STO vs ICO: A Theory of Token Issues Under Moral Hazard and Demand Uncertainty

This paper considers a financing problem for an innovative firm that is considering launching a web-based platform. Our model is the first one that analyzes an entrepreneur's choice between security tokens (via a security token offering (STO)) and utility tokens (via initial coin offering (ICO)). The entrepreneur on one hand faces a large degree of demand uncertainty on his product and on the other hand has to deal with incentive problems of professional blockchain participants who contribute to the development and sales of the product. We argue that utility tokens with profit rights are a better option for the firm compared to straight utility tokens or security tokens because they help the firm better deal with both the moral hazard problems (via profit sharing incentives) and demand uncertainty (they help the firm learn the product demand). This finding is consistent with some recent evidence. The paper also generates new predictions that have not been tested so far.


Introduction
Innovative companies account for a signi…cant share of the global market for human capital but they are often constrained in their growth potential as they have di¢ culty accessing capital markets (Hall (2009)). Blockchain-based initial coin o¤erings (ICOs) and security token o¤erings (STOs) promised to provide a new source of …nancing for such …rms. The ICO phenomenon dates back to 2013. Since then, the number and funding of projects has been growing exponentially, with over $20 billion raised by December 2018 (Coinschedule, 2018). In a typical ICO, an entrepreneur raises capital by pre-selling utility tokens I am grateful to Magdalena Brodziak, Xiehua (Richard) Ji, Victor Miglo and seminar participants at de Montfort University, London South Bank University and Coventry University for helpful comments and editing assistance.
y Birmingham City University, Birmingham, UK. anton.miglo@bcu.ac.uk. which give their owners the right to use the company's product or service once it is developed. In 2017, the next step was taken. Fintech companies started to use STOs to …nance their projects. In security token o¤erings (STOs), 1 companies sell tokenized traditional …nancial instruments, like, for example, equity where tokenholders receive rights on a …rm's future pro…ts. 2 The number of STOs is quickly growing. In January 2018 5 STOs were conducted (monthly) while in November/December 2018 there were more than 20 per month and it continues to grow. 3 ICO and STO research is quickly growing. Most papers are focused on ICOs. Theoretical papers on ICOs include, amomg others, Catalini and Gans (2018), Li and Mann (2018), Govindan and Wilson (2009), Bakos and Ha laburda (2018), Cong and Wang (2018), Garratt and van Oordt (2019) and Lee and Parlour (2018). Reserach on STOs and utility tokens with pro…t-sharing rights is in its early stages and as we are writing this article it includes several emprical papers (eg. Adhami, Giudici and Martinazzi (2017) and Ante and Fiedler (2019)) but no theoretical paper to the best of our knowledge. Respectively no paper is focused on the choice between ICO, STO and etc. eventhough for many entrepreneurs this issue seems to be very important. 4 In this article we shed some light on these unexplored questions namely what are economic ideas behind issuing security tokens or utility tokens with pro…t rights and how …rms select between di¤erent types of tokens.
Our model builds on the following observations. First, ICOs and STOs are characterized by an environement with high uncertainty. A lot of campaigns fail or turn out to be low quality or even fraud in some cases. 5 Firm success in these innovative areas depends crucially on the incentives and e¤orts of not only the …rm itself but on many particpants invloved. For example, an interesting case is Filecoin, which is setting up a network to allow peer-to-peer storage space sharing. Their success depends on action and strategies of so-called miners who are expected to be active participants of their platform. Token design issued by the platform may a¤ect the incentives of parties involved. For example, in the case of Filecoin, miners purchase tokens during the pre-sale. 6 Second, tokens serve as a learning tool for entrepreneurs regarding market demand. By observing the demand for tokens during the initial sale of tokens or by observing the token market price, the entrepreneur can learn "crowd wisdom" regarding the platform and its products. Finally, tokens have secondary markets (see, for example, the interview with BlockState CEO Paul Claudius 7 ) unlike, for example, venture capital investments. This feature of tokens makes it also di¤erent from crowdfunding which typically does not have a secondary market for investments made by funders. 8 In our model an entrepreneur with an innovative idea considers launching a web-based platform for an in…nite number of periods. The demand for the product is highly uncertain so the entrepreneur can make production decisions without learning demand or it can issue tokens prior to making production decisions. The success of the platform also crucially depends on the e¤ort provided by the entrepreneurs and blochchain participants (miners) during the development stage. In order to …nance the development of the platform, the entrepreneur can issue tokens. Utility tokens give the right to purchase a product or service on the platform while security tokens give a right on …rm pro…t. The "wisdom of the crowd" aspect of a platform kicks in when the …rm is facing demand uncertainty. Without utility tokens, production (and repsectively pricing) decisions of the …rm are not optimal. Usage of utility tokens helps the …rm to learn the demand and improve its decision-making including production (pricing) decisions. However the shortcoming of utility tokens is that they do not provide much incentive for miners to develop the product. On the other hand security tokens to not provide a ‡exible tool for learning market demand. We then analyze the trade-o¤ between security tokens and utility tokens for the entrepreneur. We show, for example, that the utility tokens will be preferred if the degree of unceratinty regarding market demand is higher (it increases the learning value of utility tokens).
Next we include utility tokens with proft rights into the basic model. We demonstrate that this type of token dominates regular utility tokens (i.e. without pro…t rights) or security tokens. Learning opportunities in terms of demand for this kind of token still exists which makes it similar to utility tokens without pro…t rights. Also in contrast to utility tokens without pro…t rights, they do a better job of incentivizing miners during the development stage.
Our model provides several predictions most of which have not been tested sofar. Interestingly though, one of our main predictions namely that utility tokens with pro…t rights can dominate utility tokens without this right is consistent with recent empirical evidence. In a subsample of 253 campaigns, Adhami et al (2017) document higher returns when tokens allow contributors to access a speci…c service including pro…t rights. Our results also provide several implications for policymakers and practitioners. First, it explains factors that should be taken into account by managers designing optimal token design for their …rms. Secondly it can help di¤erent platforms hosting ICOs and STOs compare the suitability of the di¤erent types of tokens with a variety of business factors, which ultimately can help platforms deal with di¤erent issuers and minimize risks (maximize quality).
KuCoin CEO Michael Gan explains that the advantage of why his business is doing relatively well compared to its competitors and why its tokens have an active and growing market is that their tokens have both utility value and pro…t sharing rights. It helps on one hand to provide all services to customers but also ensures …nancial incentives even when markets seems to be bearish. "...As the native token of KuCoin, KCS holders now can enjoy trading fee discount and daily KCS bonus on our platform. ....Also, KCS has gradually been accepted by increasing number of industry partners. You can now use KCS to get a loan on ETHLend, transfer KCS to your friends on Adamant Messenger, pay private expenses with KCS through Aave pay. More use cases will be unveiled this quarter." In many articles KCS is named one of the best dividend paying tokens so it has aspects of both utility tokens and security tokens. 9 There are many other examples of cryptobusinesses that use similar ideas including Binance, Medpath, XWIN, Elephant, Props, Treecoin, XOV etc. Garratt et al (2019) study the e¤ect of entrepreneurial moral hazard on ICO outcomes and …nd conditions for when an ICO is a better choice than traditional debt or venture capital. Compared to Garratt et al (2019), we also study the incentives of other blockchain participants related to moral hazard problems. In Catalini et al (2018) an ICO allows an entrepreneur to generate buyer competition for the token, which, in turn, reveals consumer value without the entrepreneurs having to know, ex ante, consumer willingness to pay. In our paper, on the other hand, tokens can help entrepreneur learn market demand in each period by observing the token price on the secondary market. Compared to the papers mentioned above, we also study STO and utility tokens with pro…t rights.
The rest of the paper is organized as follows. Section 2 describes the basic model and some preliminary results. Section 3 provides an anlysis for the model with moral hazard and demand uncertainty. Section 4 analyzes the role of utility tokens with pro…t sharing rights. Section 5 discusses the consistency of the model's predictions with observed empirical evidence. Section 6 discusses the model's robustness and its potential extensions and Section 7 is a conclusion to the study.

The Model Description and Some Preliminaries
An innovative …rm has monopoly power over its idea of creating a website platform selling a product/service for an in…nite number of periods. The platform's quality depends on the e¤ort provided by the entrepreneur (e 1 ) and the blockchain participants (we call them miners for shortness) (e 2 ). The cost of effort is e 2 j 2 , j = 1; 2. 10 During the operational stages of the platform, the demand for product in each period is expected to be driven by the following demand function: q = v p, where p is the price, q is the quantity demanded. In each period v can be either high (v h ) or low (v l ). The probability of high demand is . Let n be the …rm's operational pro…t in period n and is the discount factor. Respectively E(e 1 ; e 2 ) P n n (1+ ) n is the present value of the …rm's earnings where E(e 1 ; e 2 ) is a factor that re ‡ects the platform's quality. We assume E(e 1 ; e 2 ) = e 1 + e 2 (1) The calculations of n as well as the way the …rm's earnings will be distributed depend on the …rm's …nancing strategy. To …nance the development of the product the …rm can sell tokens. Tokens may vary in design. They can be utility tokens which give the tokenholder the right to purchase the product on the platform. They can also be security tokens which give tokenholders pro…t sharing rights. The …rm is owned by an entrepreneur. Utility tokens. Initially, i.e. before the platform is launched, the …rm sells tokens to miners for the price p 0 . 11 The total number of tokens is normalized to unity without loss of generality. As we will see, the relative fractions of tokens owned by the entrepreneur, miners and public are important. After the …rst issue of tokens is sold, the entrepreneur and the miners provide their e¤orts. Miners then trade tokens on the secondary market. After that the platform is launched. In each period, the entrepreneur sells tokens received for selling the product in the previous period. After that he determines the level of production. At the end of each period the produced items are exchanged for tokens.
Security tokens. The …rm selects the fraction of equity that will belong to security token holders and sells them during the STO to miners. After that the entrepreneur and miners select their production e¤orts. The platform is launched for an in…nite number of periods. In each period, the …rm produces its products/services and sells them to the public. The …rm's earnings are distributed pro-rata according to the number of tokens owned by each tokenholder.
First consider the symmetric information scenarios without moral hazard problems for the di¤erent types of tokens. We assume that v is given and the quality of platform E is also given and equals 1 for simplicity and it does not depend on any e¤orts made by the entrepreneur or miners.

Utility tokens
The timing of events is present in Figure 1. The platform is launched The …rm determines q 1 Products are sold to the public for tokens The entrepreneur sells tokens on the secondary market for the price pn The …rm determines qn Tokens are exchanged for products for the price Tn per item (in tokens) Figure 1. The sequence of events for utility tokens.
We begin the solution by working backwards. Consider the operational stage. In period n, the entrepreneur sells tokens for the price p n . After tokens are sold, the …rm determines q n . Tokenholders then use their tokens to buy products. Equilibrium is determined by the following conditions: 1) after selling tokens the …rm maximizes its pro…t in tokens (since tokens are the only medium of exchange on the platform), which equals q n T n (production-incentive constraint) 2) demand equilibrium: where P n is the cost of the product for the public: Taking into account (2) and (3), the entrepreneur's objective function can be written as (v qn)qn pn . The optimal q n equals (note that by the time the production decision should be made, tokens are sold and p n is determined) and the entrepreneur's pro…t (in tokens) equals: v 2 4p n From (2) and (4) we have: This implies a non-arbitrage condition for consumers (i.e the cost of tokens for consumers (p n ) equals the cost of products o¤ered by the entrepreneur taking into account the demand function): v 2 = T n p n Token market equilibrium (supply equals demand) is described by the following condition: This implies: The present value of the …rm's pro…ts equals = P n n (1+ ) n and the present value of the entrepreneur's earnings equals The second term is substracted because the entrepreneur does not sell tokens during period 1 (it is done by the miners; note that without moral hazard the results would not change if the entrepreneur sold it directly to the public).
At the beginning of period 1, miners sell their tokens on the secondary market for the value: When selling tokens, the entrepreneur's total pro…t is: under the condition that miners'pro…ts covers their investment costs We assume that there is a large number of miners so they agree to invest an amount equal to the present value of their future pro…ts. The entrepreneur's total pro…t equals

Security tokens
The timing of events is presented in Figure 2. The platform is launched The …rm determines q 1 Products are sold to the public The …rm determines qn Products are sold to !the! public Consider the operational stage. In period n there are q n items produced. The …rm's objective function can be written as (v q n )q n cq n . The optimal q equals q n = v 2 and the entrepreneur's pro…t equals: The present value of the entrepreneur's pro…ts equals = P The miners'pro…t equals: v 2 4 When choosing , the entrepreneur maximizes: under the condition that the miners'pro…t covers their investment cost The entrepreneur's total pro…t equals then Lemma 1. Without moral hazard and demand uncertainty, the …rm is indi¤ erent between the di¤ erent types of tokens.
Proof. Follows from the comparison of (6) and (7). This result is not surprising given that in the absence of any …nancial market imperfections every type of …nancing should have the same result (similar to Modigliani-Miller proposition (1958)).

Product Development, Market Uncertainty and Incentives
In this section we analyze the role of moral hazard and market uncertainty on the …rm's choice of tokens. The entrepreneur and miners provide e¤orts in the development stage of the platform that a¤ect its quality. The token design a¤ects the incentives of all the parties involved. Also the market demand for platform products is uncertain. The token design a¤ects the platform ability to learn information about market demand before the entrepreneur makes his production decisions. The timing of events for utility tokens is present in Figure 3. The platform is launched The …rm determines q 1 Products are sold to the public for tokens The entrepreneur sells tokens for the price pn and learns vn The …rm determines qn Tokens are exchanged for products for the price Tn per item (in tokens) Figure 3. The sequence of events with moral hazard and market uncertainty for utility tokens.
The timing of events for security tokens is present in Figure 4. The platform is launched The …rm determines q 1 Products are sold to the public The …rm determines qn Products are sold to public Figure 4. The sequence of events with moral hazard and market uncertainty for security tokens.
We will proceed in 3 steps. First we will consider the case with moral hazard without market unceratinty. Next we will consider the implications of market uncertainty and …nally we will consider them together.

Moral hazard
We start with utility tokens.

Utility tokens
We begin the solution by working backwards. Consider the operational stage. Similarly to the previous section, we get that the present value of the entrepreneur's pro…ts equals The di¤erence with the previous case is that the quality of the platform was given but here it depends on e¤orts provided by the entrepreneur and miners. At the beginning of period 1, miners sell their tokens on the secondary market for the value: v 2 4 At n = 0, the entrepreneur chooses e 1 to maximize Taking into account (1), this equals (e 1 + e 2 )v 2 4 (e 1 + e 2 )v 2 4(1 + ) Optimal e 1 equals: Miners chose e 2 to maximize their discounted earnings from selling tokens at t = 1 minus the cost of e¤ort: Taking into account (1), this equals Optimal e 2 equals: (9) and (11) imply It implies that (8)  When selling tokens, the entrepreneur's total pro…t is: under the condition that miners'net pro…t covers the investment cost The entrepreneur's total pro…t equals then = (2 + )v 4 32 (1 + ) 2 + (1 + 2 )v 4 32 2 (1 + ) 2 = (1 + 4 + 2 )v 4 32 2 (1 + ) 2 (14)

Security tokens
Similar to Section 2, the present value of the entrepreneur's pro…ts from operations equals E(e 1 ; e 2 )(1 )v 2 4 The entrepreneur chooses e 1 to maximize Taking into account (1), this equals Optimal e 1 equals: The miners chose e 2 to maximize: The optimal e 2 equals: (16) and (18)  And (17) equals (e 1 + e 2 )v 2 4 e 2 2 2 = (2 ) v 4 32 2 When choosing , the entrepreneur maximizes: Lemma 2. Under moral hazard, the entrepreneur's pro…t when the …rm issues security tokens is higher than with utility tokens.
Proof. Follows from the comparison of (14) and (19). Indeed the di¤erence between them can be written as 3v 4 64 2 (1+ ) 2 > 0. The idea behind Lemma 2 is that miners are better incentivized with security tokens. Miners receive part of the …rm's pro…t for a long period of time and if this part is su¢ ciently high they provide a higher level of e¤ort than with utility tokens. The entrepreneur's e¤ort is reduced but not by much since the entrepreneur keeps a large fraction of equity in any case for a long period of time in the company. Most importantly when maximizing his objective function initially, the entrepreneur has ‡exibility in terms of selecting the optimal fraction of equity for selling to miners by taking into account the cost of the miners' e¤orts and his own cost. As one can see from (16) and (18), the entrepreneur and the miner's pro…ts depend on the fraction of pro…ts o¤ered to security token holders. With a proper selection of the fraction of pro…t o¤ered to security token hoders, the …rm can provide a good combination of incentives in the case of security token issues. Under utiltiy tokens the entrepreneur does not have much ‡exibility in managing the levels of e¤orts since utility tokens do not give their holders a long-term fraction of the …rm's equity so the level of incentives that can be induced with utility tokens is smaller than it is with security tokens.

Demand uncertainty
Here we assume that in each period the demand for the product o¤ered by the platform is either v h with probability or v l . Issuing utility tokens helps the entrepreneur learn the demand and helps with production decisions.

Utility tokens
Consider the operational stage. In the beginning of each stage product demand is unknown to the entrepreneur (v n equals v h with probability and v l with probability 1 ). In period n, the entrepreneur sells tokens for the price p n . After tokens are sold, the …rm determines q n . Tokenholders then use their tokens to buy products for the price T n (in tokens). Equilibrium is determined by the following conditions: 1) after selling tokens the …rm maximizes its pro…t in tokens, which equals q n T n ; 2) demand: q n = v j P n ; j = l; h where P n is the cost of the product to the public P n = T n p n Taking into account (20) and (21), the entrepreneur's objective function can be written as (vj qn)qn pn . The optimal q n equals q n = v j 2 and the entrepreneur's pro…t (in tokens) equals: v 2 j 4p n Also note that we have: P n = v j 2 This implies a non-arbitrage condition for consumers: v j 2 = T n p n Token market equilibrium: q n T n = 1 This implies: The present value of the …rm's pro…ts equals The latter term is subtracted because the entrepreneur does not sell tokens during period 1.
In period 1, miners sell their tokens on the secondary market for the value: When selling tokens, the entrepreneur's total pro…t is:

Security tokens
Consider the operational stage. In period n the …rm produces q n items. The price of the item depends on the market demand. If it is v h the price equals p n = v h q n and if it is v l the price equals p n = v l q n . When making its production decision, the …rm maximizes its expected pro…t. The …rm's objective function can be written as ( v h + (1 )v l q n )q n . Optimal q equals q n = v h + (1 )v l 2 and the entrepreneur's pro…t equals: The present value of the entrepreneur's pro…ts equals The miners'pro…t equals: When choosing , the entrepreneur maximizes: (23) Lemma 3. Under demand uncertainty, the entrepreneur's pro…t when the …rm issues utility tokens is higher than it is when the …rm issues security tokens.
Proof. Follows from the comparison of (22) and (23). Indeed the di¤erence between them can be written as The idea behind Lemma 3 is that the …rm learns the market demand when selling utility tokens, which were collected in the previous period, on the secondary market at the beginning of each period. This is consitent with the idea of learning via "crowd wisdom".

Moral Hazard and Demand Uncertainty
In this section we analyze token design when market uncertainty and moral hazard are both present.
Proposition 1. Under moral hazard and demand uncertainty, the …rm's pro…t if it issues utility tokens equals If the …rm issues security tokens, its pro…t equals Proof. See Appendix 1. Naturally, pro…t in either case increases with the expected demand (v h , v l and ) and and decreases with the discount factor .
Proposition 2. The likelihood of selecting utility tokens increases (respectively the likelihood of selecting security tokens decreases) when increases from 0 to v l v l +v h and decreases when increases from v l v l +v h to 1; when increases; for a given value of v l is positively correlated with the di¤ erence between v h and v l .
Proof. We need to compare (24) and (25). The former is greater when The derivative of right-hand side (RHS; respectivelly LHS will be used for lefthand side) of (26) in equals: which proves the …rst part of the proposition. Indeed, the sign of (27) is determined by the sign of v l v l +v h . It is positive when < v l v l +v h and is negative otherwise. The derivative of LHS of (26) in equals 6(1+ ) 2 (1+4 + 2 ) 2 , which proves the second part. Finally note that the di¤erence between RHS and LHS of (26) can be written as For a given value of v l , the derivative of this with respect to v h v l is positive, which proves the last part.
Proposition 2 has an interesting interpretation. Points 1 and 3 are related to the degree of market uncertainty and the amount of information that the entrepreneur can recieve when learning the market demand with tokens. Indeed if = 0 or 1 the amount of information is zero since the demand is deterministic. The same holds if is either very small or very large because there is a large chance that the demand is either very high or low. However when is in the middle the degree of uncertainty is highest since the demand can go either way. The last point is also related to information since a larger the di¤erence between v h and v l implies a higher risk from misvaluing the demand. Point 2 implies that securtiy tokens are more sensitive to the value of the discount factor. If it is high then the e¤ect of security tokens as an incentive device is diminished.
The next section show that if the …rm is able to issue tokens with pro…t rights it can improve its overall outcome.

Utility Tokens With Pro…t Rights
Suppose that the …rm can issue utility tokens with pro…t rights. In this case the …rm selects the fraction of equity that belong to tokenholders and sells tokens to miners. After that, the entrepreneur and miners provide their e¤orts. Miners sell their tokens on the secondary market. The platform is launched for in…nite number of periods. At the beginning of each period n, the …rm sells tokens to the public. Then the …rm determines the level of production q n and pays dividend d n to tokenholders. Produced items are then exchanged for tokens.
The timing of events for utility tokens is present in Figure 5. The platform is launched The …rm determines q 1 Products are sold to the public for tokens The entrepreneur sells tokens for the price pn and learns vn dividends dn are paid The …rm determines qn Tokens are exchanged for products for the price Tn per item (in tokens) Figure 5. The sequence of events with moral hazard and market uncertainty for utility tokens with pro…t rights.
Lemma 4. Without moral hazard and when the demand is known, the entreprneur's pro…t equals v 2 4 .

Proof. See Appendix 2.
This result is not surprising since without market imperfections, the …rm's pro…t is the same as it is with utility tokens or security tokens (see Lemma 1).
Proposition 3. Under moral hazard and demand uncertainty, the entrepreneur's pro…t equals: Proof. See Appendix 3. The amount of earnings is positively correlated with v h , v l and and negatively correlated with .
Proposition 4. The entrepreneur's earnings in case the …rm issues utility tokens with pro…t rights are higher than they are under security tokens or utility tokens without pro…t rights.
Proof. We need to compare (24), (25) and (28). First note that (28) is greater than (25). Indeed the di¤erence between them can be written as Now compare (28) and (24). The di¤erence between them can be written as which is positive.

Implications
Our paper has several implications for an entrepreneurial …rm's choice of token design.
Proposition 2 implies that ICO is preferred to STO if the market uncertainty increases or the discount rate decreases. Although this prediction has not been tested directly it is consistent with the spirit of Amsden and Schweizer (2018). They show in their sample of 1,009 projects between 2015 and 2017 that ICO success depends negatively on venture uncertainty and positively on venture quality.
Proposition 3 implies that utility tokens with pro…t rights dominate security tokens and utility tokens without pro…t rights. The …rst part is consistent with Adhami et al (2017). In a subsample of 253 campaigns, Adhami et al (2017) document higher returns when tokens allow contributors to access a speci…c service including pro…t rights. The second part has not been tested sofar.
Our model is also consistent with the existence of a positive correlation between the platform's quality and the amount raised during an ICO (see, for example, Ante, Sandner and Fiedler (2018)). Indeed, it follows from (12) and (13) that the amount raised during the ICO equals: which means that p 0 is positively correlated with E.

The Model Extensions And Robustness
Other types of moral hazard. In our model, the moral hazard takes place because, for example, the particpants' equity stake in the …rm is reduced while his individual e¤ort is costly and this cost is not shared. This approach is very common in …nancing literature (starting with Jensen and Meckling (1976)) and typically creates an agency cost of equity …nancing. There are many di¤erent ways to analyze moral hazard issues, for example, to explicitely assume that the entreprneurs can "steal" money from the …rm. In this case the entrepreneur trades-o¤ private bene…ts from "ine¢ cient" investments and the cost incurred in the case of the …rm's bankruptcy. The entrepreneur's objective function can be made more complicated by including, for example, some bonuses from "good" investments. One can also consider an alternative function for a joint result of e¤orts provided by entrepreneurs and miners. At this point, however, we do not see which parts of our ideas can be a¤ected qualitatively without singi…cantly complicating the model's solutions so we leave it for future research.
Mixed …nancing and more types of …nancing. Unlike capital structure literature, where a debt/equity mix is a very common strategy (as opposed to pure equity or pure debt …nancing), 12 simultaneously issuing di¤erent types of tokens has not shown to be common. Nevertheless, if mixed …nancing is allowed in period 1, most results will stand. In fact, qualitatively if the …rm decidies to issue two types of tokens (utility tokens and security tokens) the results are very similar to issuing utility tokens with pro…t sharing rights. Note that this strategy seems to be quite popular in practice. For example the CEO of Minthealth Samir Damiani stated the following in one of his interviews: "You will absolutely see the rise of the security token. In fact, industry analysts and leaders predict that 25% ($20 Trillion) of the existing global equity market of $80 Trillion will be security tokens in the next 3 to 5 years, driven primarily by the massive in ‡ux of institutional capital. The security token is an incredible tool for companies as they enable stakeholders to participate in the growth of a company and reap the bene…ts of its success in an SEC compliant manner.....As for the novel dual token structure, we see this as necessary for our company, and likely will become more common in the future. Several industries can ben-e…t from incentivizing consumers. A growing spectrum of industries already have loyalty programs (think Amazon, CVS, Amex etc). As more companies leverage Blockchain, it is likely the fruits of a dual token structure will become more apparent and widely leveraged." 13 These ideas are very similar to the ones suggested in this paper. In fact, Minthealth has decided to issue two types of tokens. and its motivations are quite similar to the ideas in this article. Dual token structure is de…nitely an interesting direction for future research.
Two stages. One can assume that the …rm issues tokens in two stages. For example in case of utiltiy tokens the …rm sells a fraction t of tokens to miners and then 1 t to the public. As far as we can see, the results will not change with the introduction of this assumption however if one introduces for example two development periods in the model with two di¤erent e¤orts in each period (a dynamic extension of the model) the results will change at least quantitatively. It is hard to predict the consequnes of such a chnage so it is di¢ cult to judge if it is a promising avenue for future research.
Voting rights. One can futher extend the model by allowing the …rm to develop more than one project in the initial stage with di¤erent utilities for the entrepreneur, the miners and the public and let tokenhoders particpate in the decision-making process etc.
Asymmetric information. In our paper we focus on ex-post asymmetric information, i.e an environment where platform quality depends on the e¤ort of its deloppers. One can consider a model with ex-ante asymmetric information where the entrepreneur initially has some signals about its platform and would like to signal it to the market via tokens issue. It is an intersting avenue for future analysis but it is beyond the scope of our model. 14

Conclusions
This article is the …rst one that o¤ers a model of the choice between ICO and STO for an innovative …rm looking to fund the development of its platform. Existing literature usually focuses on ICOs. Our paper is also the …rst one that has a theoretical model of STO as well as an analysis of utility tokens with pro…t sharing rights. The topic is a highly growing area among researchers and practioners. Our model is based on two important features of innovative …rms dealing with the development of FinTech related products. First, moral hazard problems related to the developmet of platforms. The reason for this is that the quality of a platform is is highly uncertain to participants and the token design can a¤ect the incentives of the parties invloved. Secondly, tokens have secondary markets unlike venture capital investments or crowdfunding. We study how the design of tokens can help the …rm learn information about the demand by observing token price on the secondary market. We …nd that utility tokens are prefered to secutiy tokens when the degree of uncertainty is high. We also …nd that security tokens may be prefered if the moral hazard problem is important. We then analyze the role of utility tokens with pro…t sharing rights and …nd that these tokens are more pro…table for the entrepreneur compared to utility tokens without pro…t rights and security tokens. Most of our model's predictions are new and have not yet been tested but they seem to be consistent to some extent with recent empirical evidence, eg. Adhami et al (2017).