Master Thesis An empirical research into company level factors explaining post-ICO underperformance.

(1)

Master Thesis

An empirical research into company level factors explaining

post-ICO underperformance.

Student: Hugo Janse (S2448556)

MSc Business Administration - Small business and entrepreneurship

Supervisor: Arjan Frederiks Co-assessor: Samuele Murtinu

Word Count: 13587 University of Groningen

June 2018

Abstract: Recent development of the blockchain technology has led to the creation of initial

(2)

2

1. Introduction

Blockchain start-ups have raised over $6.5 billion of investment capital through Initial Coin Offerings (ICOs) in the first quarter of 2018. This in comparison to a total of $4 billion in 2017 and way up from the $96 million raised in 2016 (Coinschedule, 2018). This enormous growth in raised capital follows the growth explosion of the total cryptocurrency market itself. This enormous influx of money provides many opportunities for investors and companies looking for funding, but it is not without risks.

ICOs are based on the blockchain technology. This technology combines peer-to-peer file sharing with public key cryptography to create an open, distributed ledger that can record transactions between parties in a verifiable and permanent way (Iansiti & Lakhani, 2017; Swan, 2015). In 2008, this technology was proposed and used to create Bitcoin. A decentralized currency not backed by central authorities but by cryptography and a decentralized consensus amongst its users (Nakamoto, 2008; Swan, 2015). In the years after the launch of Bitcoin the blockchain technology was further enhanced and different cryptocurrencies started to develop based on changes made to the blockchain technology. During 2013 and 2014 a new way of distributing these different cryptocurrencies emerged. The developers of the crypto projects started to sell the new cryptocurrencies, also called tokens or coins, to investors in exchange for Bitcoin or other cryptocurrencies to fund the development of their project. This is now called an Initial Coin Offering as it is the first time the new coin is offered to the public to buy, hence the name.

ICOs share traits with Initial Public Offerings (IPOs), where shares of a company are first offered to the public on the stock market (Kaal & Dell ’erba, 2018; Sontakke & Ghaisas, 2017). During the ICO, a certain stake of the crypto project is sold to investors to raise funds for the development of the project or company, similar to an IPO. Unlike an IPO, however, the tokens do not confer ownership rights. Owning the token in most cases does not give the right to vote, receive profit or dividends from the company, it only allows the use of the tokens features. The underlying idea is that these tokens appreciate when the company is able to successfully build their platform because these tokens are needed to use it (Iurina, 2017). For example, Filecoin is a cloud storage company that lets users rent out their unused storage space on their computer (Protocol Labs, 2017). To pay for this storage space within their network they launched their own token named Filecoin during their ICO in September 2017. When there is an increase in demand for cloud storage space on their platform the value of the token is expected to rise following normal market dynamics. Investors holding this token can then sell their tokens to users who would like to use the platform for cloud storage space.

(3)

3

rendering the tokens worthless leaving the investors with nothing. This risk can be seen from the huge spread in Return On Investment (ROI) between ICOs. Some successful projects have returns on investment over 20.000% within a few years, while others are now just worth a fraction of their initial offering price (Ico Stats, 2018). Adding to this case is that most ICOs are funded in Bitcoin or Ether (the token of Ethereum, currently the second largest cryptocurrency). Both of these have had returns of over 3500% from the start of 2015 till the end of 2017 (Coinmarketcap, 2018). Out of 298 tokens that received funding over $100.000 between September 2015 and March 2018, 133 underperformed the returns of Bitcoin and 158 underperformed Ethereum in the period from their ICO till May 2018 (Tokendata, 2018). It thus shows that in about half of these cases the investors were better of doing nothing and keeping their original Bitcoin or Ether instead of trading these in for the token of the ICO. But how can these huge differences in returns be explained? Because of the very recent development of blockchain technology and ICOs scientific literature is very sparse. Only a handful of papers have looked at the emergence of ICOs, their risks, the economics behind them and their potential to provide investment capital (Conley, 2017; Iurina, 2017; Kaal & Dell ’erba, 2018; Yadav, 2017). But none of these have looked at the return on investment of the token after the ICO has finished. This return after the ICO however is the one that provides the opportunity or risk involved with buying the ICO. When buying the ICO investors are expecting a positive return on their investment. As explained earlier this return should be higher than the return of Bitcoin or Ether to offset the risk of buying the ICO instead of keeping their original cryptocurrency. When it does not provide higher ROI then Bitcoin or Ether during the same period this paper will speak of underperformance. When the ICO is underperforming, the investor is thus not having the highest returns or is even losing money on his investment. It is therefore important to find factors that lead to the underperformance of about half the ICOs. This helps investors choose which ICOs to invest in to maximize returns and minimize risk. Furthermore, we are only focussing on factors that the company having the ICO can influence. The company having the ICO is able to make certain choices on when and how to launch their new token during and after their ICO. We will be looking for these company level factors that influence the post-ICO underperformance. The research question that will be answered in this research is:

What company level factors which are present before and during the ICO can explain post-ICO underperformance?

(4)

4

The structure of this research proposal will be as follows: in section 2 the current literature will be reviewed and hypotheses will be developed using tested theories. Section 3 will describe the research approach, how data will be gathered and how this data will be analysed. Section 4 will describe and analyse the findings and section 5 will discuss and conclude.

2. Literature Review

In this section I will first argue why the IPO literature is of interest to ICO underperformance. Afterwards I review the existing literature that studies IPOs and ICOs and certain factors surrounding them that have an influence on their post-IPO/ICO underperformance.

To find factors influencing ICO underperformance I will argue that the current ICO market is a lot like the IPO market because of several reasons. First, an ICO shows similarities in what it really is. It is the first time a new token is publicly traded on exchanges in the same way that after an IPO the stock is publicly traded on the stock exchange for the first time. Second, it shows from the research of Adhami et al. (2017) that companies having an ICO sell their tokens at a large discount when compared to the first day closing price just like IPOs. This so called underpricing of ICOs is even higher than the underpricing seen with IPOs. On average the first day returns are 919.9% in their sample of ICOs (Adhami et al., 2017). This is several factors higher than the underpricing of IPOs seen by Loughran & Ritter (2004) which ranged from 7% in the 1980s to 65% during the dot-com bubble years of 1999-2000.

IPOs during the dot-com bubble show even more similarities with ICOs. This bubble was characterized by inflated valuations, enormous amounts of offerings and positive sentiment from retail investors (Chan, 2014). These characteristics can be found in the current ICO market as well, with valuations of the cryptocurrency market sharply rising and falling. The number of offerings exploding from 43 in 2016 to almost 240 in the first 4 months of 2018 alone (Coinschedule, 2016). Investors showing positive sentiment by expecting triple digit returns and often stating their ICO of choice is “going to the moon” implying the valuation is going to increase enormously (Finder.com, 2018). Like ICOs, these IPOs have been shown to underperform when compared to stock indexes (Alon & Gompers, 1997; Coakley, Hadass, & Wood, 2007; Ritter, 1991).

There are however some limitations to this comparison, ICO firms are much smaller, younger and often in the first stage of a firm’s life cycle. IPO firms on the other hand are more established companies with a median age before their IPO ranging from 5 years during the dot-com bubble till 12 years after the bubble (Loughran & Ritter, 2004). ICO firms also do not use an underwriter that helps them price and sell their token unlike IPO firms that do use an underwriter.

(5)

5

2.1. IPO underperformance

Post-IPO underperformance has been widely documented in the literature concerning IPOs. Research looking at post-IPO stock returns have shown an underperformance of nearly 9% per year for three years when compared to market indexes (Ritter, 1991). Further research by Alon & Gompers (1997) and Loughran & Ritter (1995) find underperformance for up to 5 years post-IPO. This underperformance is of interest because it strokes with market efficiency theory. According to market efficiency theory anomalies in price should only be short lived because the market efficiently prices all available information into the stock price (Fama, 1998). This seen long term underperformance thus suggests markets are inefficient. This underperformance of IPOs when compared to various benchmarks has been the subject of many studies coming up with different explanations.

Underperformance is a relative term and thus a certain benchmark must be chosen: when the IPO performs better than this benchmark in the same time frame we speak of outperformance, when it performs worse we speak of underperformance. Because most IPOs show underperformance for periods up to five years after their IPO, underperformance is used as the viewpoint (Ritter, 1991). Different benchmarks are used to show IPO underperformance. Ritter (1991) and Loughran & Ritter (1995) compare the IPO stock prices to the stock prices of firms that match in size and industry but did not have an IPO in the past 5 years. Others compare the stock prices to different market indexes like the Nasdaq, CRSP and S&P500 to show the difference in return on investment if the money was invested in this index instead of the IPO (Alon & Gompers, 1997; Ritter & Welch, 2002).

The recent study by Liu & Forester (2014) reviews the current literature on IPO underperformance and they find there are three main streams of research concerning this phenomenon. I will follow their research and discuss these three main streams of research. The first stream of research looks at IPO underpricing to explain the observed underperformance post-IPO. Underpricing is the percentage difference in price between the offering price and the first day closing price (Loughran & Ritter, 2004). This underpricing leaves a lot of potential money on the table for the issuing firm, by not asking the highest price possible (Ljungqvist & Wilhelm, 2003). The reasons for underpricing have changed throughout the years. During the 1980s the average underpricing was 7%, IPO firms were looking to maximize IPO proceeds and thus tried to price their stock as high as possible (Loughran & Ritter, 2004). During the dot-com bubble analyst coverage, side payments to CEOs and Venture Capitalists (VCs) were the main reason for increased underpricing. These actors had an incentive to underprice for their own gains which lead to a rise in the average underpricing during this bubble up to 65% (Loughran & Ritter, 2004). This higher underpricing has shown to lead to worse operating performance post-IPO. This worse operating performance is explained by the high first day returns negatively affecting the investor sentiment (Coakley et al., 2007).

(6)

6

revealed after the IPO, this information is then incorporated into the stock price leading to underperformance over a multi-year period (Liu & Forester, 2014). One of these market related explanations for underperformance that is put forward are “hot IPO markets”. These “hot” or sometimes called “bubble” market periods show IPOs earning higher than average returns in the first month. There have been a number of these periods over the last few decades (Ibbotson & Jaffe, 1975; Ljungqvist et al., 2006; Ritter, 1984). Empirical evidence has shown that companies that have an IPO during these hot markets on average underperform major stock indexes even more than during normal periods for at least 3 years (Alon & Gompers, 1997; Ritter, 1991). An explanation of this larger underperformance is that during these hot markets more companies go public because investors are willing to pay more because of optimism about growth opportunities (Ritter, 1991). When these investors see the real growth of the company after the IPO they are disappointed leading to negative views on future value causing the underperformance post-IPO (Liu & Forester, 2014; Ritter, 1991). During these hot markets there is also a disproportionate amount of “low-quality” companies which are going public for opportunistic reasons. These companies have lower productivity and post-IPO profitability (Liu & Forester, 2014; Ljungqvist et al., 2006). The research of Coakley et al. (2007) tests for empirical evidence on these findings by looking at the UK IPO market between 1985 and 2003. They confirm that during the hot market more low-quality firms went public and find that companies which have low pre-IPO quality show poor post-IPO performance. As the different scholars have found, these hot market periods attract more IPOs which underperform the indexes worse than during normal periods. Alon & Gompers (1997) show this by comparing IPOs between 1972-1992 to four market indexes: The S&P 500, Nasdaq value weighted composite index, NYSE/AMEX value weighted index, and NYSE/AMEX equal weighted index. They too find that periods with the greatest number of IPOs are associated with the most severe underperformance.

(7)

7

present value (NPV) leading to post-IPO underperformance. Liu and Forester (2014) provide evidence on this by looking at almost 7000 US IPOs, their findings show that managerial decisions on excess purchases and aggressive revenue recognition contribute to post-IPO underperformance. By focussing on revenue and making excessive inventory purchases they have trouble generating positive cash flows leading to investors being less optimistic about the company’s future leading to larger post-IPO underperformance (Liu & Forester, 2014). This comprehensive literature review on the current state of IPO underperformance research follows the research of Liu & Forester (2014) by explaining the three streams of literature describing why this underperformance is existent. These can be summarized as three factors leading to post-IPO underperformance namely: IPO underpricing, market related factors and internal company factors. These might not be the only factors influencing IPO underpricing, but they have been shown to be the main focus of many scholars and show high correlation with post-IPO underperformance. Hence why this research will also focus on these three factors by looking for comparable dynamics in the ICO market.

2.2. ICOs

ICOs were first introduced in 2013 by Willett (2013) with the Mastercoin. He proposed to build a protocol layer on top of the existing bitcoin network. On top of this new layer a new currency with its own rules could be built. He started selling Mastercoins for Bitcoin to fund the development of the project. A year later Ethereum was introduced by Buterin (2014), this new currency allowed contracts to be put into a blockchain and allows the platform to be used for decentralized applications. To run these applications the token Ether was introduced This token is the fuel of the network and is needed to pay transaction fees (Buterin, 2014). This Ether was sold to investors for Bitcoin to fund the development of the platform. The introduction of Ethereum allowed for an easy way of starting an ICO by allowing applications to be built on the network itself. These applications can also automatically trade the Ether that is being send to it into the token of the ICO. This led to the majority of developers to opt for starting their ICO on the Ethereum network instead of creating their own blockchain technology (Kaal & Dell ’erba, 2018). This development led to a massive increase in the use of the Ether token to run these applications leading to an increase in price of ether from about $0.30 when introduced to almost $1400 at its peak in January 2018 (Coinmarketcap, 2018). Currently 83% of ICOs are done via the Ethereum network with only 8% of ICOs using a custom platform (Icowatchlist, 2018).

Most of these ICOs follow a sequence that is described by Kaal & Dell ’erba (2018) this sequence consists of four stages put forward by (Aitken, 2017):

Pre-announcement: Before the announcement of an ICO, the project itself is announced on

(8)

8

need funds to develop the idea, hence the announcement of the ICO. Questions following this announcement will be answered and the business model can be adjusted accordingly.

The offer: Following these revisions a white paper is often released. The white paper is aimed

to inform investors, so they can assess the value of the project and its ICO. The white paper can describe the project’s purpose, business model and the details of the offering. These white papers are not audited by authorities, but companies like Bitsify and ICOAlert write analyst reports on them to help investors choose promising ICOs. Before the ICO is announced a pre-sale can be conducted. A pre-pre-sale is to offer selected large investors the opportunity to buy against discounted rates (Kaal & Dell ’erba, 2018). This is done so start-ups have funding to launch there ICO and run promotions (Bitcoinwiki, 2018). After the pre-ICO the start date of the ICO is announced.

PR campaign: When the ICO date is announced a promotional campaign is started to get

(smaller) investors to notice the project and its offer. Many different forms of promotions are used like advertisements on websites, Instagram, Facebook and even celebrities like Paris Hilton and Floyd Mayweather have been paid to promote an ICO on their social media (Roberts, 2017). The Securities and Exchange Commission (SEC) however has made a statement that these promotions are “unlawful if they do not disclose the nature, source and amount of any compensation paid” (SEC, 2017). This campaign will usually run till the end of the ICO.

The ICO itself: The company starts selling the tokens to investors according to the

specifications mentioned in the white paper. These specifications at least inform the investor about the total token supply, exchange rate, if there is a minimum or maximum funding goal and how or when the ICO will end. During the ICO a bonus scheme can be present. This attracts investors by giving them extra tokens or a better price when they buy early or more (Adhami et al., 2017). The ICO can be finished within a matter of seconds like Brave, they reached their funding limit of $35 million within 30 seconds and closed the ICO (Russel, 2017), or can go one for the full length of the ICO when no limit is set or reached like Filecoin. The investments are most often collected on a public Bitcoin or Ethereum address, so the community can audit the money raised and where the money is going (Kalla, 2016).

After the ICO it can take anywhere from days to months before the token is listed on an exchange. These exchanges let investors trade the token against Bitcoin, Ether, other cryptocurrencies or even euros or dollars. This allows the market to decide on its worth just like the stock market does with publicly traded companies. As was mentioned in section 1, a large spread in returns when compared to Bitcoin and Ethereum can be observed. About half of the tokens have lower returns than Bitcoin or Ethereum, these two are used as a benchmark because most ICOs are funded in Bitcoin or Ether (Tokendata, 2018). It is thus important to compare the returns of the ICO against those to see if they are higher (outperformance) or lower (underperformance) in which case the investor would have been better off not exchanging his Bitcoin or Ether for the token of the ICO.

(9)

9

Yadav (2017) provides a basic framework that provides signals for ICO investment. Randolph (2015) proposes different ways to regulate ICOs and Adhami et al. (2017) do an empirical analysis of what factors have an effect on the capability of the ICO to reach its funding goal. While these scholars fail to look at what happens to the token value, and thus the return to the investor, after the ICO has finished they do provide clues to differences between ICOs that could explain the observed underperformance. I will argue that similarities can be found which could explain the observed underperformance of ICOs based on the 3 factors which explain IPO underperformance namely: IPO underpricing, market related factors and internal company factors. These factors are chosen because they represent the 3 main research streams that try to explain post-IPO underperformance (Liu & Forester, 2014). I will provide similar factors in ICOs and they will be developed into testable hypotheses to see if these concepts are also valid when looking at ICOs.

ICO underpricing

Underpricing of IPOs has received much attention from different scholars (Booth & Chua, 1996; Loughran & Ritter, 2004; Ritter, 1991). As mentioned in section 2.1 this underpricing leads to larger underperformance. This underperformance is explained by the negative effect of underpricing on investor sentiment. During the dot-com bubble underpricing rose from an average of 17 percent (median:10 percent) in 1996 to a stunning 73 percent (median: 40percent) in 1999 (Ljungqvist & Wilhelm, 2003). This high underpricing can also be observed in the ICO market to an even larger extent. Adhami et al. (2017) observe an average underpricing of 919.9 percent when the token is first publicly traded. Our sample shows an average underpricing of 267%. We also observe that about half the ICOs underperform when compared to Bitcoin or Ethereum (Tokendata, 2018). A token which has high returns on the first day of trading gives early ICO investors an incentive to sell for a quick profit. The higher the underpricing the more profit the investor makes within a short time frame putting more selling pressure on the token. These high first day returns also provide more risk for investors that want to buy the token as it has already increased in price by a large amount. This reduces the possible returns in the future. This high underpricing of ICOs could thus have the same negative effect on investor sentiment leading to underperformance like it does with IPOs as seen by (Coakley et al., 2007). This leads to the first hypothesis:

H1: Higher underpricing of ICOs leads to larger post-ICO underperformance.

ICO market factors

(10)

10

(1995) look at the number of US IPOs between 1970 and 1990, they show a low of only 8 companies going public in 1974 and a high of 666 companies going public in 1986. They show that IPOs have smaller underperformance during periods were less companies go public, and larger underperformance when more companies go public. The ICO market also shows a large variation in the amount of companies that have an offering in a certain period. Ranging from just 43 in 2016, 210 in 2017 and almost 240 in the first four months of 2018. This market seems to be in the middle of a hot market. Because of the exceptional returns of Bitcoin, Ethereum and early successful ICOs, investors want to believe they can identify the next token that is going to offer triple digit returns. This leads to high valuations of new ICOs which incentivises companies to have an ICO because it is easier to attract a large amount of capital. This mechanic was also observed by Loughran & Ritter (1995) with IPO investors. These investors were looking for the next IPO which was going to have the same growth as Microsoft in the 90s. This positive investor sentiment leads to high valuations and IPO issuers taking advantage of this window of opportunity (Loughran & Ritter, 1995). Investors are willing to pay more because of optimism but after the investors see the real growth of the company this optimism fades leading to negative views on future value causing underperformance post-IPO (Liu & Forester, 2014; Ritter, 1991). This fading of optimism is also observed in our dataset. ICOs like Sense, Vezt and Mercury protocol, which held their ICO during the highly active month of November 2017, faded quickly after and have almost no trading volume today.

These observations lead to the second hypothesis:

H2: A higher number of ICOs in a certain month leads to larger post-ICO underperformance of ICOs which go public during this month.

(11)

11

the longer time frame. Second, providing a white paper also reduces the information asymmetry between investor and the company. By providing information the investor is better able to evaluate the quality of the company allowing them to better value the token. Having a white paper can thus lead to the investor seeing the company as a higher quality firm leading to smaller underperformance as was also argued and observed by Coakley et al. (2007) with high quality IPO firms. This brings us to the third hypothesis:

H3: Companies that provide a white paper show smaller post-ICO underperformance than companies that do not.

The second quality factor put forward by Adhami et al. (2017) is the public availability of the source code the technology of the ICO is based on. Adhami et al. (2017) find that code availability is positively related and highly significant to ICO funding success. I argue that this code availability also helps the company in the long run by decreasing the post-ICO underperformance because of two reasons. First, it provides investors with a tangible proof of concept so they can see the technology is real and get a sense of its quality (Adhami et al., 2017). This also shows the company has already put in some effort to create something and the ICO is not only fuelled on promises. Second, the source code is the basis on which the whole project is build and it is not only limited to the funding round of the ICO. It is thus likely that providing the source code also has longer term effects on the ICO. This code can be further developed post-ICO into the technology the company was promising during the ICO, providing value to the token. Which is likely to lead to smaller underperformance post-ICO. This provides the fourth hypothesis:

H4: Companies that provide the source code of their project show smaller post-ICO underperformance than companies that do not.

ICO Operating Decisions

The last hypotheses will focus on three internal operating decisions of the company having the ICO. Namely: Venture Capital involvement, direct share programs and managerial investment in inventories and projects. We will only focus on these three factors because these have been studied extensively and have shown a highly correlated and significant effect on post-IPO underperformance as stated in section 2.1. Because of the non-regulated nature of ICOs it is hard to get information about internal operating decisions because these are often not reported to the public. Hypotheses on their influence on post-ICO underperformance are thus limited to the information about internal operating decisions which most ICOs provide. We will start with VC involvement.

(12)

12

and shows credibility by gathering larger investors upfront (Adhami et al., 2017; Kaal & Dell ’erba, 2018). When a company has had a presale, this shows it has interest from large investors like VCs which can also provide guidance during the development of the project. This is not unlike the role VCs play in IPOs leading to smaller underperformance. This leads to the fifth hypothesis:

H5: ICOs that have had a presale show smaller post-ICO underperformance than companies that do not.

The second internal decision that was crucial to the underperformance of IPOs during the dot-com bubble were the directed share programs which let investors appoint shares to “family and friends” at the discounted offer price (Ljungqvist & Wilhelm, 2003). This directed share program let to more insiders wanting to sell their shares post-IPO to make a quick buck. I argue that this same mechanic of people getting in early at discounted prices looking to sell post-IPO can be seen in the ICO market as well. By offering bonus schemes, where investors get a better price the earlier or more tokens they buy, there is an incentive for these early investors to sell their tokens post-ICO for a profit, putting downward pressure on the price. This is further enhanced by the enormous average first day returns of 919.9 percent reported by Adhami et al. (2017) giving the early investors even more reason to sell their tokens. Having a bonus scheme also showed a marginally significant effect on the likelihood of a successful funding of the ICO (Adhami et al., 2017). It thus seems investors are attracted to these schemes and want to get in early at discounted prices to sell post-IPO to make a nice profit, not caring about the effects this could have on the longer term post-ICO underperformance. I therefore put forward the sixth hypothesis:

H6: ICOs that have a bonus scheme show larger post-ICO underperformance than companies that do not.

(13)

13

Figure 1. Conceptual model

I will continue with describing how this research will be conducted in the methodology section.

3. Data and Methodology

This section will describe what variables we will measure to test the hypotheses, how we build the data set for our sample, how we will analyse this sample, describes our sample and provides some thoughts on validity and reliability.

3.1. Variables

To test the hypotheses, we will have to clearly define the variables we are using. The first variable that needs to be defined is the dependent variable underperformance of the token.

Dependent variable

6-month relative return: To define the underperformance of the token we need to have a

benchmark to which we compare the return of the token. Because 83% of tokens are funded in Ether we will use Ether as the benchmark as it is most likely the investor traded his or her Ether

Post-ICO underperformance H2 Amount of ICOs

during the month

(14)

14

for the token that is sold in the ICO. The returns used in this sample are based on the buy and hold method that is consistent with methodologies used to calculate IPO returns by Ritter & Welch (2002) and Loughran & Ritter (1995). The return after 6 months is chosen because it gives the factors time to influence the underperformance of the token. It also allows for a much larger dataset than the 12-months underperformance because most ICOs are not yet 1 year old. The dependent variable 6-month relative return was calculated with the following formula and this was repeated for the variables 3 and 12 months post ICO:

(Value of token 6 months post ICO in USD / token ICO price in USD) / (Value of Ether 6

months post token ICO in USD / Value of Ether at token ICO in USD)

This variable showed strong right skewed distribution. To normalize this distribution, we continued our analysis with the natural logarithm of the above-mentioned formula. When a token offers higher returns (value >0) than holding Ether this can be seen as overperformance when it has lower returns (value <0) we speak of underperformance. This is comparable to the way Alon & Gompers (1997) set their underperformance by comparing stock returns to benchmark returns like the S&P 500 and Nasdaq.

Independent variables

Underpricing: To calculate the underpricing, the closing price of the token after the first day

of trading registered on Coinmarketcap (2018) will be divided by the ICO price. This provides a factor that represents the first-day return which is consistent with the way Loughran & Ritter (2004) calculate underpricing. This variable also showed a strong right skewed distribution. To normalize this distribution, we took the natural logarithm of this variable.

Number of ICOs in the same month: This variable was calculated by counting the number of

ICOs that happened in the same month as the ICO itself. This is consistent with how Loughran & Ritter (1995) aggregate the number of IPOs.

The following variables will be added with a binary variable taking the value of 1 if it is present in the ICO and 0 if it is not. This is consistent with Adhami et al. (2017) and lets us compare the underperformance of the ICOs that do have the variable against those that do not.

Code availability: If the ICO has a Github account with at least 1 repository it is counted as

having its source code available. Github is the usual way ICO companies provide their code (Adhami et al. 2017). We did not look at the provided code itself because of two reasons. First, this code is subject to change. Second, for the scope of this research we are unable to look at thousands of lines of code to determine if it is of high enough quality to be counted as having the source code available.

White paper availability: We searched using the Google search engine for the ICO name

(15)

15

Presale availability: If the ICO sold tokens to investors before the ICO itself this was seen as

having a presale, unconditional of the number of tokens sold or the money raised during the presale.

Bonus scheme availability: If the ICO had any type of sale incentive, which gave investors the

opportunity to get in at lower price than the normal ICO price or to get more tokens for the same price, this was noted as having a Bonus scheme.

Control variable

Total amount raised in USD: The dataset also provides the total amount of USD raised for

every ICO. No study could be found which connected the total amount raised to post-IPO underperformance or any effect it could have on the price of the token. The equivalent of this with an IPO would be the gross proceeds of the IPO. But no study was found which linked this to post-IPO underperformance. To make sure this variable has no hidden effect on our dependent variable we will use it as a control variable called Total amount raised in USD. This variable also showed a strong right skewed distribution, so the natural logarithm was calculated to normalize this distribution.

Data selection and sources

Because ICOs are unregulated there is no comprehensive list of all ICOs that have ever taken place. We will therefore have to work with a subset that is available from public information on the internet. The main source that is used is Tokendata.io (2018), this website provides an overview of the name, amount of capital raised, month of ICO and token sale price. A total of 317 tokens are recorded in this database. To exclude ICOs which have only raised a small amount which could behave different from larger ICOs, we put a bottom limit of $100.000 raised at the end of the ICO. To include as many ICOs as possible we use all ICOs that took place between January 2015, to only allow tokens that were raised after the invention of Ethereum, and the first of December 2017, so there would be at least a 6-month post-ICO datapoint. The token also must be registered by Coinmarketcap.com so historical prices can be looked up. We excluded ICOs that were not registered on Coinmarketcap within 3 months after their ICO because this would leave too much time for the price to fluctuate to still resemble underpricing.

After filtering this sample by only including ICOs that did meet these requirements: • More than $100.000 raised

• ICO took between January 2015 and 1 December 2017

(16)

16

After this filtering the final sample consists of 185 ICOs of which a full list is presented in Appendix B. For these 185 ICOs the following datapoints will be collected as described in the variables section:

• End date of the ICO to be used as the start date of the token. • Total amount of USD raised during the ICO.

• Token price at the end of the first day of trading to calculate underpricing.

• Token prices at 3, 6 and 12 months after the finish of the ICO to calculate token returns for different timeframes.

• Ethereum and Bitcoin prices for these same dates to calculate the underperformance or outperformance of the token.

• White paper availability • Code availability • Presale availability

• Bonus scheme availability

These definitions follow the approach of Adhami et al. (2017). The data will be looked up on different internet sources concerning ICOs namely: the website of the ICO project itself, Coinmarketcap (2018), Tokendata (2018), Smith+Crown (2018), ICOwatchlist (2018), Cryptoslate (2018), ICObench (2018) and Bitcointalk (2018).

3.2. Analysis

To summarize the data set, a descriptive statistics table was drawn up. This table provides information on: the mean, standard deviation and the variance inflation factor (VIF). After that, a bivariate correlation analysis was conducted to measure the degree of association between the different variables. Finally, the hypotheses were empirically tested using a linear regression. To check for robustness of the results, the underperformance of the ICO will be tested at 3, 6 and 12 months after the ICO if this data is available. This provides us with multiple time frames to test if our hypotheses hold. As a second robustness check the hypotheses will also be tested against Bitcoin instead of Ether as a benchmark.

3.3. Description of sample

(17)

17

Variables Mean SD VIF

LN 6-month relative return -.017 1.296

LN Underpricing .606 .96057 1.068

ICOs this month 19.39 8.044 1.548

White paper availability .98 .146 1.250

Code availability .79 .405 1.230

Presale availability .65 .479 1.313

Bonus scheme availability .69 .465 1.064

LN USD Raised 15.891 1.473 1.423

Table 1. Descriptive statistics

Table 2 displays the results of the bivariate correlation analysis of all variables. As the table shows there are multiple significant correlations at the 1% or 5% significance level. Underpricing has a correlation of r = 0.338 (p < .01) with the relative return of the ICO. This seems to indicate that higher underpricing leads to smaller ICO underperformance conflicting with H1. Another interesting finding is the small correlation between the number of ICOs and

underpricing of r = -.158 (p < .05). This indicates that underpricing is lower when there are more ICOs. ICOs that happen during busy months also seem to be more likely to have a white paper, code and presale and they show a moderate to large correlation to raise more funds. This

follows from the increase in ICOs during the second half of 2017. During this period presales and code availability became the norm and ICOs were raising more because of the increased investor sentiment. There is also a small to moderate correlation between Code availability and White paper availability of r = 0.292 (p < .01). Indicating that companies that publish a white paper are more likely to also publish their code and the other way around. Companies that provide a white paper also show a moderate correlation to raise more funds (r = .307, p < .01). Companies that have their code available are less likely to have a bonus scheme (r = -.171, p < 0.05) but are more likely to have a presale (r = .242, p < .01) and they raise more money (r = .256, p < .01). The last significant correlation is that companies which have a presale seem to raise more funds (r = .339, p < .01). All correlations found in our correlation matrix are small to moderate and our VIF test showed it is highly unlikely that these cause problems during the regression analysis.

Variables 1 2 3 4 5 6 7 8

1 LN 6-month underperformance 1

2 LN Underpricing .338** 1

3 ICOs this month .102 -.158* 1

4 White paper availability .066 .070 .299** 1

5 Code availability .099 .127 .192** .292** 1

6 Presale availability .040 -.036 .410** .046 .242** 1

7 Bonus scheme availability -.053 .023 .070 -.020 -.171* -.034 1

8 LN USD Raised .030 -.064 .461** .307** .256** .339** -.108 1

(18)

18

3.4. Validity and reliability

Because we are only using data provided by the ICOs themselves our data should have high validity and reliability. All our data is based on the information provided by the ICOs without any interpretation or bias. Every variable is also only based on one construct and all information used is publicly available on the internet on the sources that were mentioned in section 3.1. Our results should therefore be highly reliable because other researchers should get the same results if they would use the same ICOs. Our main issue on reliability is that we are not able to test our hypothesis on all ICOs that have ever taken place. We had to exclude ICOs because they didn’t fit our requirements or information was not generally available. We do however still have a dataset of 185 ICOs which should be enough to provide significant and reliable results which we will discuss in the next section.

4. Results

This section will discuss the results from the linear regression in detail.

4.1. Regression analysis

A linear regression was performed with all the variables including the control variable LN USD Raised to be able to test the hypotheses, the result of this regression can be found in table 3. A significant regression equation was found (F(7,177) = 4.333, p < .000), with an R2 of .146. This means the percentage of the variance of relative token return is explained for 14.6% by the independent variables in the model. This means 85.4% of the variance in the token return compared to Ethereum is because of other reasons.

We will start by testing H1: Higher underpricing of ICOs leads to larger post-ICO

underperformance. Underpricing significantly predicted token relative returns β = .365 (p < .000). This shows a highly significant positive relation between underpricing and post-ICO performance. Higher underpricing leads to smaller underperformance. This is the exact opposite of H1 and we thus reject our hypothesis.

Our H2 was: A higher number of ICOs in a certain month leads to larger post-ICO underperformance of ICOs which go public during this month. The regression shows that ICOs this month significantly predicted token underperformance, β = .189 (p < .05). This shows a small positive relation between the amount of ICOs that happen during the month and the ICO relative returns. Meaning that ICOs that take place during months with more ICOs show smaller underperformance. This leads us to reject our hypothesis.

Our H3 was: Companies that provide a white paper show smaller post-ICO underperformance than companies that do not. The regression shows that having a white paper did not significantly predict the token underperformance, β = -.011 (p > 0.05). Because of the very low number of observations that didn’t have a white paper the significance is really low of this test. This leads us to reject our hypothesis as we did not find any statistically significantly support. Our H4 was: Companies that provide the source code of their project show smaller post-ICO

(19)

19

highly insignificant ability to predict the ICO underperformance, β = .020 (p > .05). We thus reject our hypothesis.

Our H5 was: ICOs that have had a presale show smaller post-ICO underperformance than companies that do not. The results from the regression show no significant ability to predict ICO underperformance, β = -.018 (p > .05). We thus reject our hypothesis.

Our H6 was: ICOs that have a bonus scheme show larger post-ICO underperformance than companies that do not. Our results show no significant ability to predict ICO underperformance, β = -.037, (p > .05). We thus reject our hypothesis.

Variable B Std. Error β

LN Underpricing .493** .097 .365**

ICOs this month .030* .014 .189*

White paper availability -.101 .690 -.011

Code availability .065 .246 .020

Presale availability -.050 .215 -.018

Bonus scheme availability -.213 .200 -.076

LN USD Raised -.033 .073 -.037

R Square .146**

F 4.333

Table 3. Regression results, dependent variable LN 6-month relative return ** = p < .01 * = p < .05

4.2. Robustness checks

Our results were checked for robustness by comparing our results with the token returns after 3 and 6 months and when compared to Bitcoin instead of Ethereum. The results of these regressions can be found in appendix A. These mainly confirmed our findings with some exceptions. When checking for returns against Ethereum after 12 months we found no significant model (F(7,42) = 1.200, p > .10), this is probably due to the low number of ICOs that are older than 12 months (n = 50). When compared after only 3 months against Ethereum, underpricing still has a highly significant effect (B =.548, p < 0.000) but the ICOs this month and the other hypotheses show no significant effects (all p > 0.10).

When comparing to Bitcoin we find underpricing to be the only factor to have a significant effect at 3 and 6 months (p <0.01). At 12 months the whole regression model is only marginally significant (F(7, 42) = 2.078, p < .10). Only ICO’s this month has a significant effect after 12 months and this effect is now negative instead of positive (β = -.131, p <0.01).

In conclusion we reject all our hypotheses. We do however see evidence for the opposite hypotheses of H1 and H2 in certain time frames, as these have a (marginal) significant effect on

ICO performance. With higher underpricing and more ICOs in a month leading to smaller underperformance of those ICOs. H3, H4, H5 and H6 show no significant ability to predict ICO

(20)

20

5. Discussion and Conclusion

5.1. Discussion of findings

Our findings show that only 14.6% of the seen variance in post-ICO underperformance can be explained by our researched factors. When compared to IPO research this seems to be on the low side. Chan (2014) finds an adjusted R2_{of 0.616 for his model, Coakley et al. (2007) find}

an adjusted R2 of 0.252 for their model and Ritter (1991) finds an adjusted R2 of 0.070. Throughout the years researchers seem to have found more factors which are able to explain the variance better. Further research into ICO underperformance could also find more factors which have a significant effect on the underperformance to build better models explaining this. We will discuss possible factors later in section 5.4.

Only two of the researched factors seemed to have a significant effect on post-ICO underperformance. They however showed the opposite of our hypotheses. With the tested ICOs higher underpricing led to smaller underperformance, this is the exact opposite of the research on IPOs by Coakley et al. (2007). A possible explanation for this observed phenomenon might be that the ICO market is in the middle of its hot period and is still too young. This could keep the investor sentiment positive leading to high longer term returns even after the extreme first day returns we see in our sample (mean = 267%). The newness of the ICO market also limits the time frame we can look at for our patterns. We tested for underperformance 6 months after the ICO but many scholars test for periods of many years giving the effects more time to play out (Chan, 2014; Coakley et al., 2007; Jain & Kini, 1995; Ritter, 1991).

The second significant effect was that of more ICOs in a month leading to smaller underperformance. This shows the opposite effect of the research on post-IPO returns during hot markets by Ritter (1991) and Alon & Gompers (1997). This might have the same explanation as the previous observation. It seems likely we are still in a major positive sentiment era for the ICO market and it takes more time for it to settle down. We also found some correlation between the number of ICOs in a month and underpricing. While underpricing has decreased over time, the number of ICOs has been increasing, this is also shown from the significant negative correlation (r = -.158, p < 0.05). These newer companies thus seem to have learned from the earlier underpricing and are better able to set their ICO price to raise the largest amount of money. This is also shows from the increasing amount of dollars raised by newer companies which also shows a significant correlation with the higher number of ICOs in a month (r = .461, p < 0.01). This seems to support the notion that the ICO market is in a hot market. This market shows a lot of positive investor sentiment driving up the prices and attracts more companies to have an ICO. This was also observed by Ljungqvist et al. (2006) when they studied hot IPO markets.

(21)

21

As with the white paper, having the source code available did not show significant effects on the underperformance of the ICO. Adhami et al. (2017) did find this had a significant effect on the funding success of the ICO but this does not seem to carry over post-ICO. We did find a correlation between white paper and code availability. Companies are thus likely to provide both. Like the white paper availability, the difference could be in what kind of source code is available. We did not differentiate between ICOs with only a simple contract as code, from those which have coded a whole stand-alone program. Our findings show that white paper and code availability do not show similar dynamics as the quality factors in the research of Coakley et al. (2007). They did however both have a positive correlation with the amount of USD raised during the ICO. It could be that this is because more recent ICOs are more likely to have a white paper and code available or these factors could lead to more investor interest.

Having a presale did not show a significant effect on the underperformance of the ICO. It did however show to have a positive correlation with the total amount raised during the ICO. This fits with the findings of Adhami et al. (2017) where they observe that companies having a presale are more likely to successfully reach their minimum funding goal. It did however not show similarities with the VC involvement as is seen with IPOs underperformance by Jain & Kini (1995). This could be because of the very different structures of the presales between ICOs. Some allowed all investors to participate others only let large investors like VCs get involved. The information on what kind of investors are getting involved is however very limited because the ICOs are not obligated by law to disclose this information. It is thus hard to differentiate between presales which may be hindering the measurement of the VC involvement effect.

Having a bonus scheme did not show a significant effect on the underperformance of the ICO. This could be because of different reasons. First, 68.6% of our sample had a bonus scheme, and the popularity of bonus schemes seems to be increasing over time. Having a bonus scheme thus seems to be becoming the norm. Second, while it did show to have a marginally significant effect on the funding success of an ICO (Adhami et al., 2017). This effect seems to be limited to the duration of the ICO itself. After the ICO investors were not more likely to sell of their tokens even with the high observed average first day returns of 267% in our sample. This seems to stroke with the observation of IPO underperformance during the dot-com bubble where there was an increase in selling pressure leading to underperformance for people that could get shares at a discounted price as was seen by Ljungqvist & Wilhelm (2003).

Our control variable concerning the total amount raised by the ICO did not show a significant effect on the post-ICO underperformance as expected. It did however show significant correlation to ICOs this month, white paper availability, code availability and presale availability as also described earlier in this section. This could be an interesting topic for future research concerning why some ICOs raise more money than others.

(22)

22

having a negative relation with post-ICO underperformance. Meaning more underpricing and/or ICOs during the month leads to smaller underperformance. These two however, only explain 14.6% of the seen variance in post-ICO underperformance. There seem to be other factors which we did not research that could also explain part of the observed variance in underperformance we will discuss those in section 5.4 concerning future research.

5.2. Practical implications

This thesis has several practical implications. First, it shows it is hard to predict future performance of ICOs based on certain company level factors which are present before and during the ICO. This makes it almost impossible for regulators to only restrict certain ICOs to protect investors from poor post-ICO performance. As the factors that lead to this poor performance are not known. This could also explain the stance of most countries to ban all ICOs, as this is the only way they can keep investors from possible financial harm. For investors the practical implications provide the base for a possible lucrative investing strategy which is currently practiced in the ICO market. This strategy is buying the tokens during the pre-sale or during the high bonus periods and selling them immediately on the first day of trading. The mean first day return in our sample is 267%, by spreading out your investment among multiple ICOs the downward risk could also be limited. This first day return is however declining as more and more investors are using this approach. This research also shows it is very hard to determine the longer term returns of an ICO while the ICO itself is taking place. This is due to the very limited explanation of variance in post-ICO performance by the factors which are present during the ICO. Lastly, our findings have practical implications for entrepreneurs who are looking to have their own ICO. It seems the current ICO market is very favourable for these entrepreneurs. They should be looking to launch their ICO during a month with a lot of ICOs which have high underpricing. This indicates a hot market which shows to have positive long-term effects on the returns of the ICO. It is also recommended to have a white paper and code available as it is common practice these days. Having a presale and a bonus scheme also seems to become the norm so going with this trend seems to be advisable. With these practical implications also come theoretical implications which we will touch upon in the next section.

5.3. Theoretical implications

(23)

23

(2006) and Chan (2014) provides some direction for a more investor sentiment based framework. We will provide some directions for future research in the next section.

5.4. Limitations and future research

This research has several limitations which could open areas for future research. First, because of the very recent development of blockchain technology and ICOs only a very limited theoretical base is present. Therefore, this research has tried to expand this theoretical base by trying to connect the IPO literature on underperformance to ICOs. Unfortunately, our results show this comparison cannot be made and we must reject all our hypotheses. When this field of research is more developed more specific clues to factors may be found which could lead to a model which is able to explain post-ICO underperformance with a higher certainty. This development could however take years as the current IPO literature has been developed for decades. When more research has been done into the current state of the ICO market and qualitative studies have paved the way by observing certain interesting dynamics within this market we can look at certain dynamics and make more accurate predictions to develop hypothesis.

Second, this research is based on a database with data collected from different online resources which may or may not be trustworthy. If ICOs ever get regulated by the government this could lead to an obligation for them to register and report their earnings and other financial data. This would provide us with a trustworthy data set that is controlled by government agencies. Our dataset only covers a short timespan by looking at post-ICO performance after a maximum of 12 months. Our measures also seem to be taken in the middle of a hot market. The irrational behaviour of investors during these hot markets could explain our lack of findings. Future research which extends this database to cover a longer timespan could potentially find more significant results because the mechanics have a longer time to work out and it allows hot markets to cool down. Most IPO underperformance literature looks at a time span of 5 or more years which could be a good start for future research.

Third, the scope of this research only covers 185 ICOs because these ICOs needed to be completed before 2018 to have a 6-month post-ICO data point. For our third hypothesis we only had 4 companies without a white paper, a larger database could relief these problems and be a better predictor for the whole population. Because of the explosive growth of the ICO market during 2018 a huge increase can be seen in the total amount of ICOs. ICObench (2018) for example mentions they currently track 3570 ICOs from 414 different countries. Future research could be focussed on building a larger database of over a thousand ICOs so higher significance can be reached.

(24)

24

perform better than others. Looking into more detail at the presale and the bonus schemes could also provide new insights by really looking for smaller differences in how they are structured during the ICO.

Fifth, by better looking at the 2 factors which do have a significant effect, namely underpricing and ICOs during the month, we can look for factors that influence these. This could lead to a better understanding of how these factors have an effect on the underperformance and why their effects are opposite of what we expected.

Finally, as has been mentioned before this research has only found 2 factors which are able to explain 14.6% of the observed variance in post-ICO underperformance. Future research will have to search for other factors or mechanisms to build a more complete model to explain post-ICO underperformance. As mentioned in the previous section the literature on investor sentiment and its effects on IPO returns could be an interesting place to start to find more factors which have an effect on post-ICO underperformance. We can also look at the correlation matrix of this research. This shows that the total amount of dollars raised is not related to the underperformance it is significantly correlated to the amount of ICOs during the month, white paper availability, code availability and presale availability. This could thus be an interesting factor for future research as it seems to have a lot of connections to other factors.

5.5. Conclusion

(25)

25

6. References

Adhami, S., Giudici, G., & Martinazzi, S. (2017). Why do businesses go crypto An empirical analysis of Initial Coin Offerings, (January), 19. https://doi.org/10.2139/ssrn.3046209 Aitken, R. (2017). Investment Guide To “Crypto” Coin Offerings Rating Blockchain Startups.

Retrieved January 19, 2018, from

https://www.forbes.com/sites/rogeraitken/2017/01/06/investment-guide-to-crypto-coin-offerings-rating-blockchain-startups/#76b02ccb121b

Alon, B., & Gompers, P. a. (1997). The Long-Run Underperformance of Initial Public Offerings, 52(5), 1791–1821.

Bitcointalk. (2018). Bitcointalk.org. Retrieved May 5, 2018, from https://bitcointalk.org/ Bitcoinwiki. (2018). Pre-ICO - Bitcoin Wiki. Retrieved January 19, 2018, from

https://en.bitcoinwiki.org/wiki/Pre-ICO

Booth, J. R., & Chua, L. (1996). Ownership dispersion, costly information, and IPO underpricing. Journal of Financial Economics, 41(2), 291–310. https://doi.org/10.1016/0304-405X(95)00862-9

Buterin, V. (2014). A next-generation smart contract and decentralized application platform. Etherum, (January), 1–36. https://doi.org/10.5663/aps.v1i1.10138

Chan, Y. C. (2014). How does retail sentiment affect IPO returns? Evidence from the internet bubble period. International Review of Economics and Finance, 29, 235–248. https://doi.org/10.1016/j.iref.2013.05.016

Coakley, J., Hadass, L., & Wood, A. (2007). Post-IPO operating performance, venture capital and the bubble years. Journal of Business Finance and Accounting, 34(9–10), 1423–1446. https://doi.org/10.1111/j.1468-5957.2007.02055.x

Coinmarketcap. (2018). Global Charts | CoinMarketCap. Retrieved January 19, 2018, from https://coinmarketcap.com/charts/

Coinschedule. (2016). Coinschedule - Cryptocurrency ICO Statistics. Retrieved January 19, 2018, from https://www.coinschedule.com/stats.html?year=2016

Conley, J. P. (2017). Blockchain and the Economics of Crypto-tokens and Initial Coin Offerings. Vanderbilt University Department of Economics Working Paper Series, (June). Cryptoslate. (2018). Cryptocurrency News, ICO Database, Coin Rankings and Blockchain

Events | CryptoSlate. Retrieved May 28, 2018, from https://cryptoslate.com/

Fama, E. F. (1998). Market efficiency, long-term returns, and behavioral finance. Journal of Financial Economics, 49(3), 283–306. https://doi.org/10.1016/S0304-405X(98)00026-9 Finder.com. (2018). Finder Cryptocurrency Predictions – May 2018. Retrieved May 26, 2018,

from https://www.finder.com/cryptocurrency-predictions

Iansiti, M., & Lakhani, K. R. (2017). The truth about blockchain. Harvard Business Review, (February).

(26)

26

Ico Stats. (2018). ICO Stats | Track ICO Performance. Retrieved January 19, 2018, from https://icostats.com/roi-since-ico

ICObench. (2018). ICOs rated by experts | ICObench. Retrieved May 28, 2018, from https://icobench.com/

Icowatchlist. (2018). ICO Statistics - By Blockchain - ICO Watch List. Retrieved May 27, 2018, from https://icowatchlist.com/statistics/blockchain

ICOwatchlist. (2018). Finished & Traded ICO Tokens - ICO Watch List. Retrieved May 28, 2018, from https://icowatchlist.com/finished

Iurina, A. (2017). Initial Coin Offering in Gibraltar - Case study: Calidumcoin, (December). Jain, B. a, & Kini, O. (1995). Venture capitalist participation and the post issue operating

performance of IPO firms. Managerial and Decision Economics, 16(1989), 593–606. https://doi.org/10.1002/mde.4090160603

Kaal, W. A., & Dell ’erba, M. (2018). Initial Coin Offerings: Emerging Practices, Risk Factors, and Red Flags, 0–21.

Kalla, S. (2016). What is a token sale (ICO)? - Smith + Crown. Retrieved January 19, 2018, from https://www.smithandcrown.com/what-is-an-ico/

Liu, M., & Forester, C. (2014). IPOs, Operating Activities, and IPO Underperformance. Journal of Business & Economic Studies, 20(1), 1–23.

Ljungqvist, A., Nanda, V., Singh, R., The, S., July, N., Ljungqvist, A., … Stoughton, N. (2006). Hot Markets , Investor Sentiment , and IPO Pricing, 79(4), 1667–1702.

Ljungqvist, A., & Wilhelm, W. J. (2003). IPO Pricing in the Dot-Com Bubble. The Journal of Finance, 58(2), 723–752.

Loughran, T., & Ritter, J. (1995). 02_The new issues puzzle. Journal of Finance, 50(1), 23– 51.

Loughran, T., & Ritter, J. R. (2004). Why Has IPO Underpricing Changed Over Time? Financial Management, 33(3), 5–37. https://doi.org/10.2139/ssrn.331780

Nakamoto, S. (2008). Bitcoin: A Peer-to-Peer Electronic Cash System. Www.Bitcoin.Org, 9. https://doi.org/10.1007/s10838-008-9062-0

Protocol Labs. (2017). Filecoin : A Decentralized Storage Network, 1–36.

Randolph, R. (2015). The new digital wild west: Regulating the explosion of initial coin offerings. Denver University Law Review, 92(4), 697–699.

Reuters. (2018). Europe’s venture capitalists embrace virtual currency craze | Reuters. Retrieved May 27, 2018, from https://www.reuters.com/article/us-crypto-currencies-venture/europes-venture-capitalists-embrace-virtual-currency-craze-idUSKBN1HO2NG Ritter, J. R. (1984). The Hot Issue Market of 1980. The Journal of Business, 57(2), 215–240.

https://doi.org/10.1086/296260

Ritter, J. R. (1991). The Long‐Run Performance of initial Public Offerings. The Journal of Finance, 46(1), 3–27. https://doi.org/10.1111/j.1540-6261.1991.tb03743.x

(27)

27

Roberts, J. J. (2017). SEC Cracks Down on Celebrity ICOs. Look Out Paris Hilton, Mayweather | Fortune. Retrieved January 19, 2018, from http://fortune.com/2017/11/01/bad-news-for-mayweather-the-sec-is-cracking-down-on-celebrity-icos/

Russel, J. (2017). Former Mozilla CEO raises $35M in under 30 seconds for his browser startup Brave | TechCrunch. Retrieved January 19, 2018, from https://techcrunch.com/2017/06/01/brave-ico-35-million-30-seconds-brendan-eich/ SEC. (2017). SEC.gov | Statement on Potentially Unlawful Promotion of Initial Coin Offerings

and Other Investments by Celebrities and Others. Retrieved January 19, 2018, from https://www.sec.gov/news/public-statement/statement-potentially-unlawful-promotion-icos

Smith+Crown. (2018). Front Page - Smith + Crown. Retrieved May 28, 2018, from https://www.smithandcrown.com/

Sontakke, K. A., & Ghaisas, A. (2017). Cryptocurrencies: A Developing Asset Class. Proceedings - IEEE Symposium on Security and Privacy, 10(2), 104–121. https://doi.org/10.1109/SP.2015.14

Swan, M. (2015). Blockchain Blueprint for a New Age. O’Reilly books. https://doi.org/10.1016/S0197-4572(81)80089-4

Teoh, S. H., & Wong, T. J. (2002). Why New Issues and High-Accrual Firms Underperform: The Role of Analysts’ Credulity. Review of Financial Studies, 15(3), 869–900. https://doi.org/10.1093/rfs/15.3.869

Tokendata. (2018). Token Data I News, data and analytics for all ICO’s and tokens. Retrieved May 5, 2018, from https://www.tokendata.io/advanced

Willett, J. (2013). The Second Bitcoin Whitepaper, 5.

(28)

28

7. Appendix A

LN Underpricing .548** .076 .485**

ICOs this month .016 .011 .117

Presale availability .026 .170 .012

LN USD Raised -.061 .057 -.083

R Square .242**

F 8.075

Regression, dependent variable LN 3-month ETH relative return ** = p < .01 * = p < .05

LN Underpricing .314 .248 .201

ICOs this month -.054 .040 -.202

White paper availability -.669 1.071 -.110

LN USD Raised .234 .179 .220

R Square .167

F 1.200

Regression, dependent variable LN 12-month ETH relative return ** = p < .01 * = p < .05

LN Underpricing .628** .087 .481**

ICOs this month .008 .013 .053

Bonus scheme availability .073 .180 .027

LN USD Raised -.079 .066 -.092

R Square .255**

F 8.647

(29)

29

LN Underpricing .483** .100 .345**

ICOs this month -.006 .014 -.038

Presale availability .176 .225 .062

LN USD Raised -.116 .076 -.127

R Square .150**

F 4.466

Regression, dependent variable LN 6-month BTC relative return ** = p < .01 * = p < .05

LN Underpricing .338 .266 .190

ICOs this month -.131** .043 -.428**

White paper availability -.599 1.149 -.086

LN USD Raised .163 .192 .135

R Square .257

F 2.078