Volume 23 Issue 1 Article 5 June 2021 An Investigation of Factors Determining the Token Value in the An Investigation of Factors Determining the Token Value in the Blockchain-based Early Funding Mechanism Blockchain-based Early Funding Mechanism Alfreda Š apkauskienė Vilnius University, Faculty of Economics and Business Administration, Vilnius, Lithuania, alfreda.sapkauskiene@evaf.vu.lt Simona Pakė naitė Vilnius University, Faculty of Economics and Business Administration, Vilnius, Lithuania, s.pakenaite@gmail.com Follow this and additional works at: https://www.ebrjournal.net/home Part of the Business Commons Recommended Citation Recommended Citation Š apkauskienė , A., & Pakė naitė , S. (2021). An Investigation of Factors Determining the Token Value in the Blockchain-based Early Funding Mechanism. Economic and Business Review, 23(1), 55-67. https://doi.org/10.15458/2335-4216.1005 This Original Article is brought to you for free and open access by Economic and Business Review. It has been accepted for inclusion in Economic and Business Review by an authorized editor of Economic and Business Review. ORIGINAL ARTICLE An Investigation of Factors Determining the Token Value in the Blockchain-based Early Funding Mechanism Alfreda Sapkauskien_ e*, Simona Pak_ enait_ e Vilnius University, Faculty of Economics and Business Administration, Vilnius, Lithuania Abstract The research employs WLS Regression for examining the main determinants of the ICO profitability in the crowd- funding stage. The variables are divided into three main categories: financial and technological aspects, and the ICO characteristics, with the aim of verifying which parts most influence the funds raised. The results imply that financial and technological aspects might indeed have an impact on the ICO profitability. The key factors covered are the open- source code availability and the preset hard cap. Overall, the econometric analysis discloses that the amount raised during the ICO is not affected by the availability of a white paper and pre-sales, even though some researchers argue differently. Keywords: Initial coin offering (ICO), Cryptocurrencies, Blockchain, Weighted least square regression, Ether, Bitcoin JEL classifications: G1, G15, O31 Introduction C ryptocurrencies are no longer a complete novelty, as the market capitalization of dig- ital coins has skyrocketed, increasing with it public awareness significantly. It is not only in- vestors who have great interest in virtual cur- rencies, but more and more also national governments, and just as much also policy- makers as are the U.S. Securities and Exchange commission, European Securities and Markets Authority, etc. The prevalence of the improved distributed ledger technology (DTL) and crypto- currencies has fostered the growth of a new phenomenon, called the Initial Coin Offering (ICO), as the new financing instrument for entrepreneurial ventures. Generally, ICO is defined as a decentralized method of funding, where blockchain-based organizations issue new tokens (that can be sold online or used in the future to obtain products, services or profits) in exchange for the preexisting cryptocurrencies (usually, Bitcoin or Ether) (Adhami et al., 2018; Huang et al. (2020)). ICOs are simply considered an alternative to the already existing methods of fundingasare,forinstance,Venturecapitalistsor conventional crowdfunding. Moreover, the major part of ICOs has the listing stage, which has attracted high interest from both investors and traders. The occurrence of ICOs has provided companies with direct and immediate early-stage crowdfund- ing, reduced costs and intermediation fees (Adhami et al., 2018; Fisch et al., 2019), eliminated geographical boundaries and implemented high liquidity for investors (Amsden & Schweizer, 2018). The prevalence of ICO has brought many benefits also for businesses, but has likewise challenged regulationauthorities,entrepreneurs,andinvestors. Received 28 September 2020; accepted 11 February 2021. Available online 15 June 2021. * Corresponding author. E-mail addresses: alfreda.sapkauskiene@evaf.vu.lt (A. Sapkauskien_ e), s.pakenaite@gmail.com (S. Pak_ enait_ e). https://doi.org/10.15458/85451.1005 2335-4216/© 2021 School of Economics and Business University of Ljubljana. This is an open access article under the CC-BY-NC-ND license (http://creativecommons. org/licenses/by-nc-nd/4.0/). Further, ICO contributors are able to easily avert regulation rules and costs of intermediation (e.g. exchanges) that are applied to businesses, issuing their securities to investors (Masiak et al., 2019). Companies initiating token sale campaigns can collect vast amounts of money with a limited extent of information and without any insurance to par- ticipants or investors of the project. Individuals are becoming more aware of what features and signals of the ICO campaigns are reliable, transparent, and expedient, as participants are mostly prospective customers and trustworthy originators. Neverthe- less,speculatorsstillappear,sincetherearenostrict and common rules applied for the token sales market, and coins are easily exchangeable for cryptocurrency or fiat money. The interest in the Initial Coin Offerings might really be newly arisen, but the amount of the empirical analyses on the issue of the crypto- currency economy is rapidly evolving. Researchers mostly focus on the success factors (Howell, 2020; Chen, 2019; Fisch et al., 2019; Huang et al., 2020; Adhami et al., 2018), elements of value (Masiak et al., 2019; Felix & von Eije, 2019; Catalini & Gans, 2018), legal and regulation aspects (Zetzsche et al., 2018; Chanson, Gjoen, et al., 2018, b), as well as a general overview of the ICO market (Chanson, Gjoen, et al., 2018, b; Coinbase, 2019). Despite everything, there are still many questions that recent literature is only starting to undertake. In 2020, the most analyzed topic turned out to be to- wards the ICO market returns, market efficiency, and information asymmetry (Fisch& Momtaz, 2020; Domingo et al., 2020, etc.). However, there is a lack ofimplicationscomposedinregardsofthepost-ICO market and pricing. In addition, determination of theaspectsinfluencingtheICOprofitabilityhelpsto improve not only the ICO campaign performance and the settlement of success, but also market transparency. As tokens in the ICO project do not have any present value, no pricing mechanisms are applied, thus making an estimation of the time frameworkoftheprofitabilityoftheprojectdifficult. In addition, coins issued during an ICO become valuable only, when the network of the campaign matures. Many substantial forces lead to constantly increasing the demand for blockchain-based early funding, and qualifying the attributes that cause the increased value for tokens is one of the main forces thatinducesbuyercompetitionandincidenceofthe ICO projects. This paper outlines an empirical analysis of what creates value for the ICO projects. The purpose of theresearch isto investigatethemainelements that are most valued by the Initial Coin Offering investors in the crowdfunding stage. A substantial efforthasbeenmadetoprepareamethodologyfora sound and reliable analysis. Based on the data availability, examined literature, and trends in the empirical financial studies, the Weighted Least Squares Regression was considered to be the most suitable econometric analysis to examine ICOs. The analysis in addition eliminates the hetero- scedasticityandisapplicableforsmalldatasamples. A lot of attention was paid also to the assumption testing, qualitative data collection, and dataset suitability verification. Moreover, the dataset con- sists of the dependent variable (total funds raised) andthreegroupsofindependentvariables:financial aspects, technical aspects, and ICO characteristics. The relevancy of the regression analysis and statis- tical tests of the observation ability to predict the outcome is estimated using R software. Research is organized as follows, namely it starts with (1) theoretical background, continues with (2) methodology of research models, followed by (3) practical model employment, and finishes with (4) result interpretation. The first part of the research covers an overview of the ICO market, presents the processes and technical operating principles of an ICO,analyzeschallengesandopportunitiesbrought by the ICO occurrence (SWOT analysis), and in- cludes a discussion, as well as a comparison, of the diverse research findings. The second part qualifies the methods of the research analysis (WLS regres- sion), where in the methodological part the description, general considerations, assumptions, and testing are covered. The last part represents the results of the investigation and provides consider- ation of the research findings, highlighting also the limitations of the research. 1 Background and literature review The digital transformation of the company is not onlyfocusedonthenewtechnology,butalsorelates to the changes in the company business models, structure, and processes (Tomat & Trkman, 2019). Therefore, ICOs might be considered one of the ways for the company's digitalization. In general, Initial Coin Offerings are identified as an open and direct way for early funding, promoted by organi- zations and entrepreneurs in order to increase financing through cryptocurrencies in exchange for issued so-called“tokens”.Thesetokenscanthenbe sold or used later on to obtain profits, products, or services (Adhami et al., 2018; Chen, 2019; Fisch et al., 2019). Moreover, all ICOs are executed by using blockchain technology and are initially launched to fund technology-based projects. The latter aspects 56 ECONOMIC AND BUSINESS REVIEW 2021;23:55e67 are also the main ones that differ an ICO from an Initial Public offering. The other important feature thatemphasizesthedifferencebetweenanICOand the other crowdfunding methods (crowdfunding, Initial Public Offerings, Venture Capital, Angel In- vestors)isthatduringtheICO,investorsdonotbuy the underlying asset, but instead buy the money supplyofthefutureproject.Iftheprojectgrowsand the technology is well applicable, then the value of thetokenspositivelycorrelateswiththevalueofthe company. However, at the beginning of the block- chain-based crowdfunding, campaign denomina- tion of tokens is always equal to zero, and the originator's issued coins become more valuable only when the network of the ICO project matures. In the initial stage, the value of tokens strongly dependson the users'perceivedfutureutilityof the network. During the pre-sale stage, major investors are risk takers or those who firmly believe in the campaign. In the outset, the value for tokens is given byearlybirdsandtheir willingnesstopayfor theproject.Overtime,morecontributionisgivento the ICO campaign and the company begins to materialize, eventually becoming able to deliver to end users. 1.1 Process and stages of an ICO The process of an ICO is complex and only cryp- tocurrency holders can take part in the ICO project, with certain exceptions as is, for instance, the pre- sale stage. Before the creation of the ICO campaign, thefinancing-seeking company produces two smart contracts that are deployed on the blockchain plat- form to determine the key parameters of the token sale project. Despite the fact that up to 90% of the ICOs in the market are based on the Ethereum blockchain, when initiating decentralized crowd- funding, the company can nevertheless build its own blockchain. However, to create its own block- chainforanICOisexpensiveandverydifficultfrom the technical perspective. Amsden and Schweizer (2018) also indicate that the implementation of its own decentralized technology would require the company to establish an ongoing incentive mecha- nismtoattractusers inordertoverify theledger.At any rate, a smart contract that is located on the blockchain defines the hard and soft caps, the quantity of the tokens, period of the project, etc. In addition, the additional smart contract is created for token distribution and transfers that can be executed after the launch of the project. Moreover, funding is not transferred directly, after the pay- ment the subsequent process is fully automated by the predefined rules in the smart contracts. In other words,theICOcampaignautomaticallyreceivesthe access to the funding from the ICO Smart Contract and investors automatically get their portion of the issued tokens from the Token Smart Contract (Chanson, Gjoen, et al., 2018). In general, the literature identifies three stages of an ICO, namely the pre-ICO stage, the main-ICO stage,andthepost-ICO(showninFig.1asPre-sale, Disclosure of token sales, and Post-ICO stage). At the pre-ICO stage, originators of the ICO project disclose white paper to provide the information to potential investors about the key aspects of the project. The white paper contains information, as among others the prime idea, technical details, membersofthecompanywhoinitiatedtheICO,the number of tokens, and their target prices (Zetzsche et al., 2018). White papers generally do not provide anyguidelinesnorstandardsonhowtobefilledand disclosed, therefore, certain electronic documents are more detailed than the others, which, in accor- dance to Fisch (2019), causes information asymme- try in the ICO market. Besides, some entrepreneurs announce the advisory board in order to show the quality of the campaign and even employ experts (from legal, marketing, information technology de- partments) to run the ICO (Chen, 2019). Further- more,inthepre-ICOstage,pre-salesareinitiatedin order to examine the market readiness and accep- tance levels. The initiators of the ICO then provide potential investors with the possibility to participate in open or private pre-sales. Pre-sales increase the interest in the ICO, thus attracting greater attention from the public and enhancing the willingness to invest in a particular ICO (Masiak et al., 2019). Adhami et al. (2018) agree that pre-sales are one of the major factors that fosters the higher probability of the ICO project success. At the pre-sales, in- vestors are able to use fiat currency that helps to simplify the process for non-users of cryptocurren- cies and accelerates the accumulation of soft cap (Masiaketal.,2019).Domingoetal.(2020)arguethat pre-sales are in fact related to the ICO success, however, pre-sales negatively affects the project's returns. During the main stage, the company seeks to collect a predetermined hard cap and exchange the issued tokens to cryptocurrencies. At this stage, to collect more funding, initiators provide bonus schemes for participants of the campaign, which means that early birds receive more tokens for the samepricethanotherinvestors(Masiaketal.,2019). At the post-ICO stage, originators of the token sales exchange cryptocurrencies to fiat money to reach their goals of the project, which is to make an in- vestment to develop the product or service, expand the business, etc. ECONOMIC AND BUSINESS REVIEW 2021;23:55e67 57 Additionally, some ICO companies offer the op- portunity to trade their tokens on a secondary market (shown in Fig. 1: Disclosure for listing). This latter opportunity is comparable with an IPO, however, in the latter tokens/coins are traded only oncryptocurrencyexchanges(Chen,2019;Masiaket al., 2019). According to the authors, during the listing, the main factors that influence the token pricearethecompany'sdisclosuresonsocialmedia, code updates, and token sale performance in the main stage. Moreover, Chen (2019) research ascer- tains that the value of the listed tokens is very sen- sitive to low credibility and easy-interpretable signals. Investors consider the listing of tokens as a positivecause,sinceICOswithaplannedsecondary market trading possibility tend to collect more capital.Inaddition,theliquidityinthetokenmarket is considerably high. However, to seek admission to trading, the company must as a prerequisite be lis- ted on the exchange, and the preparation for listing takes time and can last up to several months. 1.1.1 Types of Tokens The very first token sale was named Initial Coin Offering. However, since the phenomenon evolved, companies have been issuing other types of sales. The currently most prevalent are three sorts of to- kens, namely currency, equity, and utility (Fisch, 2019; Masiak et al., 2019). Utility tokens can give access to the service or product created by the venture (e.g. EndChain), equity tokens represent a claim on the issuer's asset or grant contribution to the funding development (e.g. DAO), and coin to- kens are used as the medium of exchange (Adhami et al., 2018; Fisch, 2019). The issue of the different types of tokens increases the attractiveness of the ICO project and is one of the main distinctive sub- stantial features among different token sale cam- paigns. Nevertheless, Adhami et al. (2018) complain that the marketing and usage of tokens only slightly influence prosperity in the ICO mechanism, espe- cially when compared with the presale's impact on the ICO success. Furthermore, all ICOs are related to the blockchain technology, through which P2P exchangesaremadeinthefundingmechanisms.As a base, blockchain could be widely adopted in business, since every technology includes different type of data that is stored in the blocks. The type of data depends on the sort of network, e.g. Bitcoin blockchain includes information about the sender, receiver, and the number of coins traded, while the technology provides the ability to distribute data securely between non-trusted parties without intermediation (Zaninotto, 2016). 1.2 ICO problems analyzed in the literature Researchers have mostly analyzed aspects that relatetothedeterminantsofsuccessandvalue,legal aspects, information disclosure, comparison Fig.1.InitialCoinOfferingprocessandstages(includinglistingstage).PreparedbytheauthorsinaccordancewithChen(2019),Masiaketal.(2019), Chanson, Risius, and Wortmann (2018). 58 ECONOMIC AND BUSINESS REVIEW 2021;23:55e67 between diverse funding methods and ICO, haz- ards,andopportunities,andmore.Fischetal.(2019) and Masiak et al. (2019), however, analyze the mo- tives of ICO investors. They point out that motives can be broken down into three categories, namely ideological,technological,andfinancial.Inaddition, the result of their analysis shows that ICO contrib- utors are mostly driven by technological motives, becausetheyhaveaninterestandseehighpotential in blockchain-based projects. Further, Chen (2019) and Fisch (2019) analyze the asymmetry in the blockchain-based token sale market. White papers generally do not provide any guidance, standards, nor agreed regulations on how to be filled and dis- closed, and as a result, some electronic documents are more detailed than the others and can be interpreted differently by every individual. The latter issue, in accordance with the authors, causes information asymmetry in the ICO market. More- over, Huang et al. (2020) analyze the legal and reg- ulatory aspects of ICOs. The main concern of their research is why in one countrytoken sales are more prevailed than in others. As a result, ICO initiators are more interested in the countries that have well- developed financial markets, as well as a clear legal and regulatory framework for token sales. In addi- tion, Haddad et al. (2019) underline that financial innovations appear more often in the countries that have more secure internet service providers and digital technologies. On the contrary, Huang et al. (2020) argue that advanced technologies are not enough for an ICO spread, but that well-developed investment-based crowdfunding platforms are also crucial. An et al., (2020) add their finding that countries with investor protection have more developed ICOs. Chanson, Gjoen, et al. (2018) and Barsan (2017) emphasize that an implementation of regulation in the ICO market could cause law con- flicts, due to the absence of geographical bound- aries. Further, Fisch (2019), Fisch et al. (2019), Adhami et al. (2018), Amsden and Schweizer (2018), Felix and von Eije (2019), and Chanson, Gjoen, et al. (2018) analyze the ICO success factors, which is otherwise already the most broadly examined topic in the literature. Authors indicated argue that suc- cessisnotaffectedbytheavailabilityofwhitepaper, but rather by the set of open-source codes intro- duced for the ICO project. Fisch (2019) generally agreesthatanincreasedamountoffundingishighly related to the quality of code. In addition, the researcher indicates that the success in an ICO is associated with a credible commitment to the ICO, as well as with the quality of information disclosure signals. Furthermore, Adhami et al. (2018) show in their analysis that success probability increases, when the campaign not only collects earlier find- ings, but also depends on the structure of an ICO. 1.3 SWOT analysis of ICO campaigns The prevalence of Initial Coin Offerings around the world brings many benefits to the business world,butlikewiseimpartsplentyofchallengesand risks for market authorities, enterprises, investors, and more, identified in the SWOT analysis (see Table 1). The ICO participants can easily avoid regulation rules and costs applied to businesses, issuing their securities to investors in the exchange markets (Masiak et al., 2019). However, ICOs are controversial, as ventures implementing token sale campaigns could collect huge amounts of money without any insurance to contributors, investors, and provide limited data. One important aspect is that lack of regulation increases investment risk (Fisch et al., 2019). Equally important is that tokens in the ICO mechanism do not have current value, which means no pricing mechanism is applied and projectsareveryspeculative, givingahighpotential for fraud (Chen, 2019). Therefore, a few business ideas in the ICO market actually materialize. Typi- cally, companies initiating IPO already have an actual product, while ICO companies have only an Table 1.SWOTAnalysis ofInitial CoinOfferings. Preparedby the authorsin accordancewithAdhamiet al.(2018), Fischetal. (2019),Chen(2019), Huang et al. (2020), Masiak et al. (2019). STRENGTHS - Participants of ICOs provide direct and rapid funding for ventures. - Lower costs due to intermediaries and absence payment. - ICOs are openeno strict time for investment; availability for early contribution agents. WEAKNESSES - No pricing mechanism specified for token sales. - Information asymmetry exists between external investors and entrepreneurs and is especially heavy in the cryptocurrency market. - Lack of transparency in the ICO market due to the absence of mandatory disclosures. OPPORTUNITIES - Tokens can be traded on the secondary marketehigh liquidity. - Lower competitioneallow a potentially easier way to collect funding. THREATS - Lack of regulations increases investment risk. - Lack of value determination leads to a highly speculative market and high potentials for fraud. ECONOMIC AND BUSINESS REVIEW 2021;23:55e67 59 idea of the product or service, altogether making it difficult to assess the profitability of the project and the period when the project will start to produce returns (Amsden & Schweizer, 2018; Chen, 2019; Masiak et al., 2019). In the same way, high infor- mationasymmetryheavilyoccursintheICOmarket (Chen, 2019). As a result, the ICO market lacks transparency. Nevertheless, there remain many reasons why accepting innovative technologies is importantforbusiness.Adhamietal.(2018),Fischet al. (2019), and Amsden and Schweizer (2018) emphasize the main causes, which are namely that by adopting DTL, a business could reduce costs of fundraising and avoid intermediaries, besides, token mechanisms allow building a post-ICO mar- ket for their investments with high liquidity, they avoid geographical boundaries, have the open- source access to capital, and much more. Not only the latter facts, but also the authors themselves indicate that ICOs are less expensive, include fewer parties, and are an easier method to collect funding incomparisontoangelinvestorsorventurecapitals. Additionally, ICOs involve nonprofessionals by providing an easier way of participation in startup financing, hence increasing greater liquidity and reducing monitoring costs (Masiak et al., 2019). On the other hand, those investors may just follow other contributors, without taking into consider- ation and assessment any other facts without their own experience, which may lead to irrational herding behavior in the ICO markets (Masiak et al., 2019). Anyhow, investors of ICOs provide the com- panywithearly-stagefundingthatisavailabletothe venture directly and immediately. In addition, to- kens can be traded on a post-ICO market to raise funds, and the liquidity is considerably high (Fisch et al., 2019). Moreover,asthereistheabsenceofregulationsin the ICO market, no restrictions are applied to in- vestment and marketing (Amsden & Schweizer, 2018),whichleadstoeasierandfasterpreparationto collecting funds. Although, no regulations are applied for information that should be disclosed, and even more injurious is the fact that no one su- pervises the information that is disclosed. This may lead to counterfeiting of the project in order to collect more money. Due to an unregulated envi- ronment and lack of participation of parties with good public reputation, ICO contributors may be deluded by fraudulent projects (Chod & Lyandres, 2018;Zetzscheetal.,2018).Atthesametime,anICO helps to build a community of the campaign before even introducing the actual product, which helps the originators to realize the project quicker and with conditions that are more favorable. 2 Hypotheses, data and methodology The second chapter of the research describes the structure and the process of the statistical research model by providing a detailed plan of the method- ologyapplied. Asubstantialefforthasbeenmadeto prepareamethodologythatwouldproduceasound and reliable analysis. Based on the data availability, examined literature, and the trends in the empirical financial studies, the WLS Multiple Regression method was considered as the most suitable for examining the ICO value determinants in the crowdfunding stage. Below, the selected methodol- ogy is acknowledged by providing assumptions, formulas, variables, samples, time horizons, and hypotheses. 2.1 Method, time horizon and sample size As ICO historical data are short and the values of the dependent variable have great differences, the dataset violates the homoscedasticity assumption. Therefore, the Weighted Least Squares Regression analysiswasselectedtoevaluatetheinfluenceofthe chosen predictors for the total amount raised in an ICO and to examine the main factors that cause the ICO profitability. WLS regression attributes each observation with a weight that is based on the varianceof itsfitted value, hereby reducing thesum of the weighted squared residuals and eliminating the heteroscedasticity (Garson, 2013). WLS regres- sioncanbeusedforlinear,aswellasnonlinear,data in the parameters and is an efficient technique for small data samples. The final presentation of the regression model formula is provided as per below: Y i ¼ b 0 þb 1 X i þ…þb n X n þe i ð1Þ here: - b 0 e intercept; - b n e coefficients; - e i residualN 0; d 2 w i ; - ne number of observations; - X i e predictor The WLS estimates of b 0 and b n : Swðb 0 ;b n Þ¼ X n i¼1 w i y i b 0 b n X 1 2 ð2Þ where w i areinverselyproportionate,namely(1)the data points with lower variation were assigned higher weights, and (2) the data points with higher variation were assigned lower weights. After that, WLS is given as: 60 ECONOMIC AND BUSINESS REVIEW 2021;23:55e67 b 0 ¼ y w b 1 x w ð3Þ b 1 ¼ P w i x i x w y i y w P w i x i x w 2 ð4Þ where y w and x w are the weighted means. Source: mcmaster.ca 2.1.1 Time horizon ThefirstICOwasinitiatedin2013,however,itwas only in 2015 that ICOs started to increase by gath- ering more attention from society and investors. Accordingly, the data for the model were collected in the period from 2015 to 2020, where most cases were taken from the year 2017. As a result, the period of the regression analysis model is 5 years. 2.1.2 Size The amount of observations in the dataset is calculated by using G*Power software (Fig. 2). Under the confidence level of 95% (as this is the standard of empirical researches), the minimum required sample size is calculated to be 110. How- ever, in order to conduct a more reliable research model, 217 observations were included in the WLS regression, although some observations were later removed when implementing the model, which is also described in the result section. The analysis includes 217 ICO projects, where eachisdescribedbytheprojectname,thewebsiteof the project, the trading symbol, the crowd sale duration, and the white paper. Only the completed ICOs are included in the analysis, due to a lack of information and recordson the incomplete projects. In addition, the failed ICOs do not fulfill the criteria to be selected for this analysis and are thus excluded.Asthecalculatedminimumsamplesizeis 110 out of 614, projects were chosen by using strat- ified random sampling, which is one of the proba- bility sampling techniques. In stratified sampling, firstly, the population is divided into homogeneous groups (called strata), based on particular charac- teristics, and a sample is then randomly taken from each stratum (Ackoff,1953).The614projectschosen for our research were divided into seven groups under the amounts of funds raised, and from them 31 projects were then randomly selected from each stratum. This approach is applied in order to include all important sub-populations into the model (Taherdoost, 2016) and have at hand all the levels of the projects for a truly precise analysis. 2.2 Variables Most common regression analyses include two types of variables, namely the dependent variable and predictors. In particular, this multiple regres- sionanalysismodelconsistsof1dependentvariable and 15 predictors. The total funds raised are the dependent variable of this research and stands as the measure of how tokens are valued in the ICO market. This variable was chosen, because of the analyzed literature stating it as the most suitable parameter, considering the design and objectives of the research. The explained variable is expressed in U.S. dollars and is continuous. The independent variables include three types of information: (1) financial aspects, (2) technical aspects, and (3) pre- determinedICOcharacteristics.Theselectedgroups of variables were chosen in order to examine what factors determine the greater value of tokens, i.e. which tokens are most valued in the market be- tween participants. Every group of predictors was composed with regards to the analyzed literature, Fig. 2. Minimum Required Sample Size with Confidence Level of 0.95. Prepared by the authors, using G*Power Software. ECONOMIC AND BUSINESS REVIEW 2021;23:55e67 61 basedonthecommontrendsinthemarket,together with the considerations of the authors, while following the ICO investor news. The financial group of the variables consists of a hard cap, mini- mum contribution size, and soft cap (determined or not). The initial ICO characteristics that might have a significant influence on the total funds raised are token type, the total supply of issued tokens, pub- licly available token supply (estimated in percent- age), pre-sale existence, bonus scheme availability, and ICO duration. The last group of predictors in- cludes white paper availability, open-source code availability, and criteria which determine whether a particularICO accepts cryptocurrencies,fiatmoney, or both as a payment method. All independent variables are indicated in Table 2. The dependent variable, as well as the five pre- dictors (hard cap, minimum contribution size, total issued supply, public supply, and ICO duration) are all continuous. The other 10 independent variables are the so-called categorical variables. The used informationiscollectedfromTokendata(2019),ICO Drops(2020),andICOrating(2020),whiletheopen- source code was found in Github (2020). 2.3 Assumptions and hypotheses In statistical models, parametric indices usually deliberate certain characteristics about the data, model suitability, and reliability. Any nonconfor- mity of the assumptions obtained could cause an inaccurate interpretation of the findings. In some cases, exceptions might be made, but if so, they must be highly substantiated. In accordance with the analyzed literature (Pallant & Manual, 2010; Whitcomb,2012;Zaid,2015),theassumptionstested in our research are as follows: correlation, normal distribution of errors, multicollinearity, and non- correlation between the dependent variable and error terms. The acquired assumptions are then checked and verified by using Descriptive Statistics, Scatter plots, Variance Inflation Factor and Toler- ance levels, Kendall correlation matrix, R squared, and ANOVA test. Moreover, the influential points can greatly affect the slope of the WLS regression function, therefore, it was important to detect and remove any outliers from the sample. The robustness regression model wasatthetimeconsideredoneofthesolutionsfora skewed data analysis. However, the latter proved not applicable due to the singularity issue, as many categorical independent variables are included in the analysis as predictors. Wherefore, the squared Mahalanobis distance approach was selected as the most suitable classical way for multiple linear regression models. The Mahalanobis distance is a multivariatedistancemetricthatestimatestherange betweenapointandadistribution.Thisdistanceisa highly applicable measure for not only insanity detection and classification of highly imbalanced datasets, but also other unfitted cases. 2.3.1 The hypotheses of the study The general hypotheses are expressed in a theo- retical way, to test what determines the value of the ICO project in the crowdfunding stage. The main aspirationistoidentifyhowpredictorsinfluencethe dependent variable: how much the variance in the Table 2. Predictors of the regression analysis model. Prepared by the authors. Variable Description Type FINANCIAL HARD_CAP Hard cape the maximum amount that can be collected in ICO crowdfunding (USD) Continuous MIN_CONTR Minimum allowed contribution in ICO crowdfunding (USD) Continuous SOFT_CAP If the soft cap (minimum amount required to complete the project) of ICO is reached Binary ICO CHARACTERISTICS TYPE_TOKEN Type of tokens issued in ICO (utility or other) Binary T_SUPPLY Total supply of issued tokens in ICO (units) Continuous PUB_SUPPLY_PERC Supply available for investors in ICO crowdfunding (%) Continuous PRE_SALE Pre-sale availability in ICO Binary BON_SCH Bonus scheme availability Binary ICO_DUR ICO duration (time period between the start of token issue and end/listing stage) Continuous TECHNICAL WHITE_AV White paper availability Binary OPS_COD_AV Open-source code availability Binary OWN_BLOCK If ICO is based on own blockchain or on an already existing one (usually Ethereum) Binary CRP_ACC If ICO accepts cryptocurrencies Binary FIAT_ACC If ICO accepts the cryptocurrencies and fiat money Binary BOTH_CURR_ACC If ICO accepts both types of payment (cryptocurrencies and fiat money) Binary 62 ECONOMIC AND BUSINESS REVIEW 2021;23:55e67 total funds raised can be explained by the chosen predictors. Since the dependent variable stands for how valuable the token sale project is, three hy- potheses of the analysis are formed regarding the ICO crowdfunding stage and are presented below. 1) The first hypothesis refers to the first group of variables: H 1 : Financial determinants have a significant influence on the total funds raised. 2) The second hypothesis refers to the second group of variables: H 2 : The initial ICO characteristics have a sig- nificant influence on the total funds raised. 3) The third hypothesis refers to the third group of variables: H 3 : Technological aspects have a significant in- fluence on the total funds raised. As the methodology used in the research has already been discussed in the paper, the imple- mentation of the WLS Regression analysis for the token value examination in the ICO crowdfunding stage is presented together with the results in the following chapters. 3 Results The WLS Multiple Regression analysis was cho- sen in order to check how well the group of the selected independent variables (financial, technical, andpredeterminedICOcharacteristics)wereableto predict the stress levels of the explained variable (total funds raised). The main aim of using the regression model was to investigate how much unique variance of each of the predictors was explainedinthedependentvariableoverandabove other predictors. In the first place, overall 217 ob- servations were gathered, however, some data points were excluded from the sample during the assumptions check, which left us with an estimated sample size of 110. The first tested assumption was the normal distribution and linear relationship be- tween the dependent variables and predictors. A Scatter plot was employed at this point to check, if data were linearly related and normally distributed. The result shows that the sample was following a linear relationship, however, some points were in a great distance from the rest of the data, which re- veals that some outliers exist in the data sample. After the Mahalanobis distance was applied, 16 observations were identified as influential points. Therefore,outlierswereeliminatedfromthedataset and the regression equation was estimated without the influential points. A Scatter plot without influential points shows that the linear relationship between the variables does exist but is not perfect, as observations of the model are spread near the line, with some deviations here and there. The skewed points specify that the data set had some discrepancies, which identify non-normality. Nevertheless, in real life, data are not usually nor- mally distributed, and this dataset was well suited for the chosen regression method. Moreover, as the data of the dependent variable have a huge variety of values, the Weighted Least Squares Regression analysis with the standard deviation function was chosen in order to avoid heteroscedasticity bias. As discussed inthe theoreticalpart, this method is well applicable for moderate datasets and provides an optimized estimation and different types of statisti- cal intervals (Croarkin et al., 2006). 3.1 Descriptive Statistics and dataset adjustments Table 3 describes variables by involving mean, standard deviation, and the total number of obser- vationsusedinthemodel.Themodelconsistsof201 observations in total. The mean is the estimated central value of a group of numbers (the average), whilethestandarddeviationquantifiesthevariation (or dispersion) of the dataset. Table 4 shows just how well the gathered dataset fits the analysis, as it indicates the relationship be- tween the model and the dependent variable. The explained variable's total variation is estimated by its variance. This proportion is expressed by the adjusted R squared, standing at 0.308, and shows the corrected value of the R Square, which provides Table 3. Weighted Least Squares Regression: Descriptive Statistics. Prepared by the authors (R software output). Descriptive Statistics Mean Std. Deviation T_FUND_RAISED 18149509.50 1.713 WHITE_AV 0.94 0.000 ICO_DUR 29.28 0.000 T_SUPPLY 4.3496730 2.9825170 PUB_SUPPLY_PERC 0.5295 0.00000 OWN_BLOCK 0.15 0.000 OPS_COD_AV 0.76 0.000 TYPE_TOKEN 0.95 0.000 BON_SCH 0.67 0.000 PRE_SALE 0.52 0.000 FIAT_ACC 0.14 0.000 CPRV_ACC 1.00 0.000 BOTH_CUR_ACC 0.14 0.000 HARD_CAP 16936498.22 1.784 SOFT_CAP 0.38 0.000 MIN_CONTR 97.3752 0.00003 ECONOMIC AND BUSINESS REVIEW 2021;23:55e67 63 better estimates for the true dataset. The number indicatesthat30.8%ofthevarianceinthetotalfunds raised are explained by the regression equation of this model. And even though the result is not very high, the model is assumed valid as the correlation between the predictors and the explained variable exists. After the elimination of some predictors, the ANOVA test was selected to check the significance of the results. Before completing the test, the hy- potheses set were concluded, meaning that the null hypothesis indicatesthat all b j are equal to zero and that there is no statistical significance in the model. H 0 :b 1 ¼b 2 ¼…¼b n ¼0 H A :b j s0;j¼1;2;…;n As seen in Table 5, the ANOVA test shows that the analysis reached the required level of signifi- cance (p < 0.05), thus rejecting the null hypothesis. As evident from the table, the high correlation coefficient between the explained variable and predictors has only two inputs: OPS_COD_AV (0.370; condition: >0.3) and HARD_CAP (0.419, condition: >0.3). Consequently, only two variables are assumed to have an impact on the total funds raised, and it is the same variables that explain the 30.8% variance independent variable. 3.2 Hypotheses verification Finally, all assumptions of the regression analysis are met and the model is confirmed as trustworthy and reliable. Provided that the analyzed measures indicate the statistical significance of the model, the set hypotheses should be revised. Considering that only the operational code availability and predeterminedhardcapwerehighlycorrelatedwith the dependent variable, only the first and third hy- potheses can be confirmed (see below). H 1 : Financial determinants do have a significant influence on the total funds raised. H 3 : Technological aspects do have a significant influence on the total funds raised. Regarding theICO specialties, thehypothesishad to be rejected, which indicates that there is neither any statistically significant influence of the ICO characteristics for total funds raised, nor that the financial and technological aspects are more influ- ential than the ICO characteristics, i.e. in this particular model of the collected dataset. 3.3 Coefficients and formula After the hypotheses of the model were sorted out, the final WLS equation could be written. Table 6 specifies the coefficients (b), which show how much of a unique contribution is provided by each predictor in explaining the dependent variable, as well as establish the strength and direction of each independent variable'si n fluence. Table 6 reveals that the standardized coefficient (b 1 ) of Hard Cap is 0.3630(Sig.¼0.0000)andhasapositiverelationwith the total funds raised of the ICO in the crowd- funding stage. Hard Cap is considered to have the highest in- fluence on the dependent variable, as it has the highest correlation coefficient. As a result, the equationthatisusefulforpredictingthevalueofthe dependentvariable(Y)forgivenvaluesofpredictors (X) is concluded below. Y¼22190184:5087þ0:3630X 1 þ3492349:7696X 2 ð5Þ where: - the intercept is (b 0 ) ¼ 1.473, which indicates the value of the explained variable, when all pre- dictors (Operational code availability and Hard cap) are kept equal to 0. - X 1 is a Hard cap with b 1 equal to 0:3630;which shows how much the total funds raised vary in the model, when X 1 changes by one unit. - X 2 is Open-Source Code Availability with b 2 equal to 3492349:7696;which shows how much the total funds raised increase, when the open- source code of the ICO project is available to investors. The standardized coefficient (b 2 ) of Open-Source Code Availability is 3492349.7696 and has a positive Table 4.ModelFit: WeightedLeastSquares Regression. Preparedby the authors (R software output). Model Summary Model R R Square Adjusted R Square Std. Error of the Estimate 1 0.653 0.346 0.308 1.476109 Table 5. Weighted Least Squares Regression: ANOVA Test. Prepared by the authors (R software output). ANOVA Sum of Squares Df Mean Square F Sig. Regression 179.526 13 13.81 6.338 0.000 Residual 407.454 187 2.179 Total 586.981 200 64 ECONOMIC AND BUSINESS REVIEW 2021;23:55e67 relationship, just as b 1 . The latter predictor had a lower correlation with the explained variable than the Hard Cap. Nevertheless, a unique contribution of the variance in the dependent variable proves highly important and has statistical significance (Sig. ¼ 0.000). 4 Conclusions This research model uncovers the characteristics that are most valued by the ICO investors and predetermine higher funds that are raised in the blockchain-based projects. In accordance with the first research model, the key factors are the open- source code availability and the preset hard cap, meaning the greatest amount of money that can be collected during the ICO crowdfunding. Unlike the already concluded analyses, this research includes variables that are broken down into different groups, where the selection of the projects is done by using a stratified sampling technique. The econometric analysis therefore discloses that the amount raised during the ICO is not affected by the availability of white paper. Investors might not value white paper, as it does not have any certifi- cation and requirements on how it should be composed, nor is it audited. In the event of that, transparency and the quality of information are not ensured, leaving the white paper a medium for spurious interpretations and falsification. On the contrary, a set of codes of blockchain projects is highly and positively valued by the ICO contribu- tors. Even the availability of a partial set of code is assumed to be a proof-of-concept. However, code availability is more valued by the professional ICO investors, while non-professionals depend mostly on white paper. Besides, the type of blockchain of the project is not considered an important charac- teristic in the ICO mechanism, as the major part of the existing blockchains are Ethereum, with only a few of them being unique (created on own block- chain). Moreover, our research also reveals that the predetermined total supply does not have an influ- ence on the total funds raised, still, the part of the supplythat is available for the public investors does have marginal importance. Due to the analysis, the ICO project contributors appreciate more those projects that have a greater token supply, available for the public in crowdfunding, although project funds raised do not rely on any pledged growth in the supply during a particular period. Furthermore, a bonus scheme is a part of mar- ketingintheICOcampaign,usedasawaytoattract contributors. The effect of different bonus schemes should be examined separately, since within the research, in the pool together with other elements, bonus schemes did not have a statistically signifi- cant impact on the total funds raised. Pre-sale also did not have significant (only modest) affection in this analysis, however, other researchers, for instance Adhami et al. (2018), have identified pre- sales as highly and positively important in the blockchain-based early funding. Pre-sales are described as a valuable strategy to raise funds in an ICO by checking the market's readiness. Neverthe- less, as duration is also not one of the main char- acteristics that prescribes token success, as token valuereliesondemandduringasoleperiodoftime, it can be assumed that marketing strategies have a major part in the ICO project performance and should therefore be examined individually. Addi- tionally, the first model of this research proves that Table 6. Weighted Least Squares Regression: Coefficients. Prepared by the authors (R Software Output). Coefficients Unstandardized Coefficients Sig. Correlations B Std. Error Zero-order Partial Part (Constant) 22190184.5087 9623698.1168 0.0222 WHITE_AV 7391564.2501 5569121.0518 0.1860 0.0470 0.0966 0.0809 OPS_COD_AV 3492349.7696 3174827.9942 0.0000 0.2696 0.2762 0.2395 T_SUPPLY 0.0001 0.0000 0.0516 0.1324 0.1418 0.1194 PUB_SUPPLY_PERC 9620550.8425 5916369.6641 0.1056 0.1784 0.1181 0.0991 OWN_BLOCK 1461779.3060 3779001.3703 0.6993 0.0159 0.0283 0.0236 ICO_DUR 174082.5643 44293.6652 0.2727 0.0607 0.0802 0.0670 TYPE_TOKEN 3559002.4226 5967758.5339 0.5516 0.0932 0.0436 0.0363 BON_SCH 2663318.9048 2921045.7834 0.3631 0.0762 0.0665 0.0556 PRE_SALE 875031.8973 2732378.6350 0.7491 0.1103 0.0234 0.0195 BOTH_CUR_ACC 6926504.6785 3885129.5238 0.0762 0.1696 0.1293 0.1086 HARD_CAP 0.3630 0.0622 0.0000 0.4185 0.3926 0.3557 SOFT_CAP 3146585.1148 2799608.3856 0.2625 0.1159 0.0819 0.0685 MIN_CONTR 139.2003 4484.1943 0.9753 0.0791 0.0023 0.0019 ECONOMIC AND BUSINESS REVIEW 2021;23:55e67 65 thehardcaphelps investors tomeasureandforesee the ICO success. Accordingly, contributors tend to invest more in those projects that have a pre- determined maximum goal of an investment. On the contrary, based on the research results, the soft cap does not influence investors' decisions on whether to invest or not. It turns out that the type of token (as the major part of tokens is utility), both (cryptocurrency and fiat) currency acceptance, and minimum contribu- tion have no effect on the total funds raised. How- ever, many other aspects must be taken into consideration as well, when analyzing ICOs, for example, the idea of the project, market conditions, timing, team qualification, quality of disclosure channels, and more. One of the main aspects in cases when ICOs fail is that, while developing blockchain-based projects, founders sometimeslack understanding of the economic part and dimension of creating long-lasting projects. In addition, ICOs face many risks, among them most often hacker attacks due to a security flaw, but also for being spuriously recognized as a fraud by the online community, and more. Finally, ICOs have prevailed very quickly by bringing a new way of financing to early stage companies. 2017 was the most prosperous year for the ICO market. However, in the mid-2019, ICO volumesstartedtodecrease.Thisdeclinemostlikely occurred as a consequence of the regulations that policy-makers started to undertake and the uncer- tainty of future restrictions. Nonetheless, the already initiated projects have demonstrated that theyhaveinfactcreatedstronglastingbusinesses.It is therefore logical to conclude that ICOs might actually help the cryptocurrency market improve further, as the prevalence of ICOs has obviously brought about many benefits for business. This phenomenon has a potential to change the way of funding for companies, by reducing intermediation, providing secondary market liquidity, lowering costsandgivingmorecontroltoinitiators.Whatever the case, it will take a lot of time to adopt new technologies in order to replace or improve the existing conventional infrastructures. 5 Limitations of data and research model The analysis model has a couple of limitations, mostlyrelatedtodatacollection.Sincethereisalack of official websites, where aggregate ICO informa- tion is stored, the data were collected from 4 different sources. The main sources are Token data (2019) and ICO Drops (2020), which are the most valued by the ICO contributors and founders. However, not being possible to disclose all ICO projects through these two sources might affect the random selection of the dataset. References Ackoff, R. L. (1953). The design of social research. Chicago: Uni- versity of Chicago Press. Adhami, S., Giudici, G., & Martinazzi, S. (2018). Why do busi- nesses go crypto? An empirical analysis of initial coin offer- ings. Journal of the Economics of Business, 100(Nov./Dec),64e75. Amsden, R., & Schweizer, D. (2018). Are blockchain crowdsales the New'Gold rush'? Success determinants of initial coin offerings. Success determinants of initial coin offerings (April 16, 2018). https://doi.org/10.2139/ssrn.3163849. Available at: https://ssrn. com/abstract¼3163849. An, J., Hou, W., & Liu, X. (2020). Historical Determinants of Modern Finance: Evidence from Initial Coin Offerings (August 13, 2020). https://doi.org/10.2139/ssrn.3672882. Avail- able at SSRN: https://ssrn.com/abstract¼3672882. Barsan, I. M. (2017). Legal challenges of initial coin offerings (ICO).RevueTrimestrielle deDroit Familial,(3),54e65.Available at: https://ssrn.com/abstract¼3064397. Catalini,C.,&Gans,J.S.(2018).Initialcoinofferingsandthevalueof crypto tokens. National bureau of economic research working paper No. 24418, (March 2018) https://doi.org/10.3386/w24418. Chanson, M., Gjoen, J., Risius, M., & Wortmann, F. (2018). Initial coin offerings (ICOs): The role of social media for organiza- tional legitimacy and underpricing. In International conference on information systems 2018 (ICIS 2018). At: San francisco. Chanson, M., Risius, M., & Wortmann, F. (2018b). Initial coin offerings (ICOs): An introduction to the novel funding mechanism based on blockchain technology. In 24th americas conference on information systems 2018: Digital disruption, AMCIS 2018, new orleans, LA, 16-18 august 2018. Atlanta, GA USA: Association for Information Systems. Chen, K. (2019). Information asymmetry in initial coin offerings (ICOs): Investigating the effects of multiple channel signals. Electronic Commerce Research and Applications, 36, 100858. https://doi.org/10.1016/j.elerap.2019.100858. Chod, J., & Lyandres, E. (2018). A theory of ICOs: Diversification, agency, and information asymmetry (June 1, 2020). Management science forthcoming. https://doi.org/10.2139/ssrn.3159528. Available at: https://ssrn.com/abstract¼3159528. Coinbase.(2019) [online]. Available at: https://www.coinbase.com/ price/bitcoin. Croarkin, C.,Tobias,P.,Filliben,J.J.,Hembree,B.,&Guthrie,W. (2006).NIST/SEMATECHe-handbook ofstatisticalmethods.NIST/ SEMATECH,July.Availableat:https://www.itl.nist.gov/div898/ handbook/. Domingo, R. S., Pi~ neiro-Chousa, J., & Lopez-Cabarcos, M. A. (2020). What factors drive returns on initial coin offerings? Technological Forecasting and Social Change, 153. Available at: https://www.sciencedirect.com/science/article/pii/ S0040162519304275. Felix, T. H., & von Eije, H. (2019). Underpricing in the crypto- currency world: Evidence from initial coin offerings. Manage- rial Finance, 45(4), 563e578. Fisch, C. (2019). Initial coin offerings (ICOs) to finance new ven- tures. Journal of Business Venturing, 34(1), 1e22. https://doi.org/ 10.1016/j.jbusvent.2018.09.007. Fisch, C., Masiak, C., Vismara, S., & Block, J. (2019). Motives and profiles of ICO investors. Journal of Business Research, 13. https://doi.org/10.1016/j.jbusres.2019.07.036. Fisch,C.,&Momtaz,P.P.(2020).Institutional investorsandpost- ICOperformance:Anempiricalanalysisof investorreturnsin initial coin offerings (ICOs). Journal of Corporate Finance, 64. Available at: https://www.sciencedirect.com/science/article/ pii/S0929119920301231#bb0380. Garson, G. D. (2013). Weighted least squares regression. Asheboro, NC: Statistical Associates Publishers. 66 ECONOMIC AND BUSINESS REVIEW 2021;23:55e67 GitHub.(2020) [online]. Available at: https://github.com/. Haddad, C., & Hornuf, L. (2019). The emergence of the global fintech market: economic and technological determinants. Small Business Economics, 53,81 e105. https://doi.org/10.1007/ s11187-018-9991-x. Howell, S. T., Niessner, M., & Yermack, D. (2020). Initial Coin Offerings: Financing Growth with Cryptocurrency Token Sales. The Review of Financial Studies, 33(9), 3925e3974. https:// doi.org/10.1093/rfs/hhz131. Huang, W., Meoli, M., & Vismara, S. (2020). The geography of initial coin offerings. Small Business Economics, 55,7 7 e102. https://doi.org/10.1007/s11187-019-00135-y. ICO rating.(2020) [online]. Available at: https://icorating.com/. ICO Drops.(2020) [online]. Available at: https://icodrops.com/. Masiak,C.,Block,J.H.,Masiak,T.,Neuenkirch,M.,&Pielen,K.N. (2019). Initial coin offerings (ICOs): Market cycles and rela- tionshipwithbitcoinandether. Small Business Economics,1e18. https://doi.org/10.1007/s11187-019-00176-3. Pallant, J., & Manual, S. S. (2010). A step by step guide to data analysis using SPSS. Berkshire UK: McGraw-Hill Education. Taherdoost, H. (2016). Sampling methods in research methodology; how to choose a sampling technique for research. How to Choose a Sampling Technique for Research. https://doi.org/10.2139/ ssrn.3205035 (April 10, 2016). Available at: https://ssrn.com/ abstract¼3205035. Token data.(2019) [online]. Available at: https://www.tokendata.io/. Tomat,L.,&Trkman,P.(2019).Digitaltransformationdthehype and conceptual changes. Economic and Business Review for Central and South-Eastern Europe, 21(3), 351e495. Whitcomb, K. (2012). Study guide to accompany statistical techniques business and economics(15thed.).NewYork:MCGraw-Hill/Irwin. Zaid, M. A. (2015). Correlation and regression analysis textbook (p. 39). Ankara: SESRIC. Zaninotto,F.(2016). The blockchain explained to web developers, Part 1: The theory [online]. Marmelab. Available at: https:// marmelab.com/blog/2016/04/28/blockchain-for-web- developers-the-theory.html. Zetzsche,D.A.,Buckley,R.P.,Arner,D.W.,&Fohr,L.(2018).The ICO gold rush: it's a scam, it's a bubble, it's a super challenge for regulators (july 24, 2018). University of Luxembourg law workingpaperNo.11/2017,UNSWlawresearchpaperNo.17- 83,UniversityofHongKongFacultyoflawresearchpaperNo. 2017/035, European Banking Institute working paper Series 18/2018. Harvard International Law Journal, 63(2). https:// doi.org/10.2139/ssrn.3072298, 2019. Available at: https://ssrn. com/abstract¼3072298. ECONOMIC AND BUSINESS REVIEW 2021;23:55e67 67