO has not ended; If all the above conditions are met: Calculate token amounts to be issued (the price is 960 tokens per ETH) Check if the total token amount exceeds total supply Check if the total Ether amount exceeds ‘hard cap’ If all the above checks are passed: Transfer tokens to the buyer Transfer Ethers to the team wallet Broadcast the transfer event to the blockchain (items with underscores are predetermined and immutable.)  93  A.7 Blockchain Information Under Each Commitment Decision  No commitment Partial commitment Full commitment Date/time of transactions  No Yes Yes Investors’ accounts No Yes Yes Firms’ accounts No No Yes New coins generated No Yes Yes ‘Bitcoin’ received No No Yes     94  Appendix B  Appendix for Chapter 3 B.1 Proof of Theorem 1 Proof. In equilibrium, the uninformed investors can distinguish the managers of different types, so I can rewrite the uninformed investors’ pricing rule by combining the equation (3.5) and (3.6).  𝑃𝑢 ={    𝐸(𝜃|𝑡 = 𝐻𝑖𝑔ℎ, 𝑠 = 𝐺𝑜𝑜𝑑) − 𝑐, 𝑖𝑓 𝑐𝑜𝑚𝑚𝑖𝑡 𝑎𝑛𝑑 𝑑𝑖𝑠𝑐𝑙𝑜𝑠𝑒 𝑠 = 𝐺𝑜𝑜𝑑𝐸(𝜃|𝑡 = 𝐻𝑖𝑔ℎ, 𝑠 = 𝐵𝑎𝑑) − 𝑐, 𝑖𝑓 𝑐𝑜𝑚𝑚𝑖𝑡 𝑎𝑛𝑑 𝑑𝑖𝑠𝑐𝑙𝑜𝑠𝑒 𝑠 = 𝐵𝑎𝑑𝐸(𝜃|𝑡 = 𝐿𝑜𝑤, 𝑠 = 𝐺𝑜𝑜𝑑) − 𝑐, 𝑖𝑓 𝑛𝑜 𝑐𝑜𝑚𝑚𝑖𝑡 𝑎𝑛𝑑 𝑑𝑖𝑠𝑐𝑙𝑜𝑠𝑒 𝑠 = 𝐺𝑜𝑜𝑑𝐸(𝜃|𝑡 = 𝐿𝑜𝑤, 𝑠 = 𝑏𝑎𝑑), 𝑖𝑓 𝑛𝑜 𝑐𝑜𝑚𝑚𝑖𝑡 𝑎𝑛𝑑 𝑤𝑖𝑡ℎℎ𝑜𝑙𝑑 𝑠 (3.10) The conditional expectations can be expanded following the Bayesian rule.  𝑃𝑢 ={        𝐸(𝜃|𝑡 = 𝐻𝑖𝑔ℎ, 𝑠 = 𝐺𝑜𝑜𝑑) − 𝑐, =𝑎 ∗ 𝑏𝑎 ∗ 𝑏 + (1 − 𝑎) ∗ (1 − 𝑏)− 𝑐𝐸(𝜃|𝑡 = 𝐻𝑖𝑔ℎ, 𝑠 = 𝐵𝑎𝑑) − 𝑐, =𝑎 ∗ (1 − 𝑏)𝑎 ∗ (1 − 𝑏) + 𝑏 ∗ (1 − 𝑎)− 𝑐𝐸(𝜃|𝑡 = 𝐿𝑜𝑤, 𝑠 = 𝐺𝑜𝑜𝑑) − 𝑐, =(1 − 𝑎) ∗ 𝑏(1 − 𝑎) ∗ 𝑏 + 𝑎 ∗ (1 − 𝑏)− 𝑐𝐸(𝜃|𝑡 = 𝐿𝑜𝑤, 𝑠 = 𝑏𝑎𝑑), =(1 − 𝑎) ∗ (1 − 𝑏)(1 − 𝑎) ∗ (1 − 𝑏) + 𝑎 ∗ 𝑏 (3.11) In equilibrium, the manager maximizes his wealth by maximizing the share price. There are two implications: First, the commitment decision maximizes the expected payoffs before observing the signal (S). Second, the disclosure decision maximizes the payoffs after observing the signal. Before observing the signals from the informed investors, the manager already knows his type, so that he has better information about what the informed investors’ signal might be. The manager of a high type firm has updated prior that P(Success|t = High) = a, while the manager  95  of a low type firm believes that P(Success|t = Low) = 1 − a. Consequently, their assessed probability of a good signal (S) is different.  𝑃(𝑠 = 𝐺𝑜𝑜𝑑|𝑡 = 𝐻𝑖𝑔ℎ) = 𝑏 ∗ 𝑎 + (1 − 𝑏) ∗ (1 − 𝑎) 𝑃(𝑠 = 𝐺𝑜𝑜𝑑|𝑡 = 𝐿𝑜𝑤) = 𝑏 ∗ (1 − 𝑎) + (1 − 𝑏) ∗ 𝑎 (3.12) To make sure that the commitment (non-commitment) is the best strategy for the high (low) type firms, the following conditions must satisfy.  (𝑠 = 𝐺𝑜𝑜𝑑|𝑡 = 𝐻𝑖𝑔ℎ) ∗ (𝐸(𝜃|𝑐𝑑 = (𝑐, 𝑑),  𝑠 = 𝐺𝑜𝑜𝑑) − 𝑐) +𝑃(𝑠 = 𝐵𝑎𝑑|𝑡 = 𝐻𝑖𝑔ℎ) ∗ (𝐸(𝜃|𝑐𝑑 = (𝑐, 𝑑),  𝑠 = 𝐵𝑎𝑑) − 𝑐) > 𝑃(𝑠 = 𝐺𝑜𝑜𝑑|𝑡 = 𝐻𝑖𝑔ℎ) ∗ (𝐸(𝜃|𝑐𝑑 = (𝑛𝑐, 𝑑), 𝑠 = 𝐺𝑜𝑜𝑑) − 𝑐) +𝑃(𝑠 = 𝐵𝑎𝑑|𝑡 = 𝐻𝑖𝑔ℎ) ∗ 𝐸(𝜃|𝑐𝑑 = (𝑛𝑐, 𝑛𝑑)). (3.13)   𝑃(𝑠 = 𝐺𝑜𝑜𝑑|𝑡 = 𝐿𝑜𝑤) ∗ (𝐸(𝜃|𝑐𝑑 = (𝑐, 𝑑),  𝑠 = 𝐺𝑜𝑜𝑑) − 𝑐) +𝑃(𝑠 = 𝐵𝑎𝑑|𝑡 = 𝐿𝑜𝑤) ∗ (𝐸(𝜃|𝑐𝑑 = (𝑐, 𝑑),  𝑠 = 𝐵𝑎𝑑) − 𝑐) < 𝑃(𝑠 = 𝐺𝑜𝑜𝑑|𝑡 = 𝐿𝑜𝑤) ∗ (𝐸(𝜃|𝑐𝑑 = (𝑛𝑐, 𝑑), 𝑠 = 𝐺𝑜𝑜𝑑) − 𝑐) +𝑃(𝑠 = 𝐵𝑎𝑑|𝑡 = 𝐿𝑜𝑤) ∗ 𝐸(𝜃|𝑐𝑑 = (𝑛𝑐, 𝑛𝑑)). (3.14)   96  To eliminate the case where disclosure costs are so large that even good signals are not disclosed, the disclosure cost c must satisfy the following condition.  𝐸(𝜃|𝑐𝑑 = (𝑛𝑐, 𝑑), 𝑠 = 𝐺𝑜𝑜𝑑) − 𝑐 > 𝐸(𝜃|𝑐𝑑 = (𝑛𝑐, 𝑛𝑑)) (3.15) Rearranging items in (3.13), (3.14), and (3.15) yields equation (3.7), (3.8), and (3.9).                 □ 

Cite

Citation Scheme:

        

Citations by CSL (citeproc-js)

Usage Statistics

Share

Embed

Customize your widget with the following options, then copy and paste the code below into the HTML of your page to embed this item in your website.
                        
                            <div id="ubcOpenCollectionsWidgetDisplay">
                            <script id="ubcOpenCollectionsWidget"
                            src="{[{embed.src}]}"
                            data-item="{[{embed.item}]}"
                            data-collection="{[{embed.collection}]}"
                            data-metadata="{[{embed.showMetadata}]}"
                            data-width="{[{embed.width}]}"
                            data-media="{[{embed.selectedMedia}]}"
                            async >
                            </script>
                            </div>
                        
                    
IIIF logo Our image viewer uses the IIIF 2.0 standard. To load this item in other compatible viewers, use this url:
https://iiif.library.ubc.ca/presentation/dsp.24.1-0392426/manifest

Comment

Related Items