 So, hi everyone and welcome to this platform, this online platform seminar. So today the speaker is Azlem Bedrider for you, talking about hybrid platform models and joint work with Simon Anderson. So the rules I guess, I mean, most of you are used to them anyway. So if you have any questions, you can ask them in the chat, Azlem will stop from time to time for clarification questions and for more substantive ones, we can keep that at the end after Eski's discussion. I think I've said what I had to say, so Azlem, the floor is yours. Thank you. Thank you very much, Alex. Thank you everyone for being here today, good morning, good afternoon. So this is a joint work with Simon who is also here in the audience, so I'm happy that you will pick up the difficult questions. So we analyze hybrid platform models before going into our modeling setup and I want to just motivate why we are looking at these models in particular. So we have seen significant increase of e-commerce as a retail revenue in the last years, in particular during the coronavirus times. So this amounted to 10% of retail revenue in the US. So Amazon appears to be the dominant player of the e-commerce markets, both in US and in Europe. So these are the market shares that Amazon has in these countries, but situation is pretty similar in most of European countries. And Amazon has this business model called hybrid platform model that is basically Amazon enables interactions between third-party product sellers and buyers, just basically taxes than the transactions of these third-party products. And it also sells its own products, these could be private labels of Amazon like Amazon Basics or branded products that Amazon purchases from the wholesale market and resells them as a retailer on its platform. So that's what we call the hybrid platform. So if you look at the percentage of these sales coming from the third parties versus Amazon retail channel, 53% of all paid units were sold by third parties in 2019. So this is a global number. And this number is even higher in the US around 70%. So that tells us that the major or the majority of the Amazon volume of sales are coming from the third-party products compared to the retail channel. And Amazon is not the only hybrid platform. So we have Apple App Store and Google Play as a hybrid platform. They have their own apps on the app store, but also they enable interaction between third-party apps and the users. And Zalando is also another example of a hybrid platform. So why we are interested in, in particular, dominant hybrid platforms because they have raised significant antitrust concerns recently, both in Europe and in the US and in other countries I believe will follow up the suite. And in particular, in the European Commission Amazon investigation, there was a recent press release about that, which was basically emphasizing important aspects of this investigation mainly tricky concerns. So the one concern was that Amazon might be favoring its own products. So these are Amazon retail products at the disadvantage of the third parties. And Amazon might be distorting access of the third-party sellers to the consumers. That's an argument made for the basically entry level of these third-party product sellers to the Amazon marketplace. And finally, Amazon might be using third-party sales data for its own product decisions in the retail channel. So there were three key issues. And beside these concerns, of course, these were also important concerns for the US investigation. So we also see another public debate going on, which is taxing digital platforms. And France introduced tax in 2019 July, basically imposing digital tax for Amazon around 3% of its marketplace activities, because Amazon passed the threshold of the revenue both from the global sales as being a big digital platform, but also its own sales in the French market. And basically Amazon made a counter-strike by passing entire amount of this tax on to its third-party sales commissions. Basically, Amazon announced that it's going to increase its third-party commissions for French sellers in the Amazon point-fair market by exactly amount of the tax. So it's important to understand this business model of hybrid platforms also to characterize what would be the optimal taxation policy for these platforms. And another fact is that Amazon has various fees, so I don't want to spend so much time, but it's just to highlight that first-point fees vary across categories. But there might be also the same fees across different categories. So module fees are around 15%. And these are the fees applied to the third-party sales revenue. And sellers mostly don't pay fixed fees. So these are commissions over the revenue raised by the third-party sellers. But these fees range a lot. So they can go from 6% say from personal computers up to 96% for warranties. Another observation is that Amazon activity as a retailer is also different across categories. So for instance, Amazon is big in entertainment of goods and products. And it's around 20% of its total retail activity is coming from this channel. So there is this variance of how big Amazon in retail channel across categories, but also how the commissions change across categories. So we would like to build up a model, basically try to understand the main ingredients of these decisions, how big I want to be in a given category. What does it depend on? But also given that I choose as a platform, my optimal business model, how this would affect the equilibrium properties, so prices. So what we are doing is basically we provide a micro-founded model to study trade platforms, in particular we would like to capture important aspects such as the consumers making membership and transaction decisions. So they decide to enter and then they also decide whether to purchase on the platform. And the third party products have this entry decisions, which was an important point of the European Commission investigation as an access to the consumers of the monopoly platform. And the monopoly platform can influence these participation of third party sellers and also consumers and also the transactions on the platform via tool tools. It can raise revenues from commissions over the third party sales, but it can also sell its own products. So this is an endogenous decision and make profits from its own products. And then we analyze how this dual role of a monopoly hybrid platform affects basically the prices, variety and third party sellers and consumer buffet, okay? So this is what we are doing and very briefly I would like to just emphasize the previous literature and how we are connected. Basically, if you look at the two-sided market literature, we have this either participation models that are basically standard of the nightclub type models, modeling participation decisions of bias and sellers on both sides or transaction models like the payment card models introduced by Roche Troll. And we would like to incorporate both because we believe that they are important for the e-commerce platforms. And also we want to have this between group negative externality, which is mainly arising from competition between sellers in a given category. So this is important aspect of our model because it will also affect Amazon or the platform decision of whether to present its own product or not, okay? And there has been literature on that, but they were looking at the impact of these externalities on competition between platforms where different in the sense that we just look at the hybrid platform model works when we have these externalities within the product category due to the competition, okay? And this is a model of a variety provisioned by a trade platform. But again, different from these previous papers, we enable platform to charge both participation margins so it can make revenue from participation margin, but also from the transaction margin. And these margins are distinct. So this is an important part of the paper. So in our paper, basically, we will generate these two-sided network effects endogenously because of transaction surpluses affected by the platform endogenously in the model. This will be clearer when I introduce. And very recently, there have been papers. So I think they are complementary in what we are doing here because they look at different aspects of these different investigations on the hybrid platforms. So I will not want to go into details, but I'm happy to discuss at the end. So we are different from those. In particular, we have a variety aspect differentiated products. So this is the key difference from the first paper where the products are homogeneous, except for one seller being offering a higher quality. And also from the second paper, they have a very different focus. So there are capacity constraints and there are different issues like the platform can provide access to the information about the sales. And we don't do these things. And also, there is a paper by Etro. And that's also different from what we are doing here. So let me go back to the general model. So what we are doing here is, basically, we are modeling trade platform. And this trade platform enables interactions between buyers, mass of buyers, and a continuum of differentiated sellers. So what does it mean enabling interactions? Basically, sellers decide whether to come to the platform by paying a fixed cost of entry K. But this is not a fixed fee to the platform. So this is exogenous. And buyers decide whether to visit the platform. But again, they don't pay a fixed fee. And they face this intrinsic visiting cost S. And then there is this, once sellers on board, they decide also the prices. And how the timing works. The platform first decides whether it wants to also own its own product. And if it has its own product, so basically these two stages could be boiled down to one. I just put them separately to make it clear. It sets its own price of the product if it is selling its own product. In other words, it's a hybrid platform. And it also chooses this commission on the sales of the third party sellers. Okay. And then third party sellers decide whether to enter. So this is third stage and they choose their prices if they enter. And fourth stage consumers decide whether to visit the platform. And if they visit the platform, they incur this intrinsic visiting cost S. And then they discover their match values to the products on the platform. Okay. So on the platform, consumers can buy from the third party seller product. So then they get this utility, the UI, which is basically quality of the third party good minus its price plus the match value to the product. And if there's availability of the platform product, they can also buy it. The platform product and that is gives them a different quality, which is VA minus price of the product. And again, they have this match value to that product. And they can also walk away if they basically cannot find anything better than what they have as an outside option. Okay. And basically in the, so why we call this model like a mix of Lugopoli because we have this symmetrically differentiated fringe sellers on the platform. So they are infinitesimal in the sense that when they choose their prices, they do not have an effect on the equilibrium price of the market. So they are small, but the platform has a massive product. So it has this impact on the equilibrium, which makes them different than a symmetric Lugopoli framework. Okay. So we use this mixed Lugopoli framework that was adapted by early literature on the trade and the general equilibrium models. Okay. And also we have this long run aggregated games aspect in the model that simplifies our lives as we will see very soon a lot. This is given by this free entry decision of the third party sellers. What we have in mind is that we think about third party sellers having this fluid entry decisions. In other words, the entry can happen any time. So they're basically entry stage. There is no friction. So that's why we consider this entry fluent at that will give us this long run aggregated games properties of the model. Basically that will simplify again the structure of the solution. And we also adapt this Weizmann search rule, but today I'm not going to talk about it, but just let me briefly describe what it is. So we have two different versions with the search cost. So one version is when consumer incurs the cost and then discovers all the match values on the platform, including his outside option. So this I will discuss today. This is what we call the two-sided market analysis today. But there's also other version where consumers know they are outside option before visiting. So this is basically the Weizmann search rule. And they decide whether to incur the cost to discover the other match values on the platform. And indeed we have some results, but this part is still work in progress and I'm not going to present it today. So but we are confident that our qualitative results apply there too. And what are the assumptions of platform is viable. So in other words, it has sufficiently the fixed cost of entry for the pre-sellers is sufficiently low. That means when the platform charges zero commission, there will be some sellers coming to the platform. And we also assume this match values are IID, the product. So this epsilon terms are IID with the Gamble distribution. So that's why we will get this logic type demands. But again, it will be a little bit different from the standard logic due to the fact that we will have this mixed oligopoly structure. And we will solve this multi-stage game using the basically sub game perfect Nash equilibrium. So maybe Ash is a good point to pose and see if there are any questions on the model. So you see as a question of terminology, where is it mixed oligopoly rather than monopoly with competitive fringe? Yes, I think that will be clear once described the demand. So mixed oligopoly because, yeah, let's go there. Yeah, so it's not the monopoly. Because I think you'll see has in mind that when the pre-sellers are choosing their prices, they can't affect the equilibrium. So it's basically monopolistic competition framework. He's right in pointing this step. But we think about also this mixed oligopoly, like Amazon having, say, massive products, right? Each product of Amazon could also be very small, but it has this mass of products that it controls the prices. And if they are symmetric, so that's how we call it, you know, its presence comes from holding this a big junk of the products in the market, OK? So because of the interpretation, but he's right in terms of solution is monopolistic seller competition. On a more technical point of view, I'm not sure I understand the setup perfectly. So there is no search. I mean, once you take the search class, you learn all the values, right? You know, you learn all the epsilon. But if there is a continuum of sellers, doesn't that mean that there will always be several sellers for which the epsilon is at the upper bound so that those guys are essentially Bertrand competing? So you know, no, OK. So what am I missing? So maybe I take that one. You're missing the tail property of the type one extreme value distribution if you wanted to be technical. Which if we have appropriate tails, yes. Oh, you got it? Yeah. Right. So there's always something larger enough, even in the limit. So in some sense, it's the limit properties, monopolistically competitive limit properties of type one extreme value distributions and other ones where the tail is sort of doesn't go to isn't too thin. So you do have literally you have an equilibrium markup under monopolistic competition, even with a continuum of firms. But it's a good point that if you're exactly right, if we had a bounded support, then they would be actually wouldn't be first one. It would be, as you said, you would have equally high. Everyone would have the best valuation out there and they would go to zero. So I will get out the way, actually. Thank you. Just a clarification, Oslem. The sellers need to go to the platform. They cannot, consumers cannot access the sellers without the platform, right? Yes. Yes, there is no direct channel. Yeah. OK. Any other question? No. Good. Good to go. So then let me start by providing the solution of the simplest scenario where we have this visit cost not binding. So suppose that all consumers come to the platform. So this S is not binding or it's zero, basically, that all consumers are on board and it's all about whether I buy one product versus the other or nothing. So if the platform is a pure marketplace so that it doesn't sell its own product, we will have the demand for fringe seller I in this expression. So basically, it is when the fringe seller says its price PI and the equilibrium price of the product is P. Basically, so we have the symmetric sellers, right? So we get, you know, there is a continuum of them. So since they are symmetric, we have this aggregate demand being the n times exponential of this V minus P over mu, that the P is the price of the any seller. And now if I choose a price PI, I'm getting this share of the total aggregate demand. But if there is a platform product available, then basically the aggregate demand of expression of the same demand equation of seller I will be different. And now it will also depend on the product of the platform. So this is A and the price of that platform product. And there we have this mass of the platform product. So that's why we have this m in front of the platform's share of this aggregate. And that's basically how it differs between these two structures of the platform business model in terms of the demand for the seller I. And the demand for the platform product is basically its own share of the total aggregate. So these are the demand functions. Again, if there is no, all consumers come on board and then they just choose between either buying seller, third party seller product or nothing. If it's a pure marketplace and they choose between third party seller product or platform product or nothing if it's a hybrid platform. And then the seller's pricing decision is basically I get my price, but I have to pay this commission, which is kind of a volar and tax for the seller. So I get one minus T of this price or the revenue. And I get it over my sale. So it's over my demand. Here I denote the total demand by A. Basically, this is the aggregate demand or the aggregate that we are referring to because this aggregate will not be affected by the seller pricing. So it's going to be independent of what the seller chooses at its own price. And it will have different expression between these two different business models. So aggregate will be this one when it's a pure marketplace. And it will be this expression, including the platform product when it's a hybrid platform. And then we solve this simple problem, basically. We show that the fringe equilibrium price is simple perceived marginal cost by perceived because this is marginal cost divided by one minus T. So it's like at volar and tax inflating my marginal cost. And then the seller, each seller puts a standard logic markup on this marginal cost. So this is important that this pricing is the same regardless of the business model chosen by the platform because of the fact that, again, fringe sellers are infinitesimal, so they cannot affect the equilibrium. And then once we take this price at the equilibrium, we can write down the fringe variable profit. And if you take, basically, the price times one minus T minus C. So this is the margin per unit that the fringe gets. It boils down to this expression. Basically, this is the fringe markup times one minus T. So how much it can keep as opposed to paying to the platform the tax. And then this is the demand at the equilibrium price for the fringe seller. And again, the aggregate is not, now it's a function of just the T, so the commission of the platform. So the fringe sellers will come until the point that this total profit is equal to zero. So that will determine basically when the in equilibrium, the aggregate will be equal to this expression. And this is given by this pre-entry or the zero profit condition of the fringe. Important properties of the aggregate is that it is a decreasing function of the commission of the platform. So when platform raises the commission, aggregate demand goes down. And it's commission zero. Remember at the beginning, I assume that the K is sufficiently low so that there's some fringe entry. In other words, A star zero is about one. So this was our assumption on the K. And then we characterize that there also exists a T hat. This is the upper cutoff. Basically at T hat, fringe prefers to leave the market because it's too high commission for that. So this is when basically aggregate is equal to the outside goods demand, one. So then given these properties are defined, now we can also write down the number of fringe products which will basically solve the equilibrium aggregate being equal to the definition of the aggregate to start with in the pure market place case. So basically that will pin down number of fringe products as a function of a commission set by the platform, okay? So then I can move to the pure market place equilibrium. So the platform basically chooses the commission that maximizes this profit. So which is T times the price that it. So this is the tax over the revenue from the third party product sales. So I write the third party product total demand as the aggregate minus the outside demand. So then this is total demand that the platform product goods have. They are not owned by the platform but just the third party products. Again, given the pricing behavior of the seller, each seller, this is what we characterized before and the A star T pin down by the zero profit condition. And then the platform basically choose the optimal fee between the critical thresholds that I defined. Basically the platform optimal fee will be less than this cutoff T hat so that there will be some fringe firms coming to the platform in equilibrium. And then the platform in equilibrium chooses T such that aggregate value in equilibrium will be equal to one minus the fraction of the elasticity. So what are those elasticity? So if you look at the platform objective function, so you can think about it basically maximizing it is the same as maximizing the log of this function. So it tells you already that there will be some elasticity here. So the first is the elasticity of the aggregate demand with respect to the commission. Remember the aggregate demand is going down with the commission. So this elasticity is negative. So this numerator is your, this term is negative. So indeed this is an expression about one. And in the denominator, we have the elasticity of the fee revenue. So when the platform increases this commission, it trades off basically how much I can make from the margin. So it is my fee revenue per unit and this elasticity. But then I lose the aggregate demand. So that is basically the loss. So that's the trade off that pins down the optimal commission. And what we correct rise now is the illustrative equilibrium basically how it looks like. I just draw here the AT. So again, remember this equilibrium aggregate which is a decreasing function of the platform commission. And at some point where AT is equal to one, fringe is not coming anymore. So this is that T hat upper bar T hat threshold. And the right-hand side of the equilibrium condition, here I characterize for this parameter values that it is increasing. But in most of the parameters, it is indeed monotonic, but it might be non-monotonic. In other words, we might have multiple equilibria for some values of for instance, in particular when the differentiation is very high or the marginal cost is super low, that can happen. But again, this is a space rather tight. But we don't need the uniqueness. Indeed, our properties are also applying as long as we have the existence because we will do some comparative statistics analysis to the equilibrium and that will be clear next, okay? So what we do here, for instance, when we think about the fixed figure of entering for the fringe, if K goes down, so it is basically easier for fringe firms to operate on the platform, so the fixed cost is lower than this AT equilibrium aggregate bill shifts right and that will result in platform charging higher commission in equilibrium, okay? I think this is a good point to pose to get if there is any question on the pure marketplace equilibrium, no? Then we go and ask a question, what would happen if it was a hybrid platform? So then as we know already, the fringe seller's pricing behavior is the same because they are infinitesimal, so they are not affecting the aggregate. And again, since the prices are the same, their cost, their profit expression is the same and the zero profit condition will again pin down the equilibrium aggregate as before and that will give us the neutrality property of this aggregate demand or the aggregate. That means basically the equilibrium aggregate is independent of price of the platform product and the quality of the platform product. And that will simplify our lives a lot and that will be very important in analyzing this comparative aesthetics of the equilibrium, okay? So then platform profit will look like this new expression. So before it was commission times price of the fringe seller times the demand from the third party sales. Now I also have my own product, so that's why we deduct this VA from the aggregate demand. So VA is basically the how much of the platform products gets from the aggregate demand and I have my own margin from this aggregate demand, okay? So VA, I didn't put it here, but it's basically M times exponential small VA minus PA over me, okay? How much I get from the aggregate demand for my own product. And then if you solve this equation, first characterize the equilibrium price of the platform product, that gives us a very nice intuitive formula which is basically what is the opportunity cost of selling my own product on the platform. It is the cost of this product, the marginal cost, plus how much revenue I lose from the third party products when I sell this product instead, because they are competing, okay? So that's why we have this, the loss revenue, if I sell my own product, this is basically commission times the price of the third party product that I would have sold if I didn't sell my own product, okay? And mu is the standard markup on top of this opportunity cost. And one observation is that in order to, for instance, if you have the price of the marginal cost of the platform product, the same as the marginal cost of the fringe product, so CA is the same as C, we can see the platform product is higher price in equilibrium than the fringe product, okay? So at the equal marginal prices, marginal cost, the platform product is higher price. But given that the observed Amazon products being priced lower in general, again, this is anecdotal evidence, in order to get that, basically, in order to have the platform product being lower than the fringe product price in equilibrium for a given commission level, marginal cost of the platform product should be sufficiently low. And this can be indeed justified by the fact that the Amazon might have some buyer power in the wholesale market when it purchases, for instance, the branded products as opposed to these small retailers who are selling them as third-party sales on Amazon, okay? So the CA could be indeed considered as the purchasing price in the wholesale market of the Amazon product, okay? And then, yeah. If I may, before you go on, there's a question actually at the same question. So can you tell us again why it is that the existence of the first-party product does not affect third-party pricing? Does that affect the seller's pricing? So because of the, you know, the sellers cannot affect the fringe sellers, the third-party sellers are so small, so they cannot affect the aggregate demand, right? So when they are pricing, they take this aggregate demand constant, like independent of their pricing behavior. So what they basically try to fix is how much I get from this aggregate. Given that aggregate's fixed, what will be my share of this aggregate? And this is my margin. My own pricing does not depend on the pricing by other sellers, is that? Because when you introduce a big seller, the first-party one, it doesn't take the best response. Put it this way, price expression does not depend, but price expression is a function of the commission. But the commission in equilibrium will depend on whether platform product, how it is priced and how, whether there is platform product or not, okay? So the expression of the pricing for the optimal pricing of the fringe seller is the same. So the best response does not depend on other sellers pricing? Exactly, exactly. Because of, again, the fact that this each seller is so small, it cannot affect the equilibrium pricing behavior of the rival, okay? Okay, okay, move on. Okay, so then the lemma tree basically characterizes the equilibrium for the hybrid platform optimal commission. Again, there's a cutoff, but this is a different one now, because given that platform product is available on the platform, basically then this maximum, the commission above which there is no more fringe seller is now lower, right? Because it has to also compete against the platform product. So there's this another cutoff which determines if platform charges a commission above this level, there basically will be no third party sellers. In other words, platform turns itself effectively a pure reseller of its own product. And what we show is that if platform product quality is above a cutoff VA tilde, so if I have a sufficiently high quality product, indeed I prefer to turn off the third party sales channel and became a pure reseller in the model, okay? Because they are creating competition against my product and it doesn't pay off to have their variety if my product quality is sufficiently high. But if my product quality is low, then there exists a threshold to optimal commission between these levels so that that allows some fringe firms to enter, okay? And what we will do now is basically consider this area where platform is effectively hybrid. In other words, some fringe firms are there and platform is the own product and ask the question of how this equilibrium changes if the platform products gets better, okay? So basically if the quality of the platform product goes up within this region or the platform marginal cost of the product goes down, how does this affect the equilibrium? And we show that this is the main result of the paper that the hybrid platform charges a higher commission to third party sellers if its product gets better, okay? So let me describe the intuition of this result because given that it's a key for the rest of the results. So what happens when platform quality increases? So obviously platform product gets better so it's still some demand from the third party products. So that means that in equilibrium there will be fewer fringe firms coming to the platform. So this n small n will go down. It goes down such that this increasing quality of the platform product normally will increase the aggregate demand, but because of this free entry condition remember we show this aggregate demand being constant in the platform quality. So this aggregate will not change when the A goes up. So because of this fringe basically variety going down neutralizes the quality increase of the platform product. But then what happens is that platform gets a larger share of this given aggregate demand when its quality is higher. And so it puts more weight on its profit expression for the retail chain revenues or the profits as opposed to the profits that it makes on the third party sales activities, okay? In other words, this fraction increases when the small VA goes up and this market share of the third party products goes down when the platform products gets better. So then I'm balancing these two channels and I put more emphasis on my retail channel. So then it pays off to raise the commission to the third parties because they are my competitors, right? So that basically leads a higher commission in equilibrium, okay? And the same argument applies for when marginal cost of the platform product goes down. Then the corollary is that equilibrium share of the platform product increases in the market when its quality goes up or its marginal cost goes down, okay? So that can tell us some nice predictions if you compare these across categories, I show you these fees at the beginning. Indeed, if you remember, I told you that the Amazon retail channel is big for the clothes and accessories. And these are the clothes and accessories also has a relatively higher commission, for instance, 17%. I'm not saying that our model explains everything there because also the categories have different views or the different differentiations across categories, but this is just like total evidence that kind of, you know, but we predict is that when platform product is big, we predict to have the commission higher in equilibrium of our model, okay? And the consumer surplus in this model is basically the log of the aggregate demand. And this is very simple because the aggregate demand basically tracks how consumer surplus changes. And we know that the aggregate demand is function of the equilibrium commission set by the platform. And as we have already shown, this equilibrium commission goes up when platform product gets higher quality or lower cost. So basically then this is bad news for consumers, okay? Because of the fact that again, increasing quality of the platform product is neutralized by the lack of variety due to the free entry of the third party products. And what happens is that because of the business model of the platform balancing these retail profits versus the third party profits, platform puts more weight on the retail revenues profits. So that increases the commission that results in the lowering the consumer surplus, okay? But this is remarkable because in general, we know that the quality increase is good news for consumer welfare in standard IO models. But this is not true here due to the platform's business model. So in other words, Amazon was not able to control the tax on the third party products, there wouldn't be this channel that it affects the consumer welfare negatively, okay? So that's an important property. It's basically related to the, it's driven from the business model of the platform being hybrid, okay? Aslan, you have- Yeah, two minutes, no? Yeah, a couple of minutes left. Okay, so then I have to be very quick here. So I think I don't have time. What we did is basically, we look at the platform optimal choice between pure market place versus hybrid and we show that again, if the platform quality of the product is below some cutoff, platform prefers to be a pure marketplace. So this is this profit. And if it is above a cutoff, it prefers to be hybrid. Why it is the case? In general, if there is no fixed cost of the platform product, the platform is always preferred, preferring here to be a hybrid platform because it's always good to have some variety. But when you have some fixed cost of introducing your own product, then basically just creates this cutoff point. But it is always basically good news for the consumers if the policymakers ban the hybrid model in the model, okay? And then we characterize basically the oldest thresholds where the platform chooses which market mode. And important thing is that when the product differentiation in the market increases, the region where the hybrid mode becomes profitable increases. Okay? So this is another nice prediction of the model. Okay. So given that I'm over time, so in the two-sided market, so let me just go direct and say what we have as the main results that I showed you hold here. So in other words, even though we have the elastic participation demands on the consumer side, we get again the key result that the hybrid platform equilibrium commission goes up when its product has a higher quality or lower cost so that the consumers are again worse off. And again, banning the hybrid platform mode in the market benefits consumers when also be allowed this elastic participation. So two-sided market has nice interesting properties I'm happy to discuss after the talk because I think we also generate non-linear network effects in the model because of this endogenous transaction surpluses in the e-commerce model. Okay. So let me go direct to you and this is my last slide. So final points. So I showed you at the beginning this concerns about the steering. So what we can do is we can analyze in our model what does it mean steering? So just very shortcut introduction basically if platform has an ability to raise the quality of the third party products. So the small V it prefers that they are better because they are stealing demand from outside good. But it also has a possibility of increasing its own quality VA or N. Again, it prefers to do that. But if there's a trade of like if I want to push my product above when I am pushing the others product down let's suppose that steering does this. Basically I'm increasing perceived quality of my own product at the cost of lowering the perceived quality of the fringe products and this becomes profitable if platform product has a high quality. Okay. It's not always the case but if it has a high quality it prefers to do that but this is exactly the case where the commissions are high. So that's the point where this could be harmful. And for the taxation point that I raised at the beginning so basically unit tax on the platform product will benefit consumers. Again, that will lower the commission to the third parties but what happened in the French case taxing the marketplace revenues is not a good news because that basically lowers the weight that the platform puts on the third party revenues. So that basically results in increasing the commissions to the third party sellers. That's what we observed also from the Amazon reaction. I stop here. Thank you very much. And I'm happy to take other questions. Thank you very much as them. So now he's keep our eyes on and he's going to discuss the paper, Eski. Sure. So let me start by saying that I've never worked for Amazon or any other interested party here. I think when we're dealing with these sort of big policy relevant issues, it's worth reiterating that when it's the case or disclosing when it's not the case. So I don't think this paper needs a lot of motivation. This is obviously a big question. A lot of people are asking the same question. I think the key insight that I got from here is the hybrid model obviously makes life tougher for rivals. If Amazon is offering stuff, it's harder for rivals and part of what makes it harder is because then Amazon wants to benefit its own goods and so raises the fees. This tougher life for rivals is then reducing product entry, reducing variety, and that's an effect that makes consumers worse off. Now, why is that the dominant effect? Because we also have the effect that Amazon's stuff is good for consumers. Why it's necessarily the effect that the hit on rival entry is the dominant effect. That was a little bit more mysterious to me and it's hard to know going back to kind of Alex's question whether this is about a mysterious properties of what's going on in the tails and how thick these tails are and so on. And with different distributions, God knows what happens. Simon's laughing. I don't know whether that's because I'm wrong or just he gets as confused about tail properties as I do. So that's kind of the big picture of what I took from this. I'm gonna give a few comments to some more specific to the model and then sort of zoom out to the big picture a little bit. So I mean, one question here is, is consumer surplus the right measure of welfare? So we have other ways to measure welfare like total surplus and you know, Mancune Winston tell us sometimes as excess entry went when there's fixed costs. So kind of jumping from that to policy if what I was interested in doing was maximizing consumer surplus here, I would force Amazon to set the commission rate to negative infinity and get everyone to come in and be better off. So there should be some trade off against these fixed costs of entry and it's just something to think about. The action here is coming from entry and so these fixed costs are playing an important role in the model. And so I would have appreciated understanding a little bit better to think about what these fixed costs are and where they come from and in particular the extent to which they're endogenous. So I mean, one big thing for these small sellers is do I have to set up a distribution network or can I use Amazon's distribution network? You know, if Amazon is doing fulfillment for me is that like a reduction in the fixed costs for these rivals? Would the story be very different with a two-part tariff rather than with just a percentage story? I don't know, but I wanted to know. So I think those were the things that I wanted to say, like I wanted to understand the fixed costs better, you know, these Gumbel properties, two-part model, sorry, two-part tariff. The other thing, so it seems that every time I get to discuss a paper it's about Amazon being a reseller so the last paper I discussed was, you know, by Andrei Tahao and Julian and there all of the action came with the outside option to the platform. You know, it was kind of the competition for the platform and the direct sales channels. So Patrick asked this question and there with the endogenous outside option they got kind of almost opposite results. So that would seem one to think about. Now, I'm actually not completely uncomfortable with thinking of Amazon as the only game in town or, you know, if the Apple Play Store is the only game in town, but I think it's something to be aware of at least. Going back for a second to the big picture, the paper sort of started out by reminding us of the, and Oslo's presentation as well, of these key concerns of the investigation, using the data for your own product decisions, steering and other self-preferencing and distorting access. The paper is primarily focused on the last one, I think. I mean, I think this change in tea is about effecting the access of the third parties. And so, first of all, I think it'd be helpful for the paper to kind of, you know, having set up this motivation to be direct about what it's addressing, we had a little bit on steering at the end so that was helpful. You know, one thing on steering and also in the model, I didn't exactly understand how to think about this capital M, this mass that Amazon's own product sells because, you know, demand ultimately is endogenous. So thinking of it as the size of demand is a little bit confusing. I think that Yossi had a related question around the sort of micro foundation here and how to think about this M. You know, one aspect that wasn't talked about at all, but I think is one that people do worry about is how to think about the sales information from others and what that does. And that's a subtle issue. I mean, look, one paper doesn't have to address everything and I don't think it should, you should be clear about it and having sort of set up, you know, these are the key issues. If you're not going to address them, you should tell us what the effect of those other issues are and the extent to which they interact with this, you know, I mean, does the sales data, should I think of that as affecting the platform's quality or the size of demand M? Does that make a big difference in the model? I'm not sure that it does here. I mean, having set that up, it would be useful to return to that. And probably out of time, but in any case, I don't have much more to say. I, you know, I didn't, I hadn't thought through this mechanism before. There's a lot of kind of cool modeling technique in the paper as well. You know, I think the presentation and the focus shouldn't be on that modeling technique for the sake of technique, because ultimately there's a big question that we as a community need to address and get to grips with. And this is a good help on that path. Thank you Eski. A lot of comments from him. So rather than answering all of them, maybe you can pick one and answer shortly and then we have a couple of more questions. Thank you. Maybe I can just say thank you. No, I pick up the, I think you're completely right. So just maybe start from the last points about the policy implications. So our first objective in this paper is come up with some tractable framework. And then we of course would like to use that to address all of, you know, potential policy questions like tax and steering. And also I think one can introduce the uncertainty of demand and platform having some information that the third-party seller is not having that. And these could be studied once we get the tractability. And all these techniques and the modeling assumptions basically are really getting there for in terms of tractability, because we know that it's so hard to get this tractability in this, you know, two-sided market models and at the same time being realistic to capture their facts. So that's the first step basically, but I completely agree, given that this is a working paper draft that we send you. So we were not very clear about what we are doing or planning to do at the beginning, but you are right. I mean, in the final write-up we will be very clear about this. And then the issue about the, like I agree this direct channel is important. Though we have other important differences from the HAJU and ETAL paper, for instance, they have homogeneous products. So they never get hybrid mode profitable unless they have a direct channel. So they generate hybrid mode basically from the direct channel, but we have hybrid mode profitable because of variety introduced by the third party product. So product differentiations crucial here. And they have mixed strategy equilibrium. And so it's hard to make direct comparison that I agree that we have to make and clear clarify this. Two-part tariffs, as long as we have homogeneous sellers, we work on this. If you introduce two-part tariffs then the platform captures everything from sellers. So how realistic is this? So we were also questioned on this dimension when we had the other paper. And, but we have also extension in the current paper on the heterogeneous sellers in their qualities. So then we have some improv marginal sellers that enjoy some surplus. And that would make also reason to go for the total surplus analysis because currently sellers have zero profit anyway. So that's why we only consider consumer welfare. But if we have heterogeneous sellers, then I think what matters for us is that the platform cannot capture their total surplus. And that we capture by saying that you've just imposed commissions rather than having very rich contracts that enables the platform to capture everything. But you're right pointing out so this should be also clarified. Thank you again. And the tail properties I leave Simon to address that. Thank you. So, thanks for that. Let's see, there are a couple of questions. So the first one was by UC speaker. UC, are you here? Do you wanna ask a question yourself? Yes, well, I was wondering about the micro foundations of the demand structure. But I guess it's the usual, the usual discrete choice setup, isn't it? Yes. So we think about this. So you can think about like they're like, continuum of products and mess of them is owned by Amazon. But then every other product is infinitesimum and owned by the independent sellers, okay? And then if we have this IID draws, basically what we are telling is that, Amazon product because it tell Amazon or the platform owns a mass has a higher max draw because it has a max of, max of M draws, mass of the raw. So that's why it gives a, presence of the Amazon product. But again, what we're assuming there is the Amazon products are symmetric, okay? So they all generate the same VA. So this interpretation applies and this is one micro foundation. And this model also adopted by again, trade literature in the different ways. But I think this is the micro foundation that we have found from this literature as well. And is it realistic to assume that otherwise all products have the same substitution properties? Very good question. So we also think about introducing some nested structure. This is work in progress. I shouldn't promise more on that, but you're right. Maybe substitution say, because there's another thing, for instance, Amazon fulfillment, right? So then that also adds some kind of Amazon thing on third-party product, which is different from an independent third-party product own fulfillment. So we think about introducing this type of differentiation difference between the nests and within the nest a different differentiation. So nested structure, but we haven't finalized this yet. But currently you're right. He assumes symmetrically differentiated products. Yeah. Okay. Okay, thank you. We have a few questions still. So Michele Polo, Michele, do you want to ask your question? Even though Simon already sort of answered succinctly. Yeah, yeah. I think that Simon already confirmed my intuition, yes. Okay. Gary, Gary had a question. Gary, you wanna? Well, Simon answered it also, but I... Gary said yes. He said yes to both. So I thought maybe... But I had the same question that Heskey had, is that this dominant effect is that consumer surplus is lower with higher quality. So you would want to ban Amazon, but there's this issue that, why aren't consumers getting some of the benefit of this higher quality? It's like getting all the past, there's like a negative pass through in this paper. And it's kind of wondering why consumers don't get any of the benefits of this higher quality product. And that all this is steer, you're steering the consumers to the higher quality product. Consumers get none of the benefit from this. I had kind of a follow up question, which is that if you ban, and this is a question I already asked you as a last time you presented, but I thought that it would be interesting to see if you have thought about it, also that if you ban Amazon from selling its, from using the hybrid product, there's a possibility that it might, instead of becoming a pure platform, it might want to become a pure seller. And have you thought about the trade off then? Yes, I think you're right. This is basically choice between pure reseller versus pure platform. We thought that this was not the most important choice that policymakers are interested in. Basically they were more asking to question of pure market place versus hybrid, but you're right pointing out when it would become, if I ban in between region what would happen? So we can say that if the quality of the platform is closer to this lower cutoff point, then platform chooses the pure market place. And if it is closer to the higher cutoff point then it chooses the pure reseller mode in the model. But again, we don't have any other reasons to go for a volume business for instance. So there are no fixed costs for Amazon. One can also argue that this retailing business is costly. So we have the Andrea, Julian, and Julian, they have paper basically arguing that why retailing is costly and why marketplace models are very effective because of this reducing the marginal cost or the cost of activities. And we don't have that in the model. So if you had other things, then it makes even less likely that the platform goes for this reseller mode. But I mean, that kind of it's a theoretical question, but I don't know how practically relevant to consider Amazon turning itself a retailer. Suddenly, given that it has a very big processing happening on the marketplace. At least in some segments. Yeah, but it's true. It started as a retailer, right? So this is another evolution of Amazon. Amazon started as a retailer of books and it turned itself a marketplace. But we have the Apple, the opposite direction. So it started as a marketplace and then it introduced its own apps. So it's an interesting question why we have these two different paths. So we are not, yeah, this isn't, I agree. So I think one needs to compare in the other question. So if I have this choice of pure reseller, what I am doing is I am the monopoly, right? I put my monopoly markup. But if I'm offering the marketplace more, okay, I create some kind of distortion by taxing them, but it's not the monopoly, right? There's some competition. So that will be the tradeoff that one needs to look at to analyze, which mode dominate. But I think as long as the quality of the Amazon product is higher, maybe the monopoly mode might be better. I don't know. I'm just speculating, but that might be some room for further investigation. Okay, and I'm sorry, but I jumped in before you had a chance to answer, Gary. So I don't know if there was something you wanted to... No, I mean, we are already time. I'm happy to answer, but if people, you know, if you want to stop... So you're right, you're doing my job as a timekeeper. So this is the end of the hour. So we'll stop the recording now, or actually, maybe there are some interesting things that are going to happen, but I think the rules are that I'm gonna stop the recording now. But we still have a few questions, okay? So some people are supposed to ask their questions now. So anyway, those of you who want to leave, you may do so now without offending the speaker. So thank you a lot as them. Thank you. Thank you.