Content uploaded by Guillermo Callejo
Author content
All content in this area was uploaded by Guillermo Callejo on Jul 16, 2019
Content may be subject to copyright.
Empirical Analysis of Information Asymmetry in
Initial Coin Offerings
Guillermo Callejo Fernández
ES30029 Final Year Research Project
BSc (Hons) Economics
April 2019
Abstract: Raising over 13.7 billion dollars in the first five months of 2018, Initial Coin
Offerings (ICOs) have emerged as a new instrument to finance projects without the need
of intermediaries. However, the lack of regulation in most countries has facilitated the
appearance of information asymmetry, which could be causing a positive bubble in the
ICO market. This dissertation empirically analyses what signals are the most adequate to
reduce information asymmetry in ICOs. In particular, I empirically asses the determinants
factors that could have reduced information asymmetry in 132 ICOs conducted between
the 5th of June 2018 and the 31st of January 2019. The results indicate that although the
number of social media followers and the rating given by ICO experts could be useful
signals, regulation is crucial to prevent information asymmetry in this market.
2
Table of contents
1. Introduction.................................................................................................................. 3
2. Overview of Initial Coin Offerings and empirical literature review ...................... 4
2.1 Blockchain ............................................................................................................... 4
2.2 Initial Coin Offerings and tokens ............................................................................. 4
2.3 Initial Coin Offerings and other financial instruments ............................................ 6
2.4 ICO legislation ......................................................................................................... 7
2.5 Empirical literature review....................................................................................... 8
3. Information asymmetry and signalling theory in the ICO market ......................... 9
3.1 Information asymmetry ............................................................................................ 9
3.2 Signalling theory .................................................................................................... 11
4. Data and Methodology .............................................................................................. 13
4.1 Data ........................................................................................................................ 13
4.2 Methodology .......................................................................................................... 14
4.2.1 Definition of success ....................................................................................... 14
4.2.2 Explanatory variables ...................................................................................... 15
5. Results ......................................................................................................................... 18
5.1 Amount raised with respect to the objective of the ICO ........................................ 18
5.2 Success variable ..................................................................................................... 20
6. Conclusion .................................................................................................................. 21
7. Bibliography ............................................................................................................... 23
8. Appendix ..................................................................................................................... 26
3
1. Introduction
Initial Coin Offerings (ICOs) appeared in 2015 as a novel mechanism based on blockchain
technology for financing entrepreneurial ventures without the need of intermediaries.
Shortly, an ICO can be defined as a financial operation through which the promoter of a
project emits to the market a new form of financial instrument, denominated token, based
on blockchain technology. Since the second half of 2017, the ICO market has increased
substantially; from raising 90 million dollars in 2016 to obtaining more than 13.7 billion
dollars in the first five months of 2018 (Diemers et al, 2018). In fact, ICOs have by far
surpassed traditional venture capitalist funding for blockchain based companies (Hacker
& Thomale, 2018), and although to date ICOs have been essentially used by start-ups
1
, its
believed that its use will be extended to more consolidated companies (Ionov, 2018).
On the other hand, the absence of a legal framework for ICOs in most countries and the
uniqueness of this market makes it a favourable scenario for the appearance of
asymmetrical information. Adopting Akerlof´s (1970) adverse selection model of
asymmetric information and using Spence´s (1973) signalling theory model as a possible
solution, this dissertation seeks to empirically examine how to reduce information
asymmetry in the market of ICOs.
My hypothesis is that information asymmetry negatively affects the chances of ICOs
success. Therefore, following previous research papers in this area, this dissertation
assesses the determinants factors of ICO success in 132 ICOs conducted between the 5th
of June 2018 and the 31st of January 2019. With the aim of doing so, two multivariate
regression analyses have been implemented, estimating the effects of eight independent
variables
2
on a dependent variable that measures the success of ICOs. The first regression
analysis is a linear multivariate regression evaluating the effect of the explanatory
variables on the amount raised by an ICO with respect to its objective. The second is a
binary logistic regression testing the relationship between the explanatory variables and a
dichotomous variable that considers whether an ICO has been successful when the
percentage with respect to the objective is above the average.
Given that ICOs are a relatively new topic, the first part of this dissertation gives an
overview of what is blockchain, a further description of ICOs and tokens, including a
comparison of ICOs and other financial instruments, an insight of the actual regulatory
framework, and an empirical literature review. The second part begins by linking
Akerlof´s (1970) theory of information asymmetry to the current state of the ICO market
and introducing Spence´s (1973) signalling theory model as a possible solution, linking it
further to Kirmani & Rao´s (2000) signalling theory model. The third section is concerned
with the data and the methodology used for the empirical analysis, and the fourth part
inquiries into the results of the regression analyses.
1
Start-up: A newly established business (Oxford dictionary, 2019).
2
These eight independent variables have been chosen in order to analyze what could be the factors used as
quality signals to reduce the perception of information asymmetry and affect the success of ICOs.
4
2. Overview of Initial Coin Offerings and empirical literature review
2.1 Blockchain
Initial Coin Offerings (ICOs) have emerged as powerful platforms for financing new
entrepreneurial ventures. Following its name, the term Coin in this context refers to tokens
based on blockchain technology, while the terms Initial and Offer indicate the first offer
of those tokens to the public.
However, this definition is meaningless without defining what is meant by blockchain
technology. In broad terms, blockchain can be defined as an interactive spreadsheet
distributed between various participants that records every transaction (Swan, 2015). The
term block refers to every individual record; each block contains unique information about
a transaction, such as the date, the amount, and information about who is taking part in
the transaction. Consequently, these blocks are linked together, forming a chain, and every
time a block is added to the chain, it is linked to the previous block using a cryptographic
hash
3
generated from the previous block. This makes the chain extremely difficult to break
since to alter past transactions, all subsequent blocks must be altered first. Transactions
are validated by numerous computers on the Internet, creating what is known as a peer-
to-peer network. These computers on the network work together following certain
assumptions to reach a consensus and ensure that each transaction is validated before
adding it to the chain (Tapscott & Tapscott, 2016; Swan, 2015).
Blockchain technology can play an important role in addressing the problem of trust as it
eliminates the need of trusting other agents or intermediaries in a transaction, although the
user has to trust the system (Swan, 2015). As the demand for transparency from society
in governmental and financial areas is growing, blockchain allows the development of a
solid set of tools that meet these demands. In fact, there is a growing body of literature
that recognises the importance that blockchain can have not just only in economics, but in
society. The decentralization power of blockchain can be used to register not only money,
stock, or financial assets, it can also serve to register every type of contract and property,
from cars and houses, to public documents such as birth and marriage certificates without
intermediaries (Swan, 2015).
2.2 Initial Coin Offerings and tokens
An inherent problem that entrepreneurs face at the beginning of their projects is the
difficulty to attract outside capital (Schwienbacher & Larralde, 2012). For Initial Public
Offerings (IPOs) there is a legal minimal threshold to be placed on the stock market, and
business angels or venture capitalists (VCs) are out of reach for most of small companies.
Crowdfunding initiatives have appeared recently as an alternative financing instrument
3
A hash is a funciont that converts an input of letters and numbers into an encrypted compressed string of
alphanumeric characters that cannot be back-computed into the original content (Swan, 2015).
5
where a project can be financed by small investments made by people across the world.
According to Schwienbacher & Larralde (2012), crowdfunding occurs without any
intermediary; entrepreneurs raise the money directly from individuals communicating
with them through the Internet. However, these authors overlook the fact that all
crowdfunding projects need a centralized service such as Kickstarter or Indiegogo and a
payment method that ultimately involve banks
4
. Although crowdfunding third parties do
not have any control over the decisions made by entrepreneurs, the intermediary might be
the ultimate arbiter of everything, including who owns what (Tapscott & Tapscott, 2016).
Rohr & Wright (2017) compare the impact that the Internet had during the early 2000s, to
the effect that blockchain could have in finance, allowing to disseminate and trade assets
in ways that were not available before. The combination of the distributive power of the
Internet with the power of blockchain is democratising the access to capital and promoting
entrepreneurship. Accordingly, the main advantage of blockchain in finance is that it can
be used to automate the procedures of raising money (Conley, 2017). New companies
have started to create Initial Coin Offerings (ICOs) instead of going to the expense of an
Initial Public Offering (IPO) or the inconvenience of convincing a venture capitalist or a
business angel to back the company. ICOs can be viewed as a crowdfunding platform but
in which blockchain technology eliminates the need for an intermediary, including banks.
Furthermore, it is relatively common to encounter the indiscriminate use of the terms
bitcoin, cryptocurrency, token, or blockchain as if they were synonyms (Ionov, 2018).
Firstly, it is important to distinguish between cryptocurrencies and tokens. The purpose of
cryptocurrencies is to operate as a normal currency, that is, to serve as a unit of account,
store of value, and medium of exchange. Bitcoin was the first decentralized peer-to-peer
payment network, and hence, the first decentralized cryptocurrency. Bitcoin was only
constructed as a currency system and was not supposed to be a development platform for
other cryptocurrencies. However, by changing the underlying codes of the Bitcoin
platform, developers started to create new cryptocurrencies using Bitcoin´s open-sourced
protocol. Additionally, developers also started to generate their own cryptocurrencies by
creating their own protocols (Swan, 2015; Conley, 2017; Rohr & Wright, 2017).
However, as previously stated, the term coin in the context of ICOs refers to tokens and
not to cryptocurrencies. Accordingly, tokens have more uses than the mere deposit and
representation of value thanks to the flexibility of their configuration, and, moreover, the
creation of tokens is much simpler than the creation of cryptocurrencies as they do not
require code modification (Ionov, 2018). Therefore, tokens are usually built on a pre-
existing blockchain, allowing users to easily produce new tokens with liberty. Indeed,
most ICOs use the ethereum blockchain as it contains useful tools to help developers
create their own coin.
4
Kickstarter and Indiegogo are the two most important crowdfunding platforms. Both platforms apply
between 3% and 9% of the total amount raised by project (Kickstarter.com, 2019; Indiegogo.com, 2019).
6
With regard to the tokens offered in an ICO, there are different types, and they often
represent different rights. The main distinction among types of tokens is the differentiation
between utility tokens and security tokens. Utility tokens are the most popular type of
tokens. This type of token gives holders certain rights to use a certain technology or
product in the future, so they enable future access and usage of the products announced
by the company (Rohr & Wright, 2017). A user who has bought these tokens before the
platform was created can use them to enjoy the final product of the ICO, or sell these
tokens at a profit at token exchanges subsequent to the ICO.
The other type of token is called security tokens, which are similar to a normal asset
offered in an IPO. These tokens promise investors positive future cash flows in the form
of dividends or a fixed commission, and they can give rights to the token holders to
participate in the company decisions (Zetzsche et al, 2018). Security tokens can also be
traded in token exchanges markets, therefore, it is important to highlight that regardless
of its type (utility or equity), all ICO tokens can be traded in a secondary market after the
end of the ICO (Ionov, 2018).
The information about what types of tokens are sold during the ICO as well as a simple
description of the project is included in the ICO whitepaper. Given that by definition ICOs
do not need any third party, companies issuing money via ICOs avoid investment banking
and official securities exchanges, and the only document they usually issue is its
whitepaper. The whitepaper includes fundamental information for investors such as the
rights that the token include, and how many tokens is the start-up planning to issue.
2.3 Initial Coin Offerings and other financial instruments
ICOs share with crowdfunding platforms the fact that they constitute an element of social
validation of an idea in very early stages (Ionov, 2018). It is alleged that if the market
believes that the idea is good and necessary, investors will reward the project promoter
with his vote of confidence providing funding. However, from a theoretical point of view,
ICOs were born with the vocation of being an alternative for private equity and Initial
Public Offerings (IPOs).
Nonetheless, the main difference between private equity and ICOs is the target of the
investment. While in private equity operations the target is the ownership of the promoting
entity, in ICOs private investors will normally acquire digital assets that normally will not
represent a participation in the capital of the promoting entity
5
. Secondly, private equity
usually presupposes that investors analyse each investment opportunity profoundly by due
diligence
6
(Ionov, 2018), however, ICO investors do not usually go beyond the
information available on the project´s website, their social networks or the public opinion
5
There are some tokens that give rights to participate in the company decisions.
6
Due diligence refers to the steps private equity investors usually take before entering into a contract or
making a business decision.
7
available within Internet forums before investing in an ICO. Lastly, and given its
terminology, ICOs pretend to assimilate to IPOs. Although ICOs copy some of IPOs
practices, the main difference resides in the lack of regulation of ICOs. While for IPOs
there exists a strict administrative control in order to avoid irregularities, ICOs are not yet
supervised by any regulator in most countries (Ionov, 2018).
According to Rohr & Wright (2017), ICOs have surpassed by far traditional Venture
Capitalist´s funding for blockchain-based companies. Comparing the amount raised by
ICOs in the last year with the amount raised by the two main crowdfunding platforms, the
difference is even greater. According to Cointelegraph.com (2019), the total amount raised
by ICOs in 2018 was 11.4 billion dollars, while the total amount raised by the two main
crowdfunding platforms from their creation to 2018, was roughly 5.5 billion dollars in
total (Kickstarter, 2019; Indiegogo, 2019).
2.4 ICO legislation
ICOs appeared in 2015 as alternative financial instruments, however, it was not until the
introduction by some countries of the first regulatory measures in 2017 when they became
popular (Ionov, 2018). The first report about ICO regulation was issued in July 2017 by
the United States Securities and Exchange Commission (SEC). This report included ICOs
under the Securities Exchange Act of 1934, establishing that those who offer and sell
securities in the U.S were required to acquiesce the federal securities laws, regardless of
whether those securities were purchased with virtual currencies or created using
blockchain technology (SEC, 2017). A similar approach was taken in Singapore, where
the Monetary Authority of Singapore clarified that the offer of digital tokens must comply
with the applicable regulation. Moreover, China totally banned ICOs in September 2017
by order of the People´s Bank of China, reporting that 90% of the ICOs launched in the
country were fraudulent, and requesting ICO operators to return the money to investors.
In addition, the European Securities and Markets Authority (ESMA) provided its first
report on ICOs in November 2017, establishing that depending on the structure and
characteristics of the tokens, those tokens assorted as financial instruments are under the
pertinent European Union (EU) regulatory framework (ESMA, 2017). Furthermore, a
similar approach was taken by the United Kingdom (UK) Financial Conduct Authority,
which highlighted their speculative nature and that whether or not are regulated would be
decided on a case by case basis. However, there is still no regulatory framework to apply
the pertinent regulation either in the UK or the European Union and hence, ICOs remain
unregulated. Furthermore, the initial approach to ICOs in most jurisdictions of the EU was
to remain silent (Zetzsche, 2017). Accordingly, in most Europe countries (Germany,
Hungary, Croatia, Poland, Austria etc.) ICOs are still unregulated.
On the other hand, the only European country that has a transparent ICO regulatory
framework is Switzerland, where the Swiss Financial Market Supervisory Authority
(FINMA) published in February 2018 a regulatory guidance for ICOs (FINMA, 2018).
8
According to this report, each ICO has to be analysed case by case, and, furthermore, an
ICO issuer has to write a formal request with all the required information to FINMA, and
wait for a response before proceeding with the ICO.
2.5 Empirical literature review
There is a relatively small body of empirical literature that is concerned with ICOs, and
those papers that have empirically analysed ICOs have mostly only focused on which
factors affect ICO success (Fisch, 2018; Burns & Moro, 2018; Yadav, 2017; Amsden
Schweizer, 2018; Benedetti & Kostovetsky, 2018). However, to date, there is no
consensus on which is the most adequate approach to measure success in ICOs. Burns &
Moro (2018) used the four-month return on investment, the first-day returns and the total
amount raised as measures of success. This third measure was also the one chosen by
Fisch (2018) and Yadav (2017), while Amsden & Schweizer (2018) argue that the
strongest measure of success is whether the token is subsequently listed on an exchange
and trades actively.
Fisch (2018) assessed the factors that determine the amount of funding raised in ICOs by
exploring three indicators as potential signals of technological capabilities: patents,
technical white papers and high-quality source code. Accordingly, Fisch (2018) collected
data from February to August 2018 on 423 ICOs concluding that technical white papers,
source code quality, token supply, Twitter activity and the token being built in the
ethereum blockchain are crucial factors for the amount of funding raised. Furthermore,
the results of this paper suggest that those ventures with high technological capabilities
should be able to signal them to investors, as ICO investors assess them to infer the quality
and invest accordingly.
Additionally, Burns & Moro (2018) used signalling theory, human capital theory and
investor sentiment as the main factors influencing success. Correspondingly, they
explored 146 successful ICOs from June 2017 to November 2017 by analysing how the
role of ICO characteristics, team quality and market sentiment behind each ICO affect the
short-term return on investment. They concluded that the initial price of the token in the
ICO is negatively correlated to the four-month return on investment and first day’s returns,
and, noticeably, a high number of information about the ICO released prior to the ICO,
negatively affects the return on investment. Furthermore, Amsden & Schweizer (2018)
assessed how venture uncertainty, venture quality and the investor opportunity-set affect
ICO success. In line with previous studies, they found that being absent on social media
and short whitepapers negatively affect the chances of success, while a larger team size is
positively correlated with ICO success. In addition, Yadav (2017) identified the
whitepaper and the local government’s attitude towards blockchain technology as key
determinants in the total amount raised.
On the other hand, Benedetti & Kostovetsky (2018) explored the determinants of ICO
under-pricing on executed and planned ICOs, concluding that tokens are sold in ICOs at
9
a significant discount to their market price. This research found that the average returns
from the ICO price to the first market price were around 179% higher. Furthermore, it is
suggested in this paper that the high number of scams in the ICO market is not important
for investors in terms of stolen capital, because of the abnormally high overall returns of
this market. Additionally, they found empirical evidence of a strong positive relationship
between the market capitalisation of an ICO and the number of Twitter followers.
Collectively, most of these studies outline the critical role of social media and whitepapers
in the success of ICOs (Fisch, 2018; Amsden & Schweizer; Benedetti & Kostovetsky,
2018). Considering that these two factors could be used to reduce information asymmetry,
there seems to be some evidence to indicate that information asymmetry affects the
success of ICOs. However, most of the aforementioned empirical studies remain narrow
in dealing with how information asymmetry can be affecting the market of ICOs and what
can be done to reduce it.
3. Information asymmetry and signalling theory in the ICO market
3.1 Information asymmetry
Initially, as it happens with IPOs (Cohen & Dean, 2005), ICO founders and ICO investors
have unequal information about all the details of the project. However, in the ICO market,
the lack of regulation in most countries introduces greater asymmetry as there is no
institution that, at least, certifies that the characteristics of the project are legitimate. This
lack of regulation leads to a market with a high degree of information asymmetry.
Akerlof (1970) provided in his seminal article an adverse selection model to account for
the impacts of information asymmetry on a market using a straightforward example
related to the used car market in the United States. In his example, he differentiates
between good cars and bad cars, referred in his work as lemons. Buyers cannot
differentiate between lemons and good cars, but sellers can, as they have more information
than buyers. However, as buyers do not trust sellers completely, they are only willing to
pay an average price that lies between the price of lemons and the price of good cars.
Nonetheless, sellers of good cars will not see profitable to sell good cars and they will
leave the market. Accordingly, the quality of the cars will fall together with the average
price, and equilibrium is not reached as buyers will know that all cars left in the market
are lemons, leading to a market collapse.
The ICO market has all the conditions of the car market demonstrated in Akerlof´s (1970)
theory: The only information available about an ICO is what the ICO provides either in
the whitepaper or in social media, therefore, the flow of information is only one-sided,
from sellers to buyers (Fisch, 2018). Moreover, as most ICOs are still unregulated, it
creates a high incentive for token issuers to create lemon ICOs, or scams, and inflate the
token price by providing false information in their whitepapers and social media without
being legally punished (Sehra et al, 2017).
10
Therefore, this creates a paradox. Blockchain technology is supposed to be the solution
for trust in all industries where a third party to guarantee trust is needed. However, one of
the main problems of the financing platforms based on blockchain technology might be
trust and information asymmetry. Blockchain ensures that the transaction is done correctly
and it can also serve to trace the product, or the token in the case of ICOs without the need
of intermediaries. However, once investors buy the ICO token, in most ICOs they can only
be certain that the transaction has succeeded, but they can not know if the token they have
bought is from a valuable ICO or from a scam (lemon ICO). ICO investors know that they
have certain guarantees, but not others, on which a third party might be needed.
Emery (2014) extended Akerlof (1970) adverse selection model to show how rather than
collapse, bad quality assets could lead to a positive bubble that can persist when there is
favourable but incorrect information. In the ICO market there is a high incidence of scam
and bad quality ICOs
7
(Zetsche et al, 2018). Separating the market of ICOs between
scams/bad quality ICOs and good ICOs, scam/bad quality ICO tokens could be selling
above of their fundamental value, and good ICOs tokens could be selling below their
fundamental value. Consequently, either an increase in the proportion of good ICOs or an
increase in the returns from investing in good ICOs, could have expanded demand and
price, in a comparable way to which an increase in the proportion of lemons drives down
the demand and price of used cars in Akerlof´s (1970) adverse selection model.
Furthermore, there is a great proportion of investors in the ICO market who are non-expert
investors that want to get rich expeditiously by taking advantage of this new unregulated
market, investors who do not usually look beyond the ICO whitepaper and the ICO
website (Ionov, 2018, p.41). Accordingly, a sufficiently large expected return might be
causing a positive bubble in this market, due to information asymmetry and the simplicity
of accessing it given the lack of regulation.
In agreement with this proposition, Palm & Bergesen (2018) reported that if a person
managed to buy the top ten performing ICOs in 2017, he would have had an average return
of 10,400%. Consequently, Benedetti & Kostovetsky (2018), showed how ICO tokens
record particularly high average returns up to 82% on the first day of trading, with average
returns of 48% in the first 30 trading days. The abnormally high returns indicated by these
two papers could be linked to the high-return characteristics of the ICO market, or they
could be an indication of a bubble (OECD, 2019).
7
According to Satis (2018) report, in 2017 over 80% of ICO projects were identified as scams. Although
this result can be exaggerated, as this report considered scams all ICOs that were no trading in a secondary
exchange and that they did not have the underlying code open to the public.
11
3.2 Signalling theory
Signalling theory, introduced by Spence (1973) in a job market framework, gives one
possible solution to the problem of asymmetric information. In the job market, employers
lack information about the quality of job applicants, therefore, the candidates obtain
education to signal their quality to the employers, reducing information asymmetry.
Connelly et al (2011) give a further explanation of signalling theory, providing five key
elements for signalling to exist. The two main elements are the signaler and the signal; the
signalers are insiders who have information that is not available to outsiders, and which
provides signalers with a privileged perspective of the quality of the product. The other
main element is the signal. Not all actions done by the signalers are useful as signals,
therefore, they have to satisfy two main characteristic: signal observability; outsiders must
be able to notice the signal, and signal cost, the costs associated with the signal.
The other key elements of signalling theory are the receiver, the feedback and the
environment. Receivers are outsiders who do not have the information but would like to
receive it whereas feedback is the process by which receivers send information to signalers
about the effectiveness of their signal. Lastly, the signalling environment refers to how
organizations affect the effect to which signalling reduces information asymmetry. This
last element can be a key factor in correlating the ICO market to signalling theory, as
signalling is most effective in those markets for new products or products about which
consumers are uninformed (Kirmani & Rao, 2000).
Figure 1 Modification of Figure 2 in Connelly et al (2011, p.44).
Figure 1 illustrates how signalling theory can work in the ICO market. Accordingly, the
ICO market is a new market where most investors are uninformed, and in addition, is an
unregulated market where bad practices are not legally penalised. In the case of ICOs, the
signalers would be the ICOs developers, those who know the real quality of the ICO. The
signals would be all actions that the ICO owners can implement to reduce information
asymmetry. The receivers would be the ICO investors, the feedback would be all the
negative and positive comments that ICO investors send in social media and over the
internet, and the environment would be the ICO market (Fisch, 2018). ICO developers
have to find the most appropriate signals to indicate that their ICO is not a lemon ICO and
obtain a successful ICO.
The ICO developer knows
the underlying quality of
the ICO
The ICO developer
sends the signal to the
ICO investor
The ICO investor
observes and interprets
the signal
The ICO investor sends
feedback to the ICO
developer
12
On the other hand, because of the lack of regulation in this market, ICO founders may be
motivated to attempt false signalling. Therefore, if an ICO developer does not have the
underlying quality associated with the signal, but believes that the benefits of signalling
outweigh the costs of producing the signal, this signaller will probably attempt false
signalling (Connelly et al, 2011). This can be illustrated in the ICO market using Kirmani
& Rao (2000) signalling theory model:
Figure 2 Modification of the table represented in Kirmani & Rao (2000, p. 14).
Signal
Do Not Signal
High-quality ICOs
A
B
Scam ICOs
C
D
Supposing there are two types of ICOs, high-quality ICOs, and scam ICOs. Signalling is
a viable strategy when two conditions hold: when the best strategy for high-quality ICOs
is always to signal A>B, and when for scam ICOs is better to do not signal than to signal
given the costs of signalling D>C. This scenario would be ideal since ICO investors will
distinguish through signals the scam ICOs and high-quality ICOs, and only high-quality
ICOs will succeed. However, in a scenario where the payoffs are such that both types of
ICOs would be better off signalling (A>C and C>D) ICO investors would be unable to
distinguish between scam ICOs and high-quality ICOs and it will result in a pooling
equilibrium. This equilibrium is likely to occur when the gains from falsely claiming high
quality outweigh the losses from being discovered (Kirmani & Rao, 2000).
If the payoff of signalling for scam ICOs is greater than the payoff of not signalling, scam
ICOs will try to falsely report themselves as high quality ICOs. The lack of regulation in
the ICO market can encourage this practice, given that the costs associated with reporting
false information are much lower than in traditional financial instruments. Accordingly,
regulation becomes crucial to prevent the situation where C>D, if it can guarantee through
inspection mechanisms that the information issued by ICOs is legitimate. The existence
of a pooling equilibrium can dissuade signals of reducing information asymmetry and
provide more weight to the argument that the ICO market could be in a positive bubble.
The following part of this dissertation moves on to empirically analyse which signals
could reduce the perception of information asymmetry in the ICO market. My hypothesis
is the following: the lower perception of information asymmetry, the higher the chances
of an ICO being successful. Therefore, in my analysis, I empirically assess the factors that
could reduce the perception of information asymmetry and, hence, facilitate the success
of ICOs.
13
4. Data and Methodology
4.1 Data
I collected data on 132 ICOs that were completed between the 5th of June 2018 and the
31st of January 2019. This starting point was chosen considering the new report published
by the US SEC on the 4th of June (SEC, 2018), where they appointed a new chief with a
view to improving ICOs regulatory framework.
ICOs are a new financial instrument, and data on their processes and results are lacking to
a large extent. No complete database exists with clear information about ICOs. Most of
the time the only information available is the data provided by the ICOs themselves, in
particular the data that one can find on the ICO´s webpage or in the ICO´s whitepaper. In
the following analysis it was therefore necessary to carry out a parsimonious process of
data collection, selection and cleaning. Following Benedetti (2018) and Burns & Moro
(2018), the main data sources consulted were Icodata
8
and Icobench
9
for most of the
variables
10
, and Telegram.com
11
and Twitter.com
12
for the social media variables. In order
to account for discrepancies in the data and to identify missing variables; coingecko.com,
tokendata.io and icodrops.com websites were also consulted, as well as the webpages and
the whitepapers of each particular ICO
13
.
Excluded from the following analysis are those ICOs with no date information or with no
information about the amount of funds raised. Although these exclusions could entail
some degree of selection bias, especially considering that those ICOs in the website data
aggregators could be the ones that have succeeded, previous researches have followed a
similar approach (e.g. Burns & Moro, 2018). Moreover, Benedetti (2018) showed that
from the 4th quarter of 2016, 83% of ICOs were being covered by at least one of the data
aggregators, therefore, it can be presupposed that in 2018 almost all ICOs were covered
in the data aggregators used.
8
Source: https://www.icodata.io/
9
Source: https://icobench.com/
10
These variables are presented in the next section
11
Telegram.com is a messaging application that works over the internet. This app differs from its
competitors in that all messages send within the platform are encrypted (Telegram.com, 2019).
Telegram.com is the most used social media application in the ICO community.
12
Twitter.com is a social networking platform that allows registered members to broadcast short posts
(Twitter.com, 2019). Twitter.com is the most preferred social media in the ICO community to posts news
about the progress of the ICO.
13
Sources: Coingecko; https://www.coingecko.com. Tokendata; https://www.tokendata.io/. Icodrops;
https://icodrops.com/
14
4.2 Methodology
4.2.1 Definition of success
The main hypothesis of my analysis is that the perception of information asymmetry is
negatively correlated with the chances of success of ICOs. As the establishment of ICOs
is the main focus of this dissertation, the success of ICOs is the dependent variable in the
empirical analysis. Throughout this dissertation the term success is defined as the
accomplishment of an aim or purpose (Oxford dictionary, 2019).
Fisch (2018) uses the total amount raised as a measure of success, however, the absolute
amount collected is not a meaningful variable, as ICOs differ in their size and objectives.
What could be considered an acceptable amount for some ICOs, could be an undesirable
amount for others. Therefore, this dependent variable in itself fails to meet the definition
of success as it is not clear if the ICO has accomplished its aim or not. Amsden &
Schweizer (2018) suggest that the strongest measure of ICO success is whether the token
is subsequently listed on an exchange and traded actively. However, most ICOs need
between three and six months after the end of the ICO to be prepared to be listed on an
exchange. Therefore, as the sample used only includes ICOs finished in a 6-month period,
this estimator had high chances of being biased. In fact, 30.4% (see Appendix 2) of the
ICOs that were considered successful did not have their tokens in a secondary exchange.
To use a better measure to determine what can be considered success in an ICO, the
dependent variable chosen is the amount raised by an ICO with respect to its target
amount; a variable collecting in percentage how much an ICO has collected with respect
to its objective. This relative measure allows to assess to what degree an ICO has
accomplished its purpose.
Furthermore, in my analysis I build a second relative measure of success, getting
inspiration from Ahlers et al (2015). They use a binomial variable (0/1) indicating whether
a project has received the full target amount in order to measure success in crowdfunding
platforms. However, most ICOs are unlikely to reach the target amount, as they usually
stop accepting funding after they reach the maximum capital needed. Accordingly, and
following Soule (2017)
14
, I created a binary dummy variable with values 0 if the amount
raised with respect to the objective was below the average across ICOs and 1 if it was
above.
14
Soule (2017) uses a logistic regression analysis in order to analyze student success, separating
successful students and unsuccessful students arbitrary depending on their tests scores.
15
This second dependent variable was tested in a second model, namely a binary logistic
regression analysis. Equation (1) (Soule, 2017) shows the mathematical expression of the
logistic regression, which predicts the probability of Y taking a specific value (0 or 1).
The average of the amount raised with respect to the desired amount was 59% (see
Appendix 1), therefore, only ICOs above this threshold were considered successful ICOs.
Accordingly, the percentage average obtained by the successful ICOs was broadly higher
than the percentage average of the unsuccessful ICOs (94.02% vs. 21.26%) and the
percentage of tokens that were traded actively was higher in the successful ICOs (69.6%
vs. 28.6%) (see Appendix 2), demostrating the viability of this dependent variable.
4.2.2 Explanatory variables
To analyse what could be the factors used as quality signals to reduce the perception of
information asymmetry and affect the success of ICOs, eight independent variables and
one control variable were collected.
Europe (dummy variable). Nearly half of the ICOs carried out during 2017 were based
in Europe, and following this trend, 42.4% of the ICOs analysed were based in a European
country (Atomico, 2018). ICOs are still unregulated in most European countries, and, in
this sample, approximately 89.5% (see Appendix 3) of the ICOs executed in Europe were
from countries where ICOs are not regulated. Therefore, it can be seen that given this lack
of regulation, many ICO founders choose to develop their ICO in Europe given the
flexibility it bestows. As such, the variable Europe has been included in this analysis to
control for regulation differences, as ICOs launched in Europe might have a higher degree
of information asymmetry (except in the specific case of Switzerland). This variable was
included as a dummy variable with values 1 if the ICO was performed in Europe or 0 if
the ICO was executed outside.
KYC (dummy variable). In order to deal with the problems of trust and anonymity, ICOs
are increasingly introducing a system called Know Your Customer (KYC). KYC is the
process that ICOs use to verify their customers or investors. Although this process is also
being introduced in traditional finance, there is a debate in the ICO ecosystem as one of
the main reasons for investors to enter into this market is the anonymity that grants
(Benedetti & Kostovetsky, 2018). KYC has been used to fight against terrorism and
money laundering, therefore, ICOs that use this process could be given a signal of
legitimacy and absolute fulfilment of the law and anti-criminal policies.
(1)
16
ERC-20 (dummy variable). As it was stated before, most ICOs use the ethereum
blockchain as it contains useful tools to help developers to create their own coin. ERC-20
is the ethereum blockchain protocol that offers guidelines to create new tokens, providing
a standardized way to develop tokens (Tapscott & Tapscott, 2016). Developing ICOs in
the ethereum platform could have critical advantages as it may signal a higher future value
if investors assume that this standard will be established as the benchmark for ICOs (Fisch,
2018). On the other hand, the price of the tokens created with this standard could be
correlated with the ethereum price. The creation of the ERC-20 standard has been the main
driver of the quick development of ICOs in the last two years. Promoting the creation of
ICOs, and enabling the creator of the token to set a supply of tokens, keeping track of who
owns each token and facilitating the transmission of tokens between parties. Since its
adoption, ICOs funds raised growth from 90 million dollars in 2016 to 6 billion dollars in
2017 (Icodata.io, 2019). Therefore, a dummy variable taking the value 1 if the ICO was
performed in the ethereum platform, and 0 if it was not, was introduced.
Pre-ICO (dummy variable). Frequently, new ICO projects require external and
anticipated investment to cover operational and marketing costs. In order to obtain these
funds, some ICOs conduct what is known as a pre-ICO; the prior sale of the tokens to one
or several groups of investors. Commonly, this pre-sale is made at a lower price than the
initial offer and in some cases with bonuses added to the acquisition of the tokens. Pre-
ICOs are established in separate intelligent contracts and are announced by the project in
order to maintain good accounting in their cryptocurrencies. Doing a pre-ICO can be seen
as a bad signal for ICO investors, as it could mean that the ICO needs an excessive amount
of funds to develop the project. Or, on the other hand, as suggested by Fisch (2018), who
also includes the pre-ICO variable in his regression model, pre-ICOs can attract early
investors, which can be crucial for campaign success.
GitHub (dummy variable). The variable GitHub is a dummy variable that indicates if
the ICO has information on Github.com. GitHub is a website used by developers to store,
share and manage their code. In the case of ICOs, GitHub is used to enable investors to
keep track of the changes made to the code of the ICO token or the code of the product
that a particular ICO might be developing. According to Amsden & Schweizer (2018),
ICO projects without a quality code will be reluctant to establish an account on GitHub.
Therefore, being on GitHub should reduce information asymmetry.
Social media variables: Twitter & Telegram. The variables Twitter and Telegram
measure the social media success of an ICO. Given the strong community feeling in the
crypto environment, and that ICOs announce their main features mainly on social media,
this could be the main channel used by ICOs to send signals to reduce information
asymmetry, and, in addition, a high number of social media followers can increase trust
in an ICO. Furthermore, as it was pointed out in the empirical literature review, Benedetti
& Kostovetsky (2018) found that the number of Twitter followers lead to more market
capitalization, and most of the previous empirical studies use them as explanatory
17
variables (Benedetti & Kostovetsky, 2018; Fisch, 2018; Burns & Moro, 2018; Amsden &
Schweizer, 2018).
Rating. The variable rating collects the average rating given by the two main ICO rating
websites according to the literature: ICObench.com and Icodata.io. Although these ratings
have been criticized for being biased (Boreiko & Vidusso, 2018), many ICO investors
consider them as determinant factors when deciding where to invest. ICObench.com
experts give a rating from 0 to 5 based on the ICO profile, the team, the vision, and the
product of the ICO. While Icodata.io team of experts give a rating between 0 and 100
focusing on the team, the product, the market, the transparency of the ICO and the ICO’s
structure
15
. Although this variable cannot be considered a signal produced directly by the
developers of the project, it is a signal that can be followed by investors.
ICO duration. The duration of the ICO cannot be seen as a signal for investors since it
can only be known once the ICO is finished, therefore, and following what previous
researchers have done (Fisch, 2018), the duration variable has been used as a control
variable in order to prevent confounding effects. This variable indicates the duration of
the ICO, from the day that was open to the public, to the last day of the public offer. For
the purpose of this regression analysis, this variable was measured in days.
Prior studies analysing ICOs have indicated that whether the ICO has a whitepaper was
as a significant factor for the success of an ICO. However – possibly because of this
finding- all of the ICOs analysed in this dissertation had a whitepaper, therefore, this
variable was excluded from the analysis. Burns & Moro (2018), Fisch (2018) and Momtaz
(2018) used bitcoin price as a control variable. Nonetheless, for the period analysed no
correlation was found between the bitcoin price or the ethereum price and the percentage
amount raised by ICOs with respect to their objective. This lack of correlation is probably
due to the strong decay in both prices, bitcoin price dropped from $8091.83 to $3485.41
in the period covered, whereas ethereum price dropped from $672.1 to $108.9. These
results also indicate that although the extreme decay in the prices of the two main
cryptocurrencies, the ICO market has remained stable. Furthermore, although a variable
controlling for the different regulations in each country would have improved the model,
the construction of this variable was problematic given the wide diversity of regulatory
measures concerning ICOs.
15
To get an unbiased average, ICObench.com rating was multiplied by 20 in order to operate on the same
scale.
18
5. Results
I present here the empirical results in two subsections. The first contains the results of the
linear regression model analysing the effects of the independent variables on the
dependent variable that reflects the amount raised by ICOs with respect to its objective.
The second subsection presents the results of the logistic regression, which is based on the
binomial variable defined on the amount raised being above or below the average across
ICOs.
5.1 Amount raised with respect to the objective of the ICO
Table 1 OLS multiple regression analysis results on the dependent variable: percentage of the amount
collected with respect to the objective of the ICO
Variable
Coefficient
Std. Error
t-Statistic
Prob.
Constant
0.195319
0.134075
1.456792
0.1477
Europe
-0.125107
0.058172
-2.150618
0.0335
KYC
0.084082
0.072080
1.166505
0.2457
ERC-20
0.080396
0.061944
1.297895
0.1968
Pre-ICO
-0.036236
0.055682
-0.650772
0.5164
GitHub
0.014894
0.056667
0.262836
0.7931
Twitter
0.004437
0.002047
2.167149
0.0322
Telegram
0.004174
0.002023
2.063656
0.0412
Rating
0.004742
0.001975
2.400969
0.0179
Duration
-0.002894
0.000713
-4.059713
0.0001
R-squared
0.439462
Adjusted R-squared
0.398110
Observations
132
In order to measure which ICO characteristics affect the percentage of accomplishment, a
least squares regression was carried out. Table 1 represents the regression results; the
adjusted R-squared value of 0.439 indicates that the explanatory variables in this
regression model explain roughly 44% of the variance in the amount raised by ICOs with
respect to their objective. Interestingly, five out of the eight independent variables
analysed have a p-value below 0.05. Using the Breusch-Pagan test to check for the
presence of heteroscedasticity, the null hypothesis cannot be rejected (p-value of 0.86),
suggesting that there is no evidence of heteroscedasticity (see Appendix 5).
Doing an ICO in Europe has a high significant (p-value < 0.05) strongly negative effect
of -12.5%. From the 132 ICOs analysed in this sample, 89.5% of the ICOs held in Europe
did not have to conform to any regulatory framework, while 74.7% of the ICOs executed
outside Europe were performed within the bounds of a regulatory framework (see
Appendix 3). Therefore, this result suggests that doing an ICO outside Europe can give a
positive signal to ICO investors, as investors might identify the lack of regulation as a
negative signal to invest in ICOs. As it was previously stated, the lack of regulation also
entails a lack of responsibilities, suggesting that those ICOs developed in Europe tend to
be less prepared, or that investors might not have as much confidence in those ICOs.
19
Accordingly, the nonexistence of a regulatory framework
16
and the possible existence of
a higher degree of information asymmetry, might cause a negative effect in the success of
ICOs.
It is not surprising that three of the variables that were supposed to reduce information
asymmetry (the social media variables and the rating variable) are statistically significant
at the 5% level (p-value < 0.05). The results indicate that for every extra 1,000 followers
in Twitter or Telegram, the ratio of success increase by 0.4%
17
. These results match those
observed in previous studies (Burns & Moro, 2018; Benedetti & Kostovetsky, 2018;
Fisch, 2018; Amsden & Schweizer) and confirm the importance of social media in ICOs
fundraising. The number of followers that an ICO has on social media can be used by ICO
investors as a source of trust, as a higher number of followers can be perceived by
investors as a good signal to invest in an ICO. Moreover, ICOs use social networks as the
main channel to signal the quality of their product and to signal new information about
the ICO. The significant positive result of these variables indicates that as the number of
people reached with social media increases, the perception of information asymmetry
decreases, and the probability of success improves.
The positive coefficient of the variable rating, indicates the positive effect of expertise
rating in ICOs fundraising; for every extra point given by the two main ICO rating
websites, the success ratio increases about 0.47%. Although these ratings have been
criticised within the ICO community of being biased (Boreiko & Vidusso, 2018), the
positive coefficient and the significance of this variable shows that investors tend to invest
in those ICOs with higher ratings. In addition, it is important to highlight that these ratings
are usually assigned before the ICO has been carried out. Therefore, although the rating
given by ICO rating websites cannot be considered a signal produced directly by the
developers of the project, it indicates that investors use ratings as a source of trust. On the
other hand, this result is contradictory with blockchain´s logic of trust. ICOs were
supposedly established as financial platforms that do not need third parties, however it
seems that investors make use of these third parties (rating score websites) in practice.
Moreover, the control variable of the duration of the ICO is highly significant (p-value <
0.01), and negative, indicating that for each extra day that the ICO lasts, the percentage
amount collected with respect to the objective is reduced, in line with previous research
studies (Fisch, 2019). ICOs that do not reach its goal amount try to extend the ICO period
as much as possible, as well as they usually stop the fundraising period once they have
accomplished their goal.
16
The only European country with ICO regulation is Switzerland. However, in this sample, there were only
six ICOs that were carried out in Switzerland, and 5 out of these 6 ICOs were successful ICOs, supporting
the argument that the lack of regulation can be a barrier for ICO success.
17
The social media variables have been divided by 1,000 in order to account for the effect of 1,000
followers instead of 1.
20
5.2 Success variable
Table 2 Binary logit multiple regression analysis results on the dummy dependent variable success.
Method used: ml - binary logit (Newton-Raphson / marquardt steps).
Variable
Coefficient
Std. Error
z-Statistic
Prob.
Constant
-2.256768
1.112741
-2.028116
0.0425
Europe
-1.090466
0.495865
-2.199118
0.0279
KYC
0.471943
0.661572
0.713367
0.4756
ERC-20
1.035935
0.562561
1.841462
0.0656
Pre-ICO
-0.090884
0.492949
-0.184369
0.8537
GitHub
-0.243363
0.513567
-0.473868
0.6356
Twitter
0.062395
0.026003
2.399484
0.0164
Telegram
0.032718
0.018766
1.743524
0.0812
Rating
0.026865
0.017096
1.571440
0.1161
Duration
-0.033281
0.009924
-3.353437
0.0008
McFadden R-squared
0.380807
Observations
132
As it was stated before, a binary logistic regression was performed in an explanatory way,
taking as dependent variable the dichotomous variable success, constructed from an
arbitrary decision that considers an ICO has been successful (=1) when the percentage
with respect to the objective is above the average 59%, and failure (=0), when it is below
the average. The McFadden R-squared of 0.380 suggests an excellent fit, as those values
between zero and four represent a perfect model fit (McFadden, 1978). Checking for
multicollinearity, the maximum correlation was 0.53 between Telegram and Twitter (see
Appendix 7), however, only correlation coefficients above 0.90 should be considered,
suggesting no multicollinearity (Tabachnick & Fidell, 2013).
The results of this regression model (Table 2) are very similar to the previous regression
analysis results. The variables length and Europe are still significant at the 5% level, both
with negative coefficients, while the variables Telegram and rating are not significant
anymore at the 5% level (although Telegram is at the 10%).
However, the variable ERC-20 is now significant at the 10% (p-value of 0.065). This p-
value and the high coefficient (1.036) suggest that those ICOs developed in the ethereum
platform have more chances of being successful. This result is consistent with what could
have been presupposed. The adoption of a standardized way to developed tokens was one
of the main causes of the expansion of ICOs in 2017 as it attracted to the market a
considerable amount of ICO founders that were unsure on how to create a token on
blockchain technology. The significant positive value of this variable suggests that the
adoption of a standard can reduce information asymmetry, as this standard may signal a
higher future value if investors assume that it will be established as a benchmark for ICOs
(Fisch, 2018).
21
Moreover, the variables KYC, pre-ICO and GitHub were not significant in any of the
multivariate regression models. This is a remarkable result as KYC was assumed to be a
key factor in dealing with trust, and GitHub was presupposed to be one of the main
channels to signal the quality of ICOs. However, these results are in line with previous
studies, which also found these variables insignificant (Fisch, 2018; Burns & Moro, 2018).
A possible explanation for the insignificance of these signals could be that the lack of
regulation and the abnormally high returns of this market are attracting new inexperienced
ICO investors who fail to observe these signals.
6. Conclusion
The aim of this dissertation was to empirically examine how to reduce information
asymmetry in the market of ICOs. My hypothesis was that information asymmetry is one
of the main barriers of ICO success, considering success as the percentage of the amount
raised by an ICO with respect to its objective. The empirical analysis results provide
further support for this hypothesis. Generally, ICO investors do not usually go beyond the
information available on the project´s website, the social media networks of the ICO or
the online community opinion. Therefore, these were supposed to be the main sources
used for signalling by ICO developers to reduce information asymmetry and positively
affect the likelihood of success. Accordingly, the social media explanatory variables were
highly significant in both regression models, indicating that ICOs use them as a way to
signal information to their potential investors and that investors use these signals as a
source of trust. And the explanatory variable rating was also positive and significant in
one of the multivariate regression models, pointing out that rating websites
(ICObench.com and ICOdata.io in this sample) are also used by investors as a way of
receiving signals about the quality of the ICOs, as investors usually refuse to invest in
those ICOs with low rating scores.
On the other hand, I suggested using Kirmani & Rao (2000) signalling theory model that
these aforementioned signals could not be of any use in reducing information asymmetry
if the cost of false signalling is lower than its potential benefits. Correspondingly, ICO
rating websites could be biased (Boreiko & Vidusso, 2018), as some ICOs could pay to
get a higher rating falsely reporting its quality. Moreover, although obtaining a significant
number of social media followers entails a cost of time and effort, the payoff of reporting
false information through social networks can be greater than its cost given that ICOs are
not legally penalised. Without any regulation mechanism that could increase the costs of
false signalling, it could be more profitable for scam ICO founders to send false signals,
which in the end, could lead to a pooling equilibrium that could be causing a positive
bubble in the market of ICOs.
Considering this possible outcome, the best solution to reduce information asymmetry in
the ICO market is to introduce an adequate regulation framework. It was not until the
introduction of the first regulatory measures in the second half of 2017, when the ICO
market became popular. The absence of a regulatory framework causes a high degree of
22
information asymmetry in this market, which could be producing insecurity among ICO
investors and could be a limiting factor to the potential of ICOs as an alternative financial
instrument. Although further research might be required when accounting for the different
ICO regulations in each country, the significance negative value of the variable Europe in
the empirical regression analyses suggests that ICO investors penalise the lack of
regulation and the unclear position of the legal authorities with respect to ICOs.
The challenge now is how to apply the adequate regulation when it is in the nature of ICOs
to not depend on third parties. The most problematic part of regulating ICOs is to
determine the legal nature of tokens, given that they can take multiple legal forms. To
date, there are three main regulatory approaches: the total prohibition of ICOs, remain
silent, or classifying each ICO token one by one, and then applying its corresponding
legislation. This last approach is the only one that might guarantee a future for ICOs,
causing an interesting paradox, since the only way that could guarantee a future for ICOs
is to introduce a third party.
23
7. Bibliography
Akerlof G.A., 1970. The Market for Lemons: Quality Uncertainty and the Market
Mechanism. Quarterly Journal of Economics, Vol. 84, pp. 488-500.
Akerlof, G.A., 1984. An Economic Theorist’s Book of Tales. Cambridge: Cambridge
University Press.
Alexandre, A., 2018. New Study Says 80 Percent of ICOs Conducted in 2017 Were Scams
[Online]. Cointelegraph, July 13. Available from: https://cointelegraph.com/news/new-
study-says-80-percent-of-icos-conducted-in-2017-were-scams. [Accessed
Amsden, R. and Schweizer, D., 2018. Are Blockchain Crowdsales the New 'Gold Rush'?
Success Determinants of Initial Coin Offerings. SSRN. April 16. Available from:
http://dx.doi.org/10.2139/ssrn.3163849.
Atomico, 2018. The State of European Tech [Online]. Available from:
https://2017.stateofeuropeantech.com/. [Accessed 12 January 2019].
Benedetti, H. and Kostovetsky, L., 2018. Digital Tulips? Returns to Investors in Initial
Coin Offerings. SSRN. Available from: https://cdn.crowdfundinsider.com/wp-
content/uploads/2018/07/SSRN-id3182169.pdf.
Boreiko, D. and Sahdev, N., 2018. To ICO or not to ICO – Empirical analysis of Initial
Coin Offerings and Token Sales. SSRN. July 6. Available from:
http://dx.doi.org/10.2139/ssrn.3209180.
Boreiko, D. and Vidusso, G., 2018. New Blockchain Intermediaries: Do ICO Rating
Websites Do Their Job Well?. SSRN. Available from:
https://papers.ssrn.com/abstract=3263393.
Burniske C. and Tatar J., 2018. Cryptoassets: The Innovative Investor´s Guide to Bitcoin
and Beyond. New York: McGraw-Hill Education eBooks.
Burns, L., and Moro, A, 2018. What Makes an ICO Successful? An Investigation of the
Role of ICO Characteristics, Team Quality and Market Sentiment. SSRN, September,
DOI: 10.2139/ssrn.3256512.
Catalini C, Gans J, 2018. Initial Coin Offerings and the Value of Crypto Tokens. Working
Paper, 5347-18, MIT Sloan School of Management. Available from:
http://ide.mit.edu/sites/default/files/publications/SSRN-id3137213.pdf.
Chiyachantana, C., Taechapiroontong, N., Jiang, C. and Wood, R., 2004. The impact of
Regulation Fair Disclosure on information asymmetry and trading: an intraday analysis.
The Financial Review 39, pp. 549–577.
Chod, J. and Lyandres E., 2018. A Theory of ICOs: Diversification, Agency, and
Information Asymmetry. Carroll School of Management. Boston College. Available
from: http://bschool.huji.ac.il/.upload/Seminars/ICO%202018_05_10.pdf.
Clayton, J., 2017. Statement on Cryptocurrencies and Initial Coin Offerings. United States
Securities and Exchange Commission (SEC). Dec. 11. Available from:
https://www.sec.gov/news/public-statement/statement-clayton-2017-12-11.
24
Cohen, B.D. and Dean, T.J., 2005. Information asymmetry and investor valuation of IPOs:
top management team legitimacy as a capital market signal. Strategic Management
Journal, 26: pp. 683-690.
Conley P, 2017. Blockchain and the Economics of Crypto-tokens and Initial Coin
Offerings, Nashville: Vanderbilt University Department of Economics Working Papers
17-00008.
Connelly, B.L., Certo, S.T., Ireland, R.D. and Reutzel, Ch., 2011. Signaling Theory: a
Review and Assessment. Journal of Management, 37 (1), January, pp. 39-67.
Diemers D, Arslanian H, McNamara G, Dobrauz G, Wohlgemuth L. 2018. Initial Coin
Offerings, A strategic perspective. PWC.
Dowlat S. and Hodapp M, 2018. Cryptoasset market coverage initiation: valuation. Satis
Group. August 30. Avalaible from: https://research.bloomberg.com/pub/res/d37g13t-
y4Z-zRdOVQOaUywhfJw.
Emery D., 2014. Negative Bubbles and the Market for 'Dream': 'Lemons' in the Looking
Glass. University of Miami, Department of Finance: European Securities and Markets,
October 27. Available at: SSRN: http://dx.doi.org/10.2139/ssrn.2515503.
European Securities and Markets Authority, 2018. ESMA alerts firms involve in Initial
Coin Offerings (ICOs) to the need to meet relevant regulatory requirements. Nov. 13th.
Available from: https://www.esma.europa.eu/document/esma-alerts-firms-involved-in-
initial-coin-offerings-icos-need-meet-relevant-regulatory
FINMA, 2018. New recognition obligation for foreign trading venues for the trading of
Swiss equity securities. Switzerland. FINMA Guidance, Nov. 18. Available from:
https://www.finma.ch/en/news/2018/11/20181130-aktuell-anerkennung-handelsplaetze/
Fisch, C. (2018). Initial Coin Offerings (ICOs) to Finance New Ventures: An
Exploratory Study. SSRN. Available from: https://papers.ssrn.com/sol3/
papers.cfm?abstract_id=3147521
Flühmann, D. and Seiler, P., 2018. ICO financing for startups. International Briefings.
International Financial Law Review, April, pp.11-12.
Hacker, P. and Thomale, C., 2018. Crypto-Securities Regulation: ICOs, Token Sales and
Cryptocurrencies under EU Financial Law. Faculty of Law. University of Oxford.
Available from: https://www.law.ox.ac.uk/business-law-blog/blog/2018/01/crypto-
securities-regulation-icos-token-sales-and-cryptocurrencies.
Ionov, R.C., 2018. Las initial coin offerings (ICOS) y la tokenización de la economía.
Pamplona: Thomson Reuters Aranzadi.
Kirmani, A. and Rao, A.R., 2000. No Pain, No Gain: A Critical Review of the Literature
on Signaling Unobservable Product Quality. Journal of Marketing, 64(2), pp. 66–79.
McFadden, D., 1978. Quantitative Methods for Analyzing Travel Behaviour of
Individuals: Some Recent Developments. In D. Hensher and P. Stopher, eds. Behavioural
Travel Modelling. London: Croom Helm, pp. 279–318.
Metrick, A. and Yasuda, A., 2010. Venture Capital and the finance of innovation.
Hoboken, NJ: John Wiley and Sons.
25
Monetary Authority of Singapore, 2018. A guide to digital token offerings. Monetary
Authority of Singapore (MAS).
OECD, 2019. Initial Coin Offerings (ICOs) for SME Financing. OECD.
Philbeck, T. and Davis, N., 2018. The Fourth Industrial Revolution: Shaping a New
Era, Journal of International Affairs, 72(1), pp. 17–22.
Rohr, J. and Wright A, 2017. Blockchain-Based Token Sales, Initial Coin Offerings, and
the Democratization of Public Capital Markets. SSRN. DOI: 10.2139/ssrn.3048104
Schöner, M., Kourouklis D., Sandner, P., Gonzalez, E. and Förster, J., 2017. Blockchain
Technology in the Pharmaceutical Industry. Frankfurt School Blockchain Centre Working
Paper, July. Available from: http://explore-ip.com/2017_Blockchain-Technology-in-the-
Pharmaceutical-Industry.pdf.
Schwienbacher, A. and Larralde, B., 2012. Crowdfunding of small entrepreneurial
ventures. In D. Cumming, ed.,The Oxford Handbook of Entrepreneurial Finance. New
York: Oxford University Press, pp. 341-365.
Sehra, A., Smith, P. and Gomes, P., 2017. Economics of Initial Coin Offerings. Allen
Overy, Nivaura.
Spence, M., (1973). Job Market Signaling. The Quarterly Journal of Economics, 87(3),
pp. 355- 371.
Swan, M., 2015. Blockchain Blueprint for a New Economy. Sebastopol: O´Reilly Media,
Inc.
Tabachnick, B. and Fidell, S., 2013. Using Multivariate Statistics. 6th ed. Boston:
Pearson.
Tapscott, D. and Tapscott, A., 2016. Blockchain Revolution. New York: Penguin Random
House.
Yadav, M., 2017. Exploring Signals for Investing in an Initial Coin Offering (ICO). SSRN.
Sept. 1st. Available from: http://dx.doi.org/10.2139/ssrn.3037106.
Zetsche, D.A., Buckley R.P., Arner, D.W. and Fohr, L. 2018. The ICO Gold Rush: It's a
scam, it's a bubble, it's a super challenge for regulators. University of Luxembourg
Working Paper.
26
8. Appendix
Appendix 1. Summary Statistics. This table highlights the mean, standard deviation, minimum and
maximum of each variable of the 132 ICOs analysed.
Variables
N
Min
Max
Mean
Std. Dev
Amount Raised ($)
132
50,016
200,000,000
14,415,780.66
21,657,010.59
% with respect objective
132
0.00170
1
0.5929581
0.39439582
Success
132
0
1
0.52
0.501
Trading
132
0
1
0.50
0.502
Duration (days)
132
1
365
37.65
48.167
Europe
132
0
1
0.43
0.497
KYC
132
0
1
0.79
0.410
ERC-20
132
0
1
0.74
0.439
Pre-ICO
132
0
1
0.50
0.502
GitHub
132
0
1
0.55
0.499
Telegram
132
0
83,193
17,8761
17,36839
Twitter
132
0
190,112
10,0474
19,26781
Rating
132
24,50
92.50
68.6212
16.60212
Observations
132
Appendix 2. This table shows the differences between those ICOs considered
successful and those considered unsuccessful
Success
No Success
Success
Mean
Amount Raised
6463992,44 $
21676109,04 $
% with respect to objective
21,26%
94,02%
Duration
54 days
23 days
Telegram
10958 followers
24193 followers
Twitter
5169 followers
14501 followers
Rating (0-100)
61,91
74,75
Percentages
Trading
28,6%
69,6%
Europe
60,3%
27,5%
KYC
68,3%
88,4%
ERC20
65,1%
82,6%
Pre-ICO
58,7%
42,0%
GitHub
49,2%
60,9%
27
Appendix 3. Country-regulation, and Europe/Not
Total
Non Europe
Europe
Regulation
country
Non
25,3%
89,5%
53,0%
Yes
74,7%
10,5%
47,0%
Total
100,0%
100,0%
100,0%
Appendix 4. Complete results of Table 2 (p.19): OLS multiple regression analysis results on
the dependent variable: percentage of the amount collected with respect to the objective of the
ICO
Dependent Variable: Respect to capital
Method: Least Squares
Sample: 1 132
Included observations: 132
Variable
Coefficient
Std. Error
t-Statistic
Prob.
Constant
0.195319
0.134075
1.456792
0.1477
Duration
-0.002894
0.000713
-4.059713
0.0001
Europe
-0.125107
0.058172
-2.150618
0.0335
KYC
0.084082
0.072080
1.166505
0.2457
ERC-20
0.080396
0.061944
1.297895
0.1968
Pre-ICO
-0.036236
0.055682
-0.650772
0.5164
GitHub
0.014894
0.056667
0.262836
0.7931
Twitter
0.004437
0.002047
2.167149
0.0322
Telegram
0.004174
0.002023
2.063656
0.0412
Rating
0.004742
0.001975
2.400969
0.0179
R-squared
0.439462
Mean dependent var
0.592958
Adjusted R-squared
0.398110
S.D. dependent var
0.394396
S.E. of regression
0.305978
Akaike info criterion
0.542130
Sum squared resid
11.42198
Schwarz criterion
0.760524
Log likelihood
-25.78056
Hannan-Quinn criter.
0.630875
F-statistic
10.62754
Durbin-Watson stat
2.000656
Prob(F-statistic)
0.000000
Appendix 5. This table represent the results of the Breusch-Pagan-Godfrey
test for Heteroskedasticity
Heteroskedasticity Test: Breusch-Pagan-Godfrey
Null hypothesis: Homoskedasticity
F-statistic
0.517638
Prob. F(9,122)
0.8595
Obs*R-squared
4.855207
Prob. Chi-Square(9)
0.8467
Scaled explained SS
3.119697
Prob. Chi-Square(9)
0.9594
28
Appendix 6. Complete Results of Table 3 (p.21): Binary logit multiple regression analysis results on
the dummy dependent variable success. Method used: ml - binary logit (Newton-Raphson /
Marquardt steps)
Dependent Variable: SUCCESS
Method: ML - Binary Logit (Newton-Raphson / Marquardt steps)
Sample: 1 132
Included observations: 132
Convergence achieved after 6 iterations
Coefficient covariance computed using observed Hessian
Variable
Coefficient
Std. Error
z-Statistic
Prob.
Constant
-2.256768
1.112741
-2.028116
0.0425
Duration
-0.033281
0.009924
-3.353437
0.0008
Europe
-1.090466
0.495865
-2.199118
0.0279
KYC
0.471943
0.661572
0.713367
0.4756
ERC-20
1.035935
0.562561
1.841462
0.0656
Pre-ICO
-0.090884
0.492949
-0.184369
0.8537
GitHub
-0.243363
0.513567
-0.473868
0.6356
Twitter
0.062395
0.026003
2.399484
0.0164
Telegram
0.032718
0.018766
1.743524
0.0812
Rating
0.026865
0.017096
1.571440
0.1161
McFadden R-squared
0.380807
Mean dependent var
0.522727
S.D. dependent var
0.501386
S.E. of regression
0.386386
Akaike info criterion
1.008618
Sum squared resid
18.21392
Schwarz criterion
1.227013
Log likelihood
-56.56882
Hannan-Quinn criter.
1.097364
Deviance
113.1376
Restr. deviance
182.7180
Restr. log likelihood
-91.35902
LR statistic
69.58039
Avg. log likelihood
-0.428552
Prob(LR statistic)
0.000000
Appendix 7. This table shows the level the correlation between the variables of the regression analysis
Duration
Europe
KYC
ERC20
Pre-ICO
GitHub
Twitter
Telegram
Rating
Success
Duration
1.00
0.22
-0.21
-0.02
0.13
0.00
0.43
-0.04
-0.14
-0.32
Europe
0.22
1.00
-0.22
-0.01
0.14
-0.02
-0.10
-0.07
-0.07
-0.33
KYC
-0.21
-0.22
1.00
-0.01
0.00
0.06
-0.03
0.10
0.34
0.25
ERC-20
-0.02
-0.01
-0.01
1.00
0.07
0.10
0.06
0.03
0.12
0.20
Preico
0.13
0.14
0.00
0.07
1.00
-0.05
-0.14
-0.09
-0.10
-0.17
GitHub
0.00
-0.02
0.06
0.10
-0.05
1.00
0.15
0.12
0.30
0.12
Twitter
0.43
-0.10
-0.03
0.06
-0.14
0.15
1.00
0.53
0.24
0.24
Telegram
-0.04
-0.07
0.10
0.03
-0.09
0.12
0.53
1.00
0.41
0.38
Rating
-0.14
-0.07
0.34
0.12
-0.10
0.30
0.24
0.41
1.00
0.39
Success
-0.32
-0.33
0.25
0.20
-0.17
0.12
0.24
0.38
0.39
1.00
29
Appendix figure 1. Connelly et al (2011, p.44) figure 2 illustrating signalling theory elements in
the form of a timeline
Signalling Timeline
Appendix figure 2. Kirmani & Rao signalling theory table (2000, p. 14).
