 This is joint work with Wei Zhang, who hopefully will be joining us around 2.30 your time. So I want to just first start with a bit of motivation for this paper. So there have been two important trends with the booming digital economy. The first is growing popularity of cryptocurrencies and tokens. As of May 2020, there were at least 4100 tokens, not including many that have failed. And as of 2021, at least 16% of US adults have had some experience using investing and or trading cryptocurrencies. And the second, which may seem seemingly unrelated to the first, is that there's a growing tension between digital platforms and their users. When we're thinking about these digital platforms, we're thinking about very large digital platforms like Google, Facebook and Amazon. So there have been recent concerns about the exploitation of users' data privacy, antitrust concerns about competition and price discrimination in big tech. And this has led to a number of regulatory responses, especially in the EU. And I'm particularly happy to be presenting to an EU audience because a lot of the policy work we see in this area is coming from reports written by EU policymakers. So we have data privacy regulations in the EU like GDPR, we have CCPA in the US as well as the Colorado and Virginia privacy acts. South America is like GDPR in many ways, so the privacy acts they have and there's also Japan. So what we want to argue in this paper is that tokenization may represent a market-based effort to resolve the conflict between platforms and users. So crypto technology makes it possible to delegate control to a set of what are called pre-coded algorithms via, for instance, smart contracts. And a lot of this kind of underpins the excitement about Bitcoin. If you read the original white paper by Satoshi, a lot of the early enthusiasm for Bitcoin was about this notion of decentralization, moving outside the conventional monetary system, retaking control, in this case, of monetary policy. Since then, it's expanded to a variety of other applications. And if you read some papers such as Harvey et al 2020, there's a lot of excitement for decentralization's potential to democratize finance. And this is often what's referred to as DeFi or DeFi for decentralized finance. So how do we think about tokenization of digital platforms? Well, the typical way this occurs as a developer will develop a platform and then get paid by issuing points and tokens rather than traditional equity assurance. This is often what's called an ICO, although they're not as popular as they were in the past, we're just going to take this as the mechanism to illustrate what we hope is a deeper point. So tokenization is going to facilitate decentralization through what are called decentralized autonomous organizations. And I'll give you a nice quote in a second to kind of say, illustrate the enthusiasm surrounding these sorts of entities. So how does a DAO work? Well, a DEO is arguably fully decentralized and governance is done among users through sophisticated voting mechanisms. For instance, you may vote on platform improvement protocols via staking, or they may issue governance tokens, such as what you would see with Terra Luna, where Luna would be the governance token, giving voting rights for altering the Terra platform. And some examples of these platforms are listed below. Prominent ones are MakerDAO and Decred Aragon Kuiper. The most popular type of tokens used on these platforms were to call utility tokens, although there are various other extensions of it. There are coins and security tokens. But what are utility tokens for the sake of the presentation? Well, utility token is going to give a user some sort of convenience yield, like a transaction benefit. So some examples of what these convenience yield could be, well, it may be about being able to write smart contracts. I want to sell my house. I put my house as a non-fungible token, so an NFT onto the blockchain, and I'm going to sell it via that. And we can put all sorts of contingencies, such as requiring title insurance, requiring that the bank provides a loan. All sorts of contingencies can be done directly through this platform's coding mechanism. So smart contracts are a key part of the excitement, but there's also storage, such as file storage with file point in storage, gaming. And there's a lot of excitement nowadays about centralized gaming. And there's also gambling as an early application via a dragging phone. So there are a variety of different types of utilities that people can get from these platforms. We're going to stay in the abstract bubble, but I want you to keep these kind of in mind. So as I said, I would be providing a quote. This comes from the CEO of Shapeshift about a year and a half ago. So I'll just paraphrase, but essentially it's unorthodox, but its decentralization is the only way to maintain fidelity to the most important principles of crypto, specifically self sovereignty over money. So there is a new kind of firm that the centralized autonomous organization. So paraphrasing the argument that the CEO was making on Twitter is that maybe the way of doing a firm where you have shareholders and management may have been appropriate during the industrial age. But now that we have online platforms with online communities that bring together themselves and provide some sort of benefit for each other. Maybe we need a more democratic system and that's kind of what the DAO hopes to accomplish. So I'll give two examples of platforms again to try to demystify these a bit. So Filecoin goes back to 2017. It is probably the most widely cited platform in terms of academic literature is to go to example that people use for what tokenization can do. So what's Filecoin? It's going to match buyers and sellers of storage on their hard drive. So a buyer wants to store their data securely. They're going to break up that data file into hundreds of little bits. And then sellers are going to store that data on their excess hard drives. So you have excess hard drive space. You're not using it. You don't need to save a lot of things. You're selling that to a buyer who demands privacy and demand security. And then there are record locators who disassemble and reassemble the scared files. They're kind of the ones who know where all the bits are located across computers across the world. So miners are going to record these addresses of these scared file pieces on the blockchain. And the way that consensus is achieved on this decentralized platform is your what are called Filecoin improvement protocols or FIPS. So this part is a little bit less discussed among academics, but it's kind of cool. So essentially anytime the community wants to improve Filecoin, they're going to come together. They're going to read over the white paper for the proposal. They can talk on Discord and they can try to reach consensus by vote. And this is kind of novel, considering conventional platforms would have Google's management, for instance, deciding on what would happen with Google search algorithm. So another example is Tesos. So this one also started in 2017 with its ICO, but launched in 2018. So this one's a bit more closer to what would be considered Bitcoin or Ether. It's going to allow for peer-to-peer transactions and smart contracts. And governance is done via two stages and the first stage developers propose a improvement in the second stage it's voted upon and developers are paid in newly minted Tesos. So I really want to emphasize that this is a platform run by users. And that's going to be important for the key conflict we're going to try to explore. So what are the conceptual questions we have in mind for today's talk? Well, what are the pros and cons of tokenized platforms relative to their conventional equity-based counterparts? Is decentralization desirable for all platforms or can we provide some cross-sectional predictions? How do outside investors and consensus validators impact decentralization? So I'm going to first consider simplified setting while abstract from these important issues about outside investors and consensus. But I'll return to those toward the end of the talk. And we think that's important because casual observation says that these two groups are important on cryptocurrency platforms. A lot of discussion, if you look in the academic literature, for instance, is about consensus protocols. So we do want to say something about that too. So we're going to be the key insights. Well, we're going to highlight what we believe is a novel tension between network effects and decentralization. So network effects are, of course, important on these platforms. This goes back as early as Roche and Tyrol. But what's kind of new here is there's decentralization. And decentralization is going to have a cost. So the way this is going to work is each user is going to have a participation cost to join the platform. And a owner would love to subsidize the marginal user because they're increasing returns to this network effect, but decreasing marginal returns for the last one who enters. So what the owner would want to do is get as many people on the platform as possible initially through subsidization. And we do kind of see this in practice. Google and Facebook provide free services such as email and search and messaging to attract users. And through that they're able to have a large network effect and become very dominant platforms. So what we want to argue is a conventional platform has an owner with equity and control to subsidize participation. But the owner can't pre-commit not to exploit users when the profit is low. And that's going to be the key tension is this lack of commitment. So in contrast to that, we can have decentralization, which we're going to call trustless. Why? Because it's a term that refers to the notion that if you have an anonymous platform where consensus and consensus is achieved without needing a central authority such as intermediary, you don't need trust of the intermediary on the platform. So it's trustless if it's anonymous and decentralized. And this may be appealing when the expected platform fundamentals relatively weak. So we're going to argue a conventional platform makes sense. If ex ante, you think this is a really good idea because you're never going to run into the situation where you're going to want to exploit users. On the other side, if profits are expected to be low, then you won't be able to control yourself as an owner in the future. And so you may want to decentralize and basically this intermediate platform immediately to avoid this conflict of interest in the future. So that's going to be the baseline model. And as I said, we're going to have two extensions where we consider investors and consensus validators. And we're going to argue that democracy of decentralization is very fragile. So it's beautiful when you're able to give control to users and they can operate the platform. But there's always tensions to try to re-centralize the platform. So if you pay dividends on tokens, for instance, that's great for subsidization because now the marginal guy also will get a dividend. And he's going to be more excited to join the platform. But you have a dividend paying token, now it invites outside investors who just want to speculate. And if they get a controlling interest in the platform, then you're going to have re-centralization as if there's a new de facto owner. On the other side, you have consensus validators who are the ones completing all transactions. So why are the incentives only when profits are low? That'll be part of the outcome of the model. So I'll kind of defer that and hopefully that will come out in about, I think, 10 slides. But that's kind of an outcome rather than an ex ante assumption. So consensus validators are the ones who are adding these transactions to the blockchain. Proof of work is the most popular, but there's also proof of stake. It's kind of become a rising star because of the energy needs of the group of work. We're going to argue that these validators act as de facto owners. So because they have control of the platform by completing these transactions, they may have incentive to attack the blockchain also when the platform's profits are low. And so that'll kind of be the way I wrap up this talk. So there's a beautifully growing literature in this area with notable contributions from my discussants. Unfortunately, I don't have time to go through all of them, but it's a very exciting space with a lot of novel ideas. And hopefully, if you're in the space, you might see your name, if not, I apologize just for the sake of brevity, these are the ones being listed. So I'm just going to jump right into the model and try to walk through the baseline model slowly because once we get through the baseline model, everything else is just going to be some sort of documentation or adaptation of it. So consider three dates, zero one and two. There's a digital platform that's going to facilitate bilateral transactions among users. This is at a very abstract level. But think of it as I want to do file sharing. So one side will be the buyer, the other will be the supplier of storage. So if you're thinking about buying a car one side is going to sell the car to another and it might be a specialized type of car that requires a very rare fight platform to be able to find a buyer. But there are many examples that this sort of paradigm can fit into. So a T equals zero a developer is going to choose between either an equity based or token based funding scheme. So equity based is going to be kind of a conventional platform, allow Roshan to roll for instance. So the owner is just going to operate the platform from both dates. The owner is going to subsidize users at T equals one to maximize the network effect. And the way that the owner is going to be compensated is he's going to charge a fraction of surplus fees. So a fraction of every user's transaction will be eaten up by the owner to compensate him for running the platform. The other thing is that a T equals to the final dates in which there's a second round of transactions. The owner may choose to exploit the user data if the fee is low. So if the fee is low is going to be the outcome. So the owner can choose to exploit user data. But of course, because users anticipate that they're going to be exploited at time to this may undermine participation at time one by rational expectations. The reverse may be true so we can discuss that a bit later, what sort of micro foundation would need to different predictions. But I think it's still interesting to say that you can't commit you may have a potential to do something in the future. So that can come to which for that may come to the prediction of which platforms would be decentralized but I think deeper point that exploitation introduces a reason for tokenization will still be true. So it equals to the owner can choose to take what we'll call a subversive action, we're going to leave that fairly reduced form, but the important thing is users anticipate. So users know they're going to be exploited at time to that's going to undermine the desire to join the platform at time one by acting as an effective tax on their participation. So what is the alternative so that's the equity base scheme, the owner operates it for both periods. There's also a token base scheme. So the platform will issue tokens at equals one to pay off the developer and the developer exits so the owner is gone, there is no longer an owner. So once it's decentralized, there's no authority to exploit user data or to subsidize participation, although I won't show you today that we do show that users don't have incentive to vote against themselves. They're not going to exploit themselves because they're shooting themselves. So this is a frictionless platform so we're assuming all consensus frictionless again will return to consensus issues that later. So users buy tokens at time one, and then they transact with each other at time one and time to just as they would on the equity based platform. So those are the two types of schemes, either equity, or it's going to be tokens. So how do users work. So there's going to be a token is going to be a, a medallion of membership on this platform. So if you buy a token, it's going to give you a right to participate on the platform at both dates. So you can think at an abstract level you buy the token, and then when you trade with trading partner during the platforms operation, you share the token by a transact. So you trade tokens with each other, and that's how you use to pay for goods and services on the platform. So we're not going to get into the specific details of how you convert between numerair and token pricing, but essentially tokens are what give you a right to be on the platform. Yes. So that's all going to come up when I get to consensus validation. So the question is validators will not want to exploit users when they make a lot of money. That's exactly what we're going to show, which can be a risk if they exploit users that's correct. So again that's going to give us predictions about when validators will want to attack the platform. So the platform fundamental is strong. Validators don't have incentive to attack because they're going to make a lot of money. When the platform is sufficiently weak, you're going to have an adverse feedback, which is that the profits from the consensus validation are low. So if I mine or I stake, I'm going to get a low profit. And this makes the platform vulnerable to attack because at that point, you have less effort, and it's going to be expended by validators. So that's going to introduce sort of a fragility there. But again, that's getting a bit ahead, but it's a well taken point. So users, there's going to be a continuum of potential users each joining the platform. It's one, each user is going to incur a personal cost cap up. So if you want to join the platform in addition to potentially buying a token, you have to pay a participation cost cap up. So that means when you're on the platform, each user is going to get an endowment, e to the AI, where AI is some common fundamental and then some idiosyncratic components across users. So some users get more benefit than others from being on the platform, all else equal. At equals one and two, two users are randomly matched. So in each round one and two users are matched with other users. And if both users are on the platform, then they're able to transact with each other. So your utility, and we're borrowing from the international trade literature, is going to be some utility over your own endowments or your own consumption good, and then some utility over your trading partners consumption good. And a to C is going to measure the degree of complementarity. So if you don't transact so you don't find a transaction partner, then you get nothing. It's just normalized to zero. But if you do match, which is an endogenous outcome because the probability of matching depends on how many other users there are on the platform. You're going to follow the usual standard international econ type of way of solving them all, which is you're going to consume a fixed fraction of your own endowment, and the fixed fraction of the endowment of your trading partner, where the fractions are one minus a to C, and a to C, and this is rather standard in this literature. So if this is your utility, then what you might notice is that the only difference across users is that some have higher endowments than others. So if you have a high endowment, you're really excited to be on the platform. If you don't have a high endowment, you're less excited. So this is naturally going to lend itself to a cut off strategy. So the cut off strategy will be, if you're expected utility from transacting on the platform is higher than some cut off. So your expected utility is the amount that you get from your own endowments, and the expectation of the utility you get from consuming the endowment good of your trading partner. And of course, this is an endogenous out object, and you have to form expectations about who is on the platform. So there's going to be select. So as a benchmark, what is the first best equilibrium? Well, that's kind of straightforward. There's going to be a critical cut off a star first best. And if a is greater than a star first best, then everyone is on the platform and that's optimal from a planner's perspective. If a is less than first best, then the net benefits of being on the platform from transactions is going to be less than the net cost of running the platform in terms of participation user participation costs, and it's going to be best to trade with each other the platform. No, it's it's two sided platform so users meet with each other. So one is going to be a buyer one's going to be a seller and they're just going to transact with each other. So you can think of it as over a period of time each period can be like a year or two years whatever it may be, you're going to run into people and trade with each other. Yes. I'm sorry, I wasn't when I say no it's one sided. Everybody's everybody's can trade with anybody else. The two sided platform is when you've got two sides and side, people from site a trade with side B. I see so if that's your convention yeah then it would be one sided. That's not my convention I think that's the standard definition. Yeah, some of the insights you attribute to LaFontira. LaFontira, Rocher, and I'm all for giving them all the possible credit in the world but are much before them to all the people who develop the theory in the eighties. There's just a question of making sure that we keep the terminology right between what's one sided and two sided. Thank you. Thank you for the clarification. So yes, we just think of this as users are transact with each other there aren't dedicated sides I agree with that. Okay, so then that is one sided and I think it's important. Okay, no no no I'm glad to have that. So I will try not to ambiguous. That's fine we forgive you for the rest of the talk we won't we won't forgive you if we sit in the published version of a paper. I will keep that in mind thank you. Jack you may miss the fact that the paper's already forthcoming so I'm not sure whether that can be changed. We, we can certainly try I think forthcoming but it's well appreciated we appreciate that classification. So I want the name of the editor now. The editor will be up. I don't want to pull them out but you can certainly see who the editor is, it is Bruno, but I'll. We can defer that discussion to laughter. So the first past is that either everyone is on the platform or not from based on the user's participation costs. The best we'll go through the equity based scheme and then through the token based scheme. So the equity based scheme there's a representative owner as I said who has equity and control the platform. The owner can do two things one provide an entry fee or a subsidy see a T equals one. So we think of the subsidy is Google offering free search or free email. You can limit the amount of subsidy so that problem doesn't become trivial, you can subsidize up to a fraction alpha of the participation cost. And you can think of this as if you give everything away for free at a very high level, you're going to have opportunistic users come on just for the free subsidies, not really to use the platform. So on the other second thing the owner can do is have a transaction. So the transaction fee is going to be a fraction delta of the transaction surplus from any two users transaction. And the last thing the owner can do is a time to can take the subversive action s element of zero one zero means no subversion one means the version. And the owner if the owner subverts the T equals two is going to get gamma per user or gammas or reduced form benefits such as selling user data, providing aggressive advertising whatever maybe. And it's going to cost you so this is value just it's a zero sum transfer but it's value destroying for the user, the user gets minus gamma at time two. And for simplicity there's no transaction at time to for users the platform effectively blows up, but that's not really important that just helps to illustrate the point. So what is the owner's profit look like. Well the owner is going to pay out any subsidization or take any fixed cost to see across users. It's going to get the transaction surplus delta times delta for the fees at time one at time to either it gets the fees again, or it instead subverts and gets gamma per user. And the important thing is that the owner lacks commit the owner cannot commit to an action at time zero time one when selling to when issuing equity to outside stakeholders and setting fees for users. So the optimization for whether or not to subvert is based only on time to profits. So time to the owner gets the transaction fee if he does not subvert or gamma times gamma per user if he does. And of course exposed you're going to do this if the value from subverting exceeds the transaction fees at time to. So when you're going to do this when the transaction fees are sufficiently low. The key conflict again is this is an optimization at time to it doesn't take into account the full profits at time one, or the user participation decision because users are already on the platform to there's nothing they can do. So user I will join the platform if the transaction benefit that he or she receives less the transaction fee so one minus delta exceeds Kappa the cost of participating fixed fee which will be negative if it's a subsidy. And if there's subversion instead of getting the benefit, they're going to face a cost minus gamma s. Subversion effectively is going to ask act like an additional cost on the user. If subversion is anticipating. So the possibility of owner subversion is going to add to user participation costs and also reduce the benefit. So this is like the key tension for the owner the owner exposed can control itself. So if fees are low, he's going to exploit the users, but ex ante he would like to commit not because users anticipate this so less users run the platform initially further defeating the fees that the owner will get. So what's the equilibrium look like there's going to be two cuddles a star e and a star star e. So if a is greater than the first cut off, everything is great platform is strong the owners going to fully subsidize as much as he can. So again, the owner still going to subsidize as much as possible to get as many people on the platform as he or she can. It's going to set a fee that's somewhat similar to that. Without subversion. So again, the owner still going to subsidize as much as possible to get as many people on the platform as he or she can. It's going to set a fee that's somewhat similar to that without subversion, but now the owner is going to take a subversive action that equals to. So users are going to anticipate this and they're going to change how they cut off of which they joined from the same fundamental. So without subversion more people join with subversion, you anticipate this. So now less users joined because they face a higher effective participation cost. And what's kind of interesting is because of the network effects. User participation, owner profit and social surplus are all decreasing in the degree of exploitation. So even though ex post it's great if you can exploit ex ante itself defeating because of the network effect. As soon as some users get off the platform because of this exploitation, more users are going to leave because there's a reverse network effect going on. So network effect is great when you're building a platform. It also defeats you very quickly when you go against it. So below the second cut off the platform breaks down. So that that's kind of the equity based platform utility based platform is slightly different. The developer is going to cash out a time one by selling tokens is going to set up a bunch of pre coded algorithms that have no ability to exploit users will return to that again in the extension. So the way this is going to work is a user buys a token to join the platform at time one, transacts with another match user at time one two. And in addition to the personal cost is going to pay some token price P. There's no subsidy of user participation because again all the owners going to do is sell the tokens and walk away. So in addition to your fixed cost of Kappa, you also have to buy a token. So again in the background there's perfect consensus, there's majority rule for tokens, it's going to be a frictionless token platform to give tokens the best chance at being a superior funding scheme. So the owner's profits are slightly different now the owner just gets the token sales so that's the number of tokens sold for this object here times the price per token. And ZT is just a transformation of the cut off at which users join. So users will join if a is great if their AI is greater than or equal to some a hat T and stay off the platform other ones. So the key difference this is a marginal users participation decision is this indifference condition here for the marginal user notice the difference from this expression, instead of transaction fees in the substance. Instead, you're just paying the token price P over here. And because it enters as Kappa plus P. It's effectively raising the cost of entering. So tokens are inherently costly to start because there's no subsidization. In fact, you're charging users even more to use the platform. So P deters users. And that's further exacerbating the network. The token price is determined by the marginal user, because this is the last guy who enters determines the token price, compared to in the case of the equity platform, the transaction fees were closer to like equity it's based on the average user. So automatically we see the tokens are giving up a lot. They make act as a tax on users deterring participation, and they're somewhat inefficient at extracting revenue, they get the marginal users convenience yield, whereas transaction fees get the act. So you're giving up a lot as an owner to use tokens. So what's the platform equilibrium look like there are only two cut offs now because there's no subversion. So below a star star T platform breaks down just as before. But above there you have a unique cut off where platform operates for two periods, users trade with each other, and there's no subversive action. So now I just want to compare these two schemes. Given a level of subversion, if a is sufficiently high the equity platform dominates the token plot. As you'd expect if there's no incentive to subvert. So you're using to issue tokens because you're giving up a lot of revenue. Similarly, given a gamma sufficiently high so the incentive to exploit is sufficiently high token platform dominates equity plan. So for a given a, if gamma sufficiently high you support. And so you're going to want to issue tokens. Similarly if gamma for fixing gamma is sufficiently low you want to support. So we're kind of building out the space of which you would want to use tokens. So higher user participation owner profit and social surplus for tokens when you use tokens and higher user participation surplus and profit for the equity platform when you opt to use that. So what this suggests is we can construct a threshold. The threshold is more general than what I'm showing here. But for the sake of fixing ideas suppose there's a normal prior. You're going to choose equity if your prior belief about the platform is above some threshold and more generally it's based upon the distributions of your beliefs. But if you're sufficiently optimistic about the platform a that is sufficiently high, then you're going to want to do equity. And if you're not that confident and you're worried about subversion undermining the network effect. That's when you're going to think about tokens. So the prediction of this model is token platforms tend to have relatively weak ex ante fundamentals expose they can be great. But from an ex ante perspective, this concern really shows up for those platforms. And there's some empirical evidence that documents skewed distributions for instance for ICO proceeds suggest that for the most part they have on average low payoffs, but occasionally you get some right tail. And we think that's kind of consistent. So can we have decentralization with tokenization. Sorry, with subsidization. So they weren't going to go now just in the last five minutes to the two extensions. So suppose instead of issuing utility tokens. The optimal from a social surplus point of view the optimal there is everyone's on the platform if the fundamental sufficiently strong. So if you have that first best cut off, you want everyone on, and if it's below you don't want anyone in that case. Yes. Over two forms you have which one is the better is tokenization better or is. I mean, if you could have a law saying you're not allowed tokenization or you're not allowed. You know which one you would choose. So it goes social surplus is perfectly aligned with the user participation in the setting, ignoring inequality of transfers. So depending on which one you think it's higher participation that's the scheme of plan with one use in this sense they're aligned between two. So that's great because that gets to this last point here where I talked about how subsidization interacts with centralization. So in this case we're going to have it that the owner can issue something a bit different than utility. Instead of just getting convenience yield. Suppose you can also pay dividends. So you can pay it like a stock dividend based on the transaction fees of users. With only users, you're going to get the first best out because now you can pay out huge dividends to the marginal user the marginal users going to join you can get everyone on the platform with the proper choice of substance via these these dividends that the equity token case. So the optimal here it's kind of trivial but illustrates the point 100% transaction fees so all users get the same payoff everyone joins if it's within if it's socially optimal to join. And the owner retain zero state. Basically it's it's endogenous the optimal for the owner to sell it because they want everyone on the platform that maximizes the value of the tokens that they sell. So what this says is that seems like assigning cash flow rights to subsidize tokens improves on utility tokens which it does. You can create a harmony between decentralization substance if you could provide dividends via tokens that acts as a subsidy just as the seed did in the equity platform, and that's great that gets everyone on the platform. So it seems like this is the best of both worlds. So we want to argue though is this introduces some bad incentives. Suppose an outside investor can buy. This outside investor doesn't care about using the platform they just want. If they don't use the platform they don't pay a fixed participation. So the investor what the investor can do is if the investor buys enough tokens, they get enough voting rights to be able to control the platform and time to. So if you acquire over 50% of the tokens, you can then vote to act like an owner, in which case of course you can take that subversive action that the owner could if it were an equity platform. Is that when the fundamental is low and subversion would be optimal. You might have a in that outside investor buy enough tokens to be able to act like the owner and support the platform time to. And this is kind of a vicious cycle. Why well if users anticipate being exploited at time to that lowers the value of tokens at time one. If it lowers the value of tokens, it's easier for the owner to us are for the investor. That's kind of a self defeating kind of self filling expectation. So the way we kind of think about this has been pointed out is the issue is that this governance algorithms can't distinguish between users and investors. At least this is a current issue with these platforms. So you think of maker DAO and Kyber, you have waiting votes by stick token holdings. So they don't distinguish between who uses the platform and just someone who's holding the tokens for their returns. So, so what we suggest is maybe one should wait token sorry wait votes by platform usage. So we should try to make sure only users vote on the platform, not just people who own the tokens, which is kind of the current standard. And I show for example maker DAO and Kyber as two platforms that kind of do that right now. So my final point is I left aside the issue of consensus. So a big issue on these platforms is that you need to have anonymous record keepers, keeping track of transactions and adding them to the blockchain. There's a huge literature on this, both for proof of work and now the more contemporary proof of state. So the last point we're going to make at a very high level is giving control and cash flow rights to record keepers can reintroduce this commitment problem. And what we do is we put together a game I'll describe at a high level, if all the validators behave they get the transaction fees just as the owner would in the equity platform. But if the fundamental was weak, then you can have someone attack a platform one of the validators may go row. We show in a mixed strategy game theoretic competition between the honest minors and the road validator that the road validator is more likely to see at attacking the platform, when the fundamental is weak, because in that situation validators are devoting less computing power or devoting less of a stake to maintaining the platform. So we think that this is somewhat consistent with empirical evidence because up until this point, all the attacks on block chains have been through these platforms with weaker fundamentals and lower prices. So feather coin Bitcoin cash verge, they tend to have smaller market caps and less usage. Those tend to be the ones at least we've seen until now, being attacked as a platform. So let's summarize and apologize for being a minute over. So decentralization through tokenization comes with both costs and benefits. The new insight hopefully that came across is decentralized tokens allow platforms to pre commit, not to exploit user data at the cost of having no owner to subsidize user participation. And that's important because we need to maximize the network. And this decentralization is fragile. There are many incentives to decentralize the platform to reintroduce the commitment problem. Two examples we give is outside investors may acquire majority state. This seems particularly relevant because a lot of the push for broader adoption has been to money managers and to institutional investors in addition to retail and consensus validators may attack the blockchain that the user suspects. So thank you so much for listening to me and I look forward to my discussions comments. Thank you, Michael. You really kept the time pretty well, give me quite a few questions so how to have a better is the expert on platforms and blockchain. And she's going to discuss this paper for five minutes or so. Share my slides. Thank you, Julian. And thank you very much for inviting me to discuss this paper. Because this paper is is already forthcoming in a journal finance. I'm not going to go via typical route of the discussant where I'm going to suggest improvements on the paper, because that's a little bit beside the point. Especially that here we have. I mean we really wanted this paper to be presented to this audience platform audience of bottom researchers, because it is of great interest to us. Instead of giving comments on you know nitty gritty details about the papers and how about if you change this or that or add another assumption. I wanted to focus on on actually talking to the audience about how we can use the findings of this papers and the modeling approach to forward our research on platforms. So with my apologies to the authors it's not going to be particularly useful for improvement of this paper but then that probably would not be socially optimal at this point. So this, this paper really built on a lot of a lot of interesting aspects of platforms that we have already we have already looked at in platforms literature. So it seems that finance literature got really interested in platforms with defy and with with with ICOs and with tokens and we could see it from Michael's literature that all the papers referred in the literature, and we're coming from 2018 onwards, which is kind of a natural natural meeting point of finance and platforms. But in platforms, we actually have a quite rich literature and platform governance that goes back, you know, much, much is much older than So I want to kind of point out two aspects of this literature. One is that we have Andres results from 2006 about the importance of commitment and commitment of platform to future pricing. And this is a very, very similar dynamics. If a platform cannot commit in the future to a particular pricing and therefore has incentive to exploit the users who have already joined. Then, or using the users that already joined exploit the future users, then this is going to affect the incentive of the users to to join in the beginning of the platforms existence. The paper is a bonus in the two sided papers with Jack was saying is that we first have our video game developers, let's say and then the gamers, but if the platform cannot commit to the pricing for the second time it's not going to for the second site is not going to attract the first site. And, you know, there's two types of far of users that we typically think about in two sided context. I kind of wanted to make this remark here given that as Jack already talked about two sidedness. We can use a similar logic to thinking about the first period users and second period users because they are in a way different if you attract the users in the first period it makes it easier to attract users in the second period, and it gives the same incentive to exploit us when we attracted one site and now we want to make money on the on the second side. Okay, then we have a literature I mean it's not the main point of platform strategy research, we typically focus on platforms that are centrally managed. And what, what, what Michael of the paper is equity platform, but there is a significant research focusing on open platforms where where the users are managing the platforms making decisions. And the main examples here are Linux and Wikipedia and how those platforms work and how they create network effects and how they compete against the platforms like Linux against against Microsoft, for example. And what we see in that literature is that, for a given size on the network, open platform will open this kind of decentralized platform will offer higher customer surplus, because there is no centralized platform provider who will exploit and charge high prices for the service, but at the same time, the open platforms network is smaller. So on net there may be a negative, the negative that open platform maybe less preferred by the users, even though the platform is not exploiting them. And why is open platform open open platforms network smaller is because there is no incentive I don't know incentives to subsidize early users and grow the network, exactly because there is, there is no proprietor or the of the network who can later recruit the recruit the investment of subsidizing. So, this kind of open platforms that have been traditionally analyze in platforms literature, they always have open access, which is the same as having no cost. Right, so you join a platform and you're not paying clues wiki pdr Linux and so on, because traditionally before blockchain and tokenization, we did not have a way to monetize. In the decentralized platform, the moment you have payment system the moment you pay with a credit card you need to have someone who is centralizing the central authority in the platform. And, and that kind of opens up this question that if we move from traditional platforms to blockchain platforms and, you know, for some of you this slide maybe familiar because I took it from one of my older presentations is that with the with with blockchain and smart contracts, we, we can allow or we can have algorithmization of governance via dollars for example, and that may give us a unique opportunity new opportunity to have those open platforms and co platforms to have to be at the same time to monetize and be able to monetize. And now the question is, can they perform better against that the standard platforms that we have a kind of equity platforms or not. And this is where you know we had this question as an open question area and platforms, and your paper is already first step to show us that that may not necessarily be enough. I basically called the centrally managed platforms to lose platforms of coming, coming back from the, you know, the early descriptions, and then the decentralized platforms that may allow for decentralization monetization the same time. Nakamoto platforms. Right. So, in this paper what we are seeing that also when we have tokenization and we use those tools of decentralization with monetization where the, where the users can make a can make profit where the, the token can have a price and we can maintain this monetization. We still see a similar effect as we had seen for the traditional open platforms that could not charge for access, even though the model is much more complicated and now we can, we can see that those forces are much, much stronger because it was not, it was not just a problem of monetization. So, the decentralized platform increase user surplus, but for the same size of the network but achieve a smaller size because no incentives to, there are no incentives to subsidize earlier on. What, what Michael and today are showing is that it is actually possible to overcome the shortcoming of the traditional open platforms with equity tokens. Right, and this is where you're showing that you can have this first best with decentralization and there is subsidization, if there is transfer of transfer of payoffs between the first period and the second period via equity token, and then the centralized way. However, it works only if the equity tokens are not transferable as if I may paraphrase, and that kind of relates to a different literature that we had in platforms and this was on the platform based digital currencies. And this was referring to Facebook credit Amazon coin War of Warcraft gold kind of the traditional platforms now that the centralized platforms. But the research there was basically showing that if we have transferability of those of those coins. Then the transferability is detrimental to network effects, and this is why platforms very often when they set out those the currency or coins on the platform for use on the platform. They very often restrict functionality. They don't want it to be generally accepted means of payment because they want to maintain that the currency on the platform to increase network effects. There seems to be this flavor of this off of a similar of similar kind of dynamics that you want to keep it on the platform to keep it with the users you don't want that the revenue streams to spill outside of the platform. So, what it very interestingly shows is that while this decentralized combining decentralization with monetization that is offered by smart contracts and blockchains. And help us solve some of the issues that we have identified earlier in platforms literature, it also it also introduces new new issues new aspects and new problems. And that gives us more topics to work on as platforms researchers. So where can we, where can we take it from here. This is the last slide. Five minutes to the hour so we're going to be a little bit time for others. So I will just just basically say that we can look into the different structure of the platforms that may lead to different results. So, make your doubt and file point are different in their nature than Bitcoin or Ethereum. And you can analyze it the same way but they're different and we have similarly in in platforms of two sided one sided platform and so on. There's potentially conflict of interest between the users which you completely abstract away and it may have a big impact on how the platform functions. We talk about optimal design of talking talking functionality. And what is really important for me is the competition, either between the open platforms or between the decentralized platform and centralized platform and how they how this will, this will play out. So those are kind of open questions that I wanted to throw out out there. Thank you and thank you Julian for letting me go over time.