Impact of Regulations on ICO Performance

Размещено на http://www.allbest.ru//

Размещено на http://www.allbest.ru//

GOVERNMENT OF THE RUSSIAN FEDERATION

NATIONAL RESEARCH UNIVERSITY

HIGHER SCHOOL OF ECONOMICS

Institute for Statistical Studies and Economics of Knowledge

Master of Governance of Science, Technology, and Innovation PROGRAM

MASTER THESIS

Impact of Regulations on ICO Performance

Student

Alexey Chekalin

Research Advisor

Dr. Yury Dranev



List of abbreviations

DLT -- Distributed Ledger Technology

ICO -- Initial Coin Offering

USA -- United States of America

PoW -- Proof of Work

PoC -- Proof of Concept

BTC -- Bitcoin

SEC -- The United States Securities and Exchange Commission

P2P -- Peer-to-Peer

R&D -- Research & Development



Summary

In this paper, approaches to Initial Coin Offering (ICO) regulation are explored and described. The impact of some of the regulative initiatives, infrastructural, business and IT-related factors on the short-term performance of ICOs is assessed through the application of event study methodology. To provide the basis for the analysis, a general explanation of blockchain and distributed ledger technology (DLT) is provided. The data on the global ICO market is consolidated and an overview of the ICO results, geography and efficiency is provided.

The challenges of gathering information on the coin offerings for the event study and choosing a proper model for expected return are addressed and a number of approaches are discussed. A basic framework for ICO sampling is derived. Cumulative returns from two short-term event windows of 30 and 90 days after the ICO are estimated and used as dependent variables for separate linear regressions.

The interpretation of the results of the analysis shows that certain regulatory restrictions can affect the cumulative abnormal returns with statistically significant coefficients. Implications for future policies on ICOs are outlined in the Discussion section, as well as limitations of the work and suggestions for further research.

coin offering regulative investor



Introduction

ICO is a crowdfunding mechanism that allows businesses to attract money through issuing tokens that can represent ownership of company shares, be a means of payment for the company's future products, services or have other value inside the company's ecosystem. Token Generation Event (TGE) is believed to be a better name for the process as it more accurately reflects the mechanics (Narayanan et al., 2016). The latest definition currently used in academic research is introduced by Adhami, Giudici and Martinazzi (2018) and encompasses the views of Pilkington (2016), Chohan (2017), Sehra, Smith and Gomes (2017): `an open call, through the Internet, for the provision of cryptocurrencies in exchange for tokens generated through smart contracts and relying on the blockchain technology, allowing the pledger to enjoy an exclusive right or reward or financial claim'.

The current definition by SEC (2017) explains that `a virtual currency is a digital representation of value that can be digitally traded and functions as a medium of exchange, unit of account, or store of value, but virtual tokens or coins may represent other rights as well'.

With the rapidly ongoing establishment of distributed ledger technologies, the attraction of capital through token sales can become the easiest way of raising funds due to high liquidity of cryptocurrencies and low transactional costs. Even today with the help of new smart contract platforms Ethereum and Waves it has become increasingly cheap and fast to set up your own personal blockchain and develop a distributed application around it. At the same time, the regulatory answers from governing institutions all around the world are sporadic and vary from complete bans to government support.

Scholars have become increasingly interested in identifying the proper regulatory and infrastructural measures that will help businesses with their ICO campaigns and at the same time making them legal and transparent, as most current approaches seem to lack in efficiency (Lai, 2018).

In order to identify the key policies needed to support token economics in a legitimate way, one needs to assess the current impact of the pioneer regulatory attempts and understand, what factor can positively influence ICO performance.

Thus, the aim of this research will be to assess the influence of regulation on ICOs and find out, whether current approaches provide with statistically significant results.

In this thesis I will

explain the technology behind ICOs

outline the stages a project undergoes in order to carry out an ICO

present the current state of the ICO market and an overview of token sales during the recent years

present results of the literature review on approaches to cryptocurrency regulation

go through the application of event study method and linear regression to assessment of ICO performance

present results of the short-term event studies on empirical data on a sample of ICOs from 2015- 2017.

outline points for further discussion of the topic and explain the limitations of the work

In the first chapter I will focus on the theory behind token sales. This will be followed by a general overview of the token market and current approaches to regulation. Then I will elaborate on the used methods and interpret the results.



Theoretical background of ICOs

Blockchain and Distributed Ledger Technology

Concept

Essentially, an ICO can be viewed as issuing a new cryptocurrency that is linked to the products or services provided by the issuing company (Conley, 2017). The newly created tokens are exchanged for existing cryptocurrencies, in most cases Bitcoin or Ethereum. ICO is a complex phenomenon that combines the principles of shared economies, crowdfunding, advanced cryptography and computer science (Chohan, 2017). To understand how the ICO model of financing works, a general overview of the DLT and specifically, blockchain, has to be provided.

Blockchain was conceptualized in 2008 and put into practice as a backbone for Bitcoin in 2009, however the first ideas on applying cryptography to currencies can be traced back to the works by David Chaum in 1983 (Narayanan et al., 2016). The initial concepts by Chaum combined with influences from Wei Dai's B-money, Nick Szabo's Bit Gold, Adam Back's Hashcash and Hal Finney's Reusable Proof-of-Work were the key predecessors that allowed for the first iteration of blockchain to emerge (Narayanan et al., 2016). It is unclear whether a single creator or a group of developers developed the technology. Bitcoin whitepaper focuses solely on the inner workings of the system and does not mention the team that developed the technology (Nakamoto, 2008).

In the recent years blockchain has become one of the most hyped technologies, even despite its complexity and reasonable skepticism from certain experts (Constellation Research Inc., 2016). It was nearly at the Peak of Inflated Expectations in the Gartner's Hype Cycle in 2016 and it almost hit the Trough of Disillusionment in 2017, with an estimation to hit the Plateau in 5 to 10 years (Gartner, 2017).



Figure 1: The Gartner Hype Cycle for Emerging Technologies, 2017

Source: Gartner, 2016.

Figure 2: The Gartner Hype Cycle for Emerging Technologies, 2018

Source: Gartner, 2017.

Most Blockchain initiatives are still in alpha or beta stage and the development worldwide is hindered by the lack of proven use cases (Constellation Research Inc., 2016). At the same time, particular examples of application are exploited with intensity, including the ICO mechanism, which is claimed to have helped businesses worldwide get over $5.6 billion in funding in 2017 (Token Data, 2017).

At its core blockchain is a distributed secure database. It consists of a string of blocks, each one being a record of data, that has been encrypted and given a unique identifier with the help of a hash function (Nakamoto, 2008). There is a number of computers connected to the network, called `mining computers'. They gather all the inquiries to make changes to the database, validate them and form data blocks. For example, if one was to transfer bitcoins to a particular address, this transaction would have to be collected by one of the mining computers on the bitcoin blockchain network, verified and added to a block.Once the amount of data is sufficient to complete a block, it is broadcasted to the network,and all thecopies of the database that are stored by other participants, receive an update.To understand how exactly a blockchain system works, we will go through the two main components: cryptography and the consensus mechanism.

Cryptographic hash-function

A hash function is used to code different chunks of information with unique identifiers of fixed size. It essentially allows to speed up processing large data through dividing the information into pieces and homogeneously naming them according to the algorithm. The output is called hash valueor simply, a hash.When the output value of hashing two different pieces of information is the same, a collision occurs. The main challenge for designing a hash function is to create an algorithm that allows minimal collision. The procedure must also be deterministic, which means that the results of hashing for a particular input value should be consistent every time. To sum up, each hash must be unique, but repeatable.

For instance, a hash function called MD5 takes 1462 milliseconds to hash 500 megabytes of data and turn it into a 16-byte string. At the same experts find it to be `cryptographically broken and unsuitable for further use' (Dougherty, 2008). That happens because MD5 is vulnerable in terms of possible collisions. Chinese researchers famously were the first to design an algorithm for MD5 collision in 2004 (Wang, Feng, Lai, Yu, 2004). In 2005 there was a successful attempt of identifying collisions for MD5 hash function in under 6 hours using just a PC (Klima, 2005).

Each block of a distributed database such as blockchain, each block is chained with the previous block using a hash value which represents a unique identification number of the data within the block. For instance, Bitcoin's blockchain works with the function SHA 256 being run over the block's data to compress it into a hash. Given just the hash value you cannot recreate the data contained within it, due to the one-sided nature of the hash function.

All blocks created after the first one are securely chained with the previous one. Thus, once recorded, the data in any given block cannot be altered without the alteration of all subsequent blocks. The system of interdependent hashing makes a distributed database particularly secure. You can only input the data to the database and once it is recorded it is almost impossible to change. Therefore, data stored on the blockchain is generally considered incorruptible.

Consensus mechanism

The blockchain database is maintained by a large number of computers that are incentivized to provide computing resources by earning some form of tokens in exchange. But these computer nodes in the network themselves cannot be trusted individually. Thus, it is required for the system to provide mechanism for creating consensus between scattered or distributed parties that do not need to trust each other, but rather need to trust the mechanism. Blockchain is designed to support a system which does not need a centralized component to verify the alterations of the database, so the network relies on a distributed consensus algorithm (Nakamoto, 2008).

As in the case of the bitcoin blockchain this is realized on such a form of challenge, that no one actor on the network is able to solve this challenge consistently more than anyone else on the network. Miners compete to add the next block on the chain by racing to solve a very difficult cryptographic puzzle. The first to solve the puzzle wins the lottery. As a reward for his or her efforts, the miner receives a small amount of newly created bitcoins and a small transaction fee. Such scheme is called proof of work.

The consensus algorithm, like Bitcoins proof of work, functions to ensure that the next block in the blockchain is the one and only version of the truth. It also keeps powerful adversaries from derailing the system. To successfully tamper with the blockchain you would need to alter all the blocks on the chain, redo the proof of work for each block and take control of more than 50% of the peer-to-peer network.

DLT, Blockchain 2.0 and Ethereum

The technology blockchain provides is suitable for storing transactions that contain secure value or can be used as a token of trust. These transactions are called distributed ledgers. Modern blockchains are capable of supporting a whole ecosystem of the geographically dispersed nodes and transactions between them without any centralized regulation. The distributed ledgers can act as empty vessels for any type of transactional data. They are capable of maintaining a database of currencies, real estate ownership rights, patents, votes, identities, health data (PwC, 2017). Moreover, the distributed database can be automated through self-executing smart contracts -- pieces of code which can contain all the terms of agreement and be stored on blockchains

The evolution of the technology led to the so-called Blockchain 2.0, which is designed as a network on which developers could build applications. This was made technically possible by the development of the Ethereum platform. It is an open-source public blockchain based distributed computer platform, featuring smart-contract functionality. It provides a decentralized Turing-complete virtual machine which can execute computer programs using a global network of nodes. Currently, Ethereum is the largest and most popular platform for building distributed applications. The Ethereum platform is believed to have caused the epidemic of ICOs with its own token sale success: the company attracted over $18 million in 2014. Ethereum's own tokens are used to create your own distributed databases with smart contracts on the platform.

Tokens and coins

Coins or also known as altcoins (alternative to Bitcoin) and most of the time they utilize the Bitcoin's original open source blockchain networks and protocols to exist. For example, Namecoin, Peercoin, Litecoin, Dogecoin and Auroracoin are coins. It is important to note, that some altcoins can be based on their own original blockchain protocols and still be called `coins'. This happens due to the fact that they can only act as a means of payment, most of them are just another version of Bitcoin.

A token is a quantified unit of value that is recorded on the blockchain. It might be likes on social media, a currency, the integrity of an ecosystem or an electrical unit. Token networks consist of a network of independent nodes but can self-organize to create a distributed management system through incentive structures. Thus, blockchain is not only an information technology, but also an institution technology as it allows to design incentive structures in the form of token economics and convert centralized organizations in shared network infrastructures.

Through adding a layer of trust and value exchange the blockchain enables the society to create new organizational structures and redefines the way we can share responsibilities and interact with each other. However, much of what the blockchain promises will only be possible given parallel advancements in the internet of things, datification and advanced analytics.

ICO mechanics

Researchers claim that in most cases, ICOs are used to attract funds for projects, connected with DLT and supporting infrastructure, blockchains, and cryptocurrencies (Sehra et al., 2017). Some reports claim that in 2017 less than 40% of the projects were related to DLT or distributed applications (Token Data, 2017; PwC, 2017). However, researchers believe that an ICO for companies is not involved in DLT will most likely be less beneficial. This is due to the nature of token price dynamics. All cryptocurrencies eventually lose their value if they are not actively traded or used in the project's own DLT ecosystem.

ICOs and IPOs have a lot in common, the key difference is the approach to value assessment and distribution (Sehra et al., 2017). The goal of the company during a coin offering is to attract as much capital as possible during a time window that is announced prior to the token sale. However, the tokens' market value in uncertain until they are listed on exchanges. In short, the ICO process can be divided into three steps:

Figure 3: Stages of the ICO process

Source: own elaboration

The whole process from announcement to listing can take up from 3 months to years depending on the funding goal, marketing, and preliminary agreements with exchange platforms. According to some researchers, the average sale window of ICOs lasts for approximately 27 days, `but the duration is heterogeneous -- some ICOs close in a few days, whereas other are open for some months' (Adhami et al., 2018). However, there is no definite consensus on the average length of an ICO process, as both the token sale window and the fundraising goals are limited only to the ambitions of the projects' teams. A token sale by EOS, a company developing a decentralized operating system to power distributed applications, has been going on in stages for almost a year with the goal of selling close to 1 billion tokens. The first five days of the sale were devoted to an initial unlimited ICO, with 200 million tokens distributed afterwards proportionally to the contributions. This was followed by the second stage consisting of 350 independent token sale windows 23 hours long each with 2 million tokens distributed after each individual sale is over.

Whitepaper

In order to carry out an ICO a company has to prepare a `whitepaper' or a `token sale term'. This document usually contains business-related information as well as technical nuances and all the data necessary for the investors willing to participate in the crowd sale:

a description of the products or services the company provides

a business plan and a correlating roadmap with financial milestones

a team description

details of the IT protocols and the blockchain infrastructure used to issue the tokens

token price, supply and planned distribution

There is no standardized structure, so some documents can be more technical than others. There have been a few attempts at addressing the challenge of whitepaper evaluation by businesses, individual experts or media, however, there is still no consensus on the contents and layout of the documentation. According to some researchers, most ICO white papers lack a thorough justification of the use of blockchain and implementation of tokens in their business model (Wilson, 2017; Hileman, G., & Rauchs, 2017).

Token distribution

Not full amount of token is distributed through ICO. There is no generally accepted scheme on token distribution. More than 50% of all tokens should go on sale during private sale, pre-sale and sale, otherwise the ICO will be considered technically invalid. In ICO cases considered further, typically 55-60% of tokens were distributed during sale periods.

10% of tokens are usually held for the team and shared among developers and managers. It is common to set limitations for the team on selling tokens. Most common is that they cannot sale 5-10% of their tokens during next 6-12 month. (EY, 2017) So, they are more motivated to work in the product in future and multiply their income.

1-5% of tokens are held as a bounty -- reward for user's promotion or early buying of tokens. Such bonus is an important part of most ICOs as it lets shift marketing facilities to users, who get additional tokens for posting ICO news on their blog or attracting customers other ways. While decision on buying tokens is usually made because of emotions rather than rational approach, bounty plays a significant role in sales.

There are no strict rules for the remaining tokens. They can be distributed through sales, held for additional bonuses and community or for infrastructure needs of the project.

Tokens are sent to customers after confirmation of payment. However, they are blocked during pre-sale and remain inactive until ICO is not finished. They cannot be sold or exchanged in this period. When ICO is finished, tokens are automatically activated.

There are basics principles in tokens distribution:

Every customer gets equal price (during the same period of sales)

No one gets anything for free

All users have equal chances to participate

Transaction cost should be as small as possible

Stimulus for developers

Token sales may be capped or uncapped. In first case, limited amount of tokens are distributed for a fixed period of time. It usually provokes hype due to Fear of Missing Out. Uncapped distribution means that tokens are divided proportionally to amount of investment. In such case investors do not know, how much will the token cost until the end of sale.

Token market price is not easily predicted as it gets stable only after a period of time when it is sold or exchanged on the market. So, when a token is listed on exchanges, it gets reevaluated according to existing supply and demand.

If project's token is supposed to be used on the market and not only for internal use within the project, listing is a crucial part of an ICO. It means that is accepted as an exchange mean for other tokens or cryptocurrencies on specific platforms. It makes buying and selling the token easier and confirms its legitimacy.

However, the project also has to have a solid need for listing. It cannot be speculative reason like pumping token's price. Listing let's projects enlarge and diversify its audience and creators have to keep in mind, whether they need it or not.

There are two main exchanges, Bittrex and Poloniex, and several smaller. Each has its own requirements for listing, some are more open, some, like Bittrex, have stricter rules. Generally exchanges require:

Coin name, its description, logo or trading symbol. Exchanges check whether projects is not scam, speculative and doesn't breach security of users.

Approved accounts of one or several team members. Once account is approved and can be linked to a specific individual, it saves time for both exchange and project.

Github or any other open-source link for due diligence. It works as a proof that the team has a functional digital product.

Usage limitations

Despite the fact that blockchain community positions ICO as a breakthrough financing method, it has its own limits, first of all, technological.

Blockchain lets independent participants reach consensus without centralized intermediary. On the other hand, it is slow and expensive database. Lack of justified necessity in blockchain is a common problem among ICO projects (EY, 2017). Blockchain based crowdfunding might be appropriate for cases, when:

There is a long chain of parties in they need coordination.

Parties are independent and not trustworthy.

Existing centralized intermediary is more expensive than blockchain or there are safety risks.

Parties approve of full transparency of any records.

Projects aims at `tokenizing' a specific asset.

Nobody controls more than 50% of nodes.

There are clear rules and oracles for smart-contract.

These are cases, when blockchain is a more effective decision then existing ones, therefore ICO is an appropriate way of funding, However, open blockchain is not a perfect match for other cases such as when:

Speed of transactions is required.

There are a few nodes or there is no chance in reaching consensus.

Confidentialityisrequired.

Conditionsof a contract may change during its execution.

There is complex pricing or course manipulation risk

Volatility of the token rate leads to an increase in the cost and loss of competitiveness of the product or service

It is necessary to provide «right to be forgotten»

That provokes two kinds of problem. First one is ICO projects that use blockchain not as an effective problem-solving mean, but as a hyped way to raise funding. However, hype is gradually evaporating and the second problem appears -- despite many predictions, practical usage of public blockage in a variety of industries, except for financial and infrastructure, is limited due to the low speed and high transaction costs.

Another limitation for ICO is limited capacity of platform. The most popular platform for ICO is Ethereum, which is overloaded because of its popularity. Growing demand increases the cost of Ether and, consequently, the cost of ICO (EY, 2017). VISA makes 24 000/second, whereas Ethereums's capacity is only 20 (FabricVenturexЧToken Data, 2018).



ICO market overview and regulation

ICO market

Sources

2017 will go down as the year that ICOs dominated a large part of the daily conversations in the worlds of cryptocurrencies, decentralized web economics and venture capital. More than $5,6 billion of capital was raised in 2017. This compares to 1 billion dollars of venture investing in blockchain startups in the same timeframe and 240 million dollars raised by token sales in 2016.

Due to the lack of centralized authority and official general statistics on the market, the research on the state of ICO price dynamics, the number of projects that and industrial cross-section is mostly limited to business reports. The analysis in all of them is performed using data gathered from the open sources on the internet, media, cryptocurrency exchanges and the projects' whitepaper (JP Morgan, 2018). The sources that have the largest databasesand are most frequently used include:

CoinDesk

Coinbase

TokenMarket,

TokenData

Coinschedule

These sources aggregate information on the price dynamics of certain tokens, on the amount of raised funds and can provide insights on the economic sense of the tokens, the industry and general information on the projects that announce crowd sales. Therefore, the data in them can vary. The most recent reports that are publicly available:

EY research: initial coin offerings, 2017

JP Morgan, Decrypting Cryptocurrencies: Technology, Applications and Challenges, 2018

Societe Generale: cryptocurrencies, bitcoin and blockchain -- an education piece on how they work, 2018

Autonomous NEXT: Crypto Fund List, 2017

Fabric Ventures and TokenData: The State of the Token Market, 2017

CoinDesk: State of Blockchain, 2018.

The data on ICO market Researchers believe the rising demand on cryptocurrencies and tokens has recently been fueled by a radical increase in the price of Bitcoin.

Figure 4: Funds raised from 2015 to October 2017, cumulative sum, billion USD

Source: EY, 2017

ICO results

In a recent report by EY the cumulative market capitalization of 150 projects on the day of ICO announcement was compared to their capitalization in December of 2017. The authors concluded that the main contributors to an almost 4000% growth of the portfolio were Ethereum and other projects that develop blockchain infrastructure (EY, 2017).

During 2017, a total of 902 startups announced the start of an IPO campaign. Out of them 435 were successful and raised $5,6 billion (FabricVenturexЧToken Data, 2018). Average Capital Raised in 2017 was $12,7, however median capital raised was $4,4.

We can divide unsuccessful projects into two groups:

Failures/Refunds:131 projects reported that they failed to meet their minimum threshold ?

Unreported:347 sales did not report the end result of their token sales.?

10 largest token sales raised 25% of all capital. It's notable, that 5 projects out of these 10 focus on Blockchain infrastructure.

Table 1. 10 largest token sales in 2017

Project	Sector	Raise
Tezos	Blockchain infrastructure	$230 498 884
Filecoin	Blockchain infrastructure	$200 000 000
Sirin Labs	Other	$157 885 825
The Bancor Protocol	Blockchain infrastructure	$153 000 000
Polkadot	Blockchain infrastructure	$144 347 146
QASH	Trading&Exchange	$108 174 500
Status	Blockchain infrastructure	$107 664 904
Kin	Payments	$98 500 326
COMSA	Finance	$95 614 242
TenX	Finance	$83 110 818
Total		$1 378 796 646

Source: FabricVenturexЧToken Data, 2018

It is important to state, that success rate of ICOs decreased throughout the year. 93% of projects were successful in June, whereas only 25% could collect necessary amount in November.



Figure 5: Success rate of ICOs from June to November, 2018

Source: FabricVenturexЧToken Data, 2018

Geographical diversity

There were 56 different countries in which legal entities behind different tokens are registered or located. It means, that community is geographically diverse. However, capital behind those ICOs is no so widespread -- top 10 countries have raised more than 75% of all capital.

Figure 6. ICO projects by country/jurisdiction, million USD

Source: EY, 2017

Analysis of Token Data shows that on average tokens have returned 12,8 times the initial investment in dollar terms versus 7,7 times for Ethereum and 4,9 times for Bitcoin during 2017. However, closer look shows that returns are skewed towards a handful of tokens issued in the first quarter of 2017, when the ICO hype had not fully erupted. Avarege token returns have been trending down since. Additionally, a breakdown of median returns still shows outperformance by tokens but paints a much more nuanced picture.

Approaches to ICO regulation

Why ICO needs to be regulated?

ICO is binded with huge cashflow and provokes interest from hackers. 10% of ICO projects lose money because of hackers' activities, who use hype, anonymity, lack of centralized regulator and irreversibility of blockchain operations. At the same time founders are more focused on accumulation of investors, rather than security issues.

Not only projects themselves are victims, but also inverstors. Popular methods are substitution of the project wallet address, getting access to private keys and theft of funds from wallets. Most popular tool are fishing schemes, that are still relevant due to low awareness about the subject. Using fishing hackers can steal up to $1,5 million a month. It also leads to leakage of personal data and endangers team members' reputation.

Considering potential losses and crime capabilities, regulators cannot avoid the phenomenon of ICOs and crypto economy in general. Majority of governments are following the same pass: from ignoring through active discussion to prohibition or regulation of IC (EY, 2017).



Business sources on ICO regulation

The data from the reports provide us with the following information. ICO is already banned in China and South Korea, however, other countries opt for regulation rather than prohibiting. (EY, 2017). In order to get a more holistic view, academic databases are to be examined.

Table 2. Regulation of ICO and token in different countries

Country	2016	2017, Jan - Jun	2017, Jul - Sep	2017, Sep - Nov
China	Active discussion	Active discussion	Ban	Ban
USA		Active discussion	Regulation	Regulation
GB		Active discussion	Active discussion	Active discussion
Canada			Regulation	Regulation
Hongkong			Active discussion	Regulation
Australia			Active discussion	Regulation
Switzerland	Active Support	Active Support	Active Support	Active discussion
Japan		Regulation	Regulation	Regulation
Malaysia	Active discussion	Active discussion	Active discussion	Active discussion
Estonia			Active discussion	Regulation
Russia	Active discussion	Active Support	Active discussion	Active discussion
South Korea			Active discussion	Ban

Source: EY, 2017

Academical sources on ICO regulation

The legal status of ICOs in unclear both form the point of global and national regulators, as is the status of cryptocurrencies in general. However, the challenge of deriving regulation frameworks and policies for cryptocurrencies has been recognized since 2011 (Conley, 2017; Yadav, 2017) Due to the scarcity of the academic work on the topic, a census-like approach for mapping the existing body of literature was applied. A similar method is used by Holub & Johnson (2017) to qualitatively assess the influence of Bitcoin on the academic literature.For the literature review on the challenges of ICO regulation, the following databases were examined with the key words `ICO', `cryptocurrency', `regulation', `token', `coin' and the dates of publication between 2011 and 2018:

Table 3. Overview of the academic databases

Database	Raw results	Left out	Refined results
EBSCOhost Research Databases	812	Not relevant, news and magazine articles: 794	18
Jstor	14	Not relevant: 9 Duplicates: 2	4
AEA Journals	1		1
Science direct	120	Not relevant: 106 Duplicates: 6	8
Emerald	13	Not relevant: 10	3
Springer Link	20	Not relevant: 16	4
SSRN	250	Not relevant: 226 Duplicates: 17	7

Source: own elaboration

The key problem that academic research identifies in regulating cryptocurrencies is their outstanding liquiditywith minimal transactional costs (Elendner, Trimborn, Ong & Lee, 2015). Their efficiency in capital attraction have exceeded VC and IPO investments and this poses a threat to the stability of the traditional financial markets. Another commonly recognized problem is the so-called `scam companies', that are set up exclusively for ICOs and end up not delivering on their promises to investors or even not distributing the tokens (Harwick, 2016). Token holders usually invest in a future product, which may not have any existing proof of concept. Therefore, a number of countries have already started deriving regulatory frameworks for the token crowd sales.

Unlike traditional businesses with a history of financial transactions or a representation on the stock exchange, most crypto platforms cannot generate revenue to offset costs. Moreover, cryptorojects may not sometimes even have employees in the traditional sense that create and advertise the platform product. Some researchers argue, that most if not all of the raised fundsareattracted in cryptocurrencies, which proposes a challenge of legitimization and taxation of the money (Flьhmann & Hsu, 2018).

ICO investors have no preemptive rights or other anti-dilution protections. If the promoters decide to issue more reserve tokens to additional investors, the ICO investors may be diluted in the future. The only real control token holders have to protect themselves is to sell their tokens post-ICO (Kaal, 2018).

In addition to that, Jackson (2018) claims that token holders typically do not receive a liquidity preference that would protect them in the case of bankruptcy or termination of the platform they invested in. If they go bankrupt, token holders have nothing after the debt holders and outside creditors were satisfied with the liquidation value of the entity. By contrast, in a typical venture capital seed stage investment, the venture capital fund should typically obtain at least a simple liquidity preference. This allows venture capital funds to reclaim their initial seed investment before other creditors are satisfied (Kaal, 2018). Researchers identify the following countries as the most supportive of ICOS (Jackson, Kaal, 2018):

United Kingdom

Switzerland

Singapore

Lithuania

Australia

Germany

Canada

Israel

Lichtenstein

Luxembourg



Methodology and approach

Event study

In this chapter I willelaborate on the choice of the research methods and present a few cases application of such methods for cryptocurrencies. This is followed byan explanation of the process of sampling, gathering data, and assessing ICO performance. I will also focus on defining the key factors that can influence token prices after the ICO and deriving hypotheses for the regression.

To understand which approach to estimate the impact of external factors on performance could be applied to ICO, I decided to study the most common methods of analysis that are used with IPOs, as the similarities between the two fundraising mechanics have already been established by the media and certain researchers (Conley, 2017). IPO performance is most commonly analyzed through either event studies, accounting based methods or case studies. According toWang & Moini (2012)over 90% of the academical work on the impact of M&A announcements on stock prices used event studies as the primary method of performance evaluation. In the overview of event study literature and related econometrics challenges, Kothari & Warner (2007) find the works by Fama, Fisher, Jensen, and Roll (1969), Brown & Warner (1980, 1985), Fama (1991, 1998) and Campbell, Lo, and MacKinlay (1997) to be most credited by the researchers and refer to them as `classical' or `standard' event study references. They also emphasized that the core framework of the method has not changed for over 30 years, so its conditions and limitations are well-understood (Kothari & Warner, 2007).

Event study method is used to assess the influence of certain time-bound events on the returns of the asset prices. It is versatile, as there is no limitation on the types of events that can be researched and, thus this method is also used in political science, marketing and IT and can be applied to both short-term and long-term assessment.Event studies are also proven to be applicable for measuring the effects of regulation and assessment of damages in legal liability cases (Kothari & Warner, 2007; Lamdin, 2001; Bushnell, Chong,and Mansur, 2009). The core idea behind the method is the efficiency of the financial market, which results in the fact that all new information is always represented through the stock prices. Therefore, the impact of certain events can be measured with the help of statistical analysis. And vice-versa, long-term event studies can be used to test market efficiency. The return is decomposed intonormaland abnormal components. Assuming the time of the event is t = 0, the return on a sample security i for the period t related to the event equals:

With being the normal or predicted return on the stock, and being the unexpected component. Therefore, the abnormal return equals:

The historical data of the stock prices is used to identify abnormal returns during the time around the event. There are severalmodels of identifying normal and, consequently, abnormal returns on the market. The three simplest and most common approaches are:

Constant mean return model

Market adjusted model

Market model

The constant mean return model is based around the idea that the return on a security is not subject to change over time and equals to its historical average. Under this approach:

Where is the abnormal return, is the observed return and -- the mean return over a given period.The key drawback of the model is that it does not account for the changing conditions on the market. On the other hand, Brown and Warner (1985) claim that the constant mean return model often provides with outcomes similar to those of more complicated models, especially in the short-term period.

The market adjusted model implies that an indicator of a market-level return can be used as a measure of expected return.Under this approach:

Where isa return on a market portfolio for a time period of t. In economic research, is usually represented by a broad-based stock index return, for instance, S&P 500.

The market model requires a regression analysisto understand the parameters of linear relation between the return on a particular security and the market return.The data used for the regression is taken from a time period before the event, called estimation window. The results of the regression provide the researcher with the means to predict the normal return of a specific stock for a given time period.

Under this approach, the abnormal return is the difference between the estimated return and the observed return:

Where is the predicted return from the regression equation above. This model allows for a more sophisticated and detailed approach: it includes the individual risk profile and market behavior of the sample stock. The market model is regarded as standard and is the most commonly used model in the census of over 500 event studies examined by Kothari & Warner, 2007.

There are other complex models, for instance CAPM orFama And French 3 factor modelthat incorporate more factors and are capable of providing more accurate risk adjusted results. Yet, they require more data and a better-established efficient market than the one ICO tokens currently form.

In order to narrow down and specify the period of observation it is essential to choose an event window, which is usually the time period before and after the event. While assessing IPO performance, the public offerings themselves are treated as events with the event window stretching forward. Windows of less than 12 months are generally considered short-term (Kothari & Warner, 2007).

Application of event study in cryptoeconomics

There is limited evidence of event study application in qualitative research in the field of cryptoeconomics. A study on Bitcoin price drivers by Seys (2016) is concentrated around two research tools. The first in an OLS regression model which aims at explaining the relation between the bitcoin price index is USD and a number of technical, financial and infrastructural factors that include currency rates between USDand CNY, the difficulty of mining, prices of oil and gas, number of Wikipedia searches for `Bitcoin'total volume of Bitcoin and Bitcoin trading volume in China.The second research tool is an event study which aims at exploring the behavior of the Bitcoin price index around positive and negative media announcements.

Seys (2016) uses a market model with MSCI World Index representing the market portfolio return and an estimation window of 120 days prior to the first event to come up with a regression equation. The researcher assumed Bitcoin to be significantly more sensitive to the media than the stock market and used a short event window of 5 days: 2 before and 2 after the event. During 8 out of 26 examined eventsthrough 2013 to 2015 the Bitcoin price index performed significantly different to the chosen market portfolio index, with negative news seemingly provoking larger abnormalities more quickly. Positive news also correlated with abnormal performance, however in a different, slowerfashion. Seys (2016) concludes that the Bitcoin market is still not completely efficient, with skepticism for positive news and exaggerated declines in demand after negative news.

Another example of event study application in this field is the research by Adhami, Giudici, G and Martinazzi (2018), which is devoted to identifying the determinants of ICO success. As in the previous case, it is accompanied by a regression-based analysis. The authors assumed that access to certain information regarding the token sale, for example the detailed explanation of the token distribution or the location of the team is a more significant factor of ICO success, that the media attention to the project or the bounty campaign. The researchers used a logistic regression model to understand the relations between the independent binary coded determinants (information availability, industry, token type, etc.) and the dependent variable of ICO success and found statistically significant positive coefficient for the availability of the source code. However, most of the other coefficientswere found to be not significant, including the availability of white paper.

The event study in the research was aimed at identifying connections between abnormal returns on the token market and media announcements. A total sample of 202 publicly available tokens was used to calculate abnormal returns using the market model with a market portfolio represented by an index of five cryptocurrencies, weighted by market capitalization: Bitcoin, Bitcoin Cash, Ethereum and Ripple.A sample of 18 events from 2016-2017 were analyzed. The event study showed statistically insignificant abnormal returns after most good news and significant negative performance after half of the negative announcements.

There is conclusive evidence of event study method application in academic research devoted to IPO performance, impact of regulation and some use cases substantiating the possibility of applying event studies for cryptoeconomics. Therefore, I decided to use the method for the goals of this thesis.To conduct an event study, one has to:

Identify an event window

Choose a model for expected returns

Select criteria for the sample

Gather information

Measure abnormal returns

Test them for statistical significance

Analyze and interpret the empirical results

Event window and market model

Judging from the insights of the previous research on token sales established in this thesis, one can assume high volatility on the market and fast response to new information. Moreover, gathering historical data on tokens that lost their value or on companies that completed ICOs but disappeared from the public eye because of unrealized expectations may impose a serious challenge due to the lack of a centralized regulated statistical database. Therefore, short-term event windows were chosen for this particular research.

Seys (2016) used daily returns on the Bitcoin price index, which is updated every two minutes by CoinDesk and chose an event window of only five days with the event itself being in the middle. As my goal in the thesis is assessment of the impact of regulatory factors, I opt for a larger event window and use daily returns for 1 month and 3 months after the tokens are listed on an exchange and made available for public trade. A month is considered to consist of 30 days, contrary to the classic 20-trading-days-per-month approach for IPO event studies. The reason for that being the absence of the trading day concept for tokens. Event windows longer than that 3 months might be less indicative in the context of cryptoeconomics, based on the assumption that most tokens inevitably lose their value after they are listed. This is caused by the investors selling the tokens in the first few months after the ICO in order to get the profits after the initial inflated pricing (EY, 2017).

In the case of assessing initial ICO performance there is no estimation window and, consequently, both constant mean and market models cannot be properly implemented, which leaves the option of an adjusted market model. Further drawing parallels with IPO performance research, a market portfolio index CRIX is used as a source of expected returns. CRIX is an index comprised of 20 cryptocurrencies and tokens weighted proportionally to their market capitalization (see Appendix 1). It was conceptualized in 2016 by Trimborn & Hдrdle and is calculated with the data provided by websites CoinGecko, Coinmarketcap and Blockchain.info. The earliest record which was estimated retrospectively is 01.04.2014. The index is updated every five minutes with closing prices for each day calculated at midnight. The data can be extracted from the website with the help of an API request.

It is possible to derive an index of your own for an event study through putting together a portfolio of assets that could give a valid representation of the market. In the case of token market, even using either Bitcoin or Ethereum by themselves as proxies for expected return can be considered valid options. However, using CRIX provides a more holistic view, as it contains information on a number of blockchain-based currencies of different types and functions, thus, making it the most and currently the only broad index for the world of cryptoeconomics. Moreover, CRIX is looked after by a team of researchers who ensure the quality of calculations and regular updates. For the purpose of this research I used daily closing CRIX indices.

Approaches to sampling

In order to obtain historical data on the prices of tokens starting from the listing date with the websites of the projects, their white papers and the exact amount of capital raised during token sale, I needed to find a database on completed ICOs. The information on the attracted funds and success of ICOs appears to be scattered across multiple complementary sources with the largest databases on TokenData.io and Coinmarketcap.com. The latter of them has open API which allows to gather both historical and current data on the token prices, circulating supply, total market capitalization and 24-hour trade volume of 1592 cryptocurrencies (Coinmarketcap.com, 2017). Unfortunately, there is no tool that allows to sort the database by the function of the cryptoasset and get only tokens or only coins. It also not possible to rely on Coinmarketca.com as a source of any additional information.

It is necessary to point out that it is manageable to gather all the announcements of ICOs through browsing BitcoinTalk.org or subreddits on Reddit.com that are devoted to token crowd sales or cryptocurrencies. These are the two main social platforms for the community of enthusiasts and cryptoinvestors and all the ICO-related marketing campaigns start there. Once again, there seems to be no automated tool to get data on the completed ICOs, especially if the crowd sales did not reach record-breaking amounts of attracted capital.

Therefore, a list ofcompleted ICOs had to be handpicked through searching blockchain-related media websites:

CoinDesk.com

Hackernoon.com

CoinBase.com

Steemit.com

TokenMarket.net

Coinschedule.com

CoinGecko.com

IcoAlert.com

IcoList.com

Next.autonomous.com

BitsonBlocks.net

Masterthecrypto.com

Insidebitcoins.com

BTCmedia.org

The search resulted in a preliminary sample of 211 completed ICOs from 2015-2017. It was possible to find out the total amount of raised funds for 156 projects and only 130 of those had complete white papers available on their websites. Out of those 103 were chosen for the final sample, as the data gathered on them from the website and whitepaper was sufficient and their tokens were listed at least 3 months ago. 29 out of 103 were ICOs from 2018, 47 -- 2017, 22-- from 2016, 5-- from 2015.

This dispersion of projects can be explained by two factors. Firstly, the sample is somewhat proportionate to the amount of ICOs conducted throughout these years, with 2017 being the record holder for most token offerings per year. Secondly, data from more recent ICOs is more likely to be still available. Previous data can be put down by both the media and the projects' teams in case of product launch failures. The historical data of each ICO was downloaded from Coinmarket.com, other information was taken from the websites and white papers of the projects (see Appendix 2), as well as GitHub, where the availability of the source code was checked.