 Let me first of all disclose that I've done consulting work for Amazon on antitrust matters. There is indeed a big debate on the double role of some hybrid marketplaces as Amazon and also others as hosting third-party products on their platforms and at the same time providing their own products. And there is a view sometimes called the New Brand Asian View associated with suggesting that there is a systematic incentive for these platforms to promote their own products setting worse conditions for third-party sellers and for instance in particular increasing the commission rates applied to these sellers and these can harm consumers in the long run. From an economic point of view there is a first obstacle to this view due to the business model of these hybrid marketplaces because they monetize at the same time on third-party products through the commissions and of course on their own products through diet margins. So it is not obvious that there is an incentive to bias customers in one or the other direction and it is even less obvious that if there is a bias in one direction this goes against the necessarily the interest of consumers. There were some papers recently published here I mentioned some focusing on this issue, CAFAR and the policy paper did stress the importance of business models for antitrust purposes. The other papers I have one from last year emphasizing both static and dynamic incentives and a platform that is there to stay usually as an incentive to internalize both static and dynamic incentives over its product introduction. There is a paper by Aguete and Wright that is now forecoming that stresses another aspect, hybrid platforms strengthen competition on the platform and this creates benefits for consumers in itself and they also stress other issues related to self-preferencing and imitation. However my talk today is more related and as a matter of fact inspired by two papers that came out last year. One is by Yuzuke Zenio who happens to be discussed on today and the other is by Simon Anderson and Oslem a bit of a day fully and these papers do a lot of different things but I see the basic frameworks as providing a common framework that they think it's the appropriate one to analyze the welfare impact of these hybrid marketplace because it is based on a model with product differentiation between all the products present on the marketplace. In these papers they use discrete choice models leading to a logic demand systems and they adopt monogalistic competition between sellers. This is becoming the standard analyzing platforms for the simple reason that there are so many sellers on Amazon and also other platforms that this is not the natural way to go. We also endogenous enter of sellers to endogenize the set of products that are made available on these marketplaces. Now these two papers reach in the baseline frameworks two different conclusions and this is what inspired me to explore more of the issue. So in Zenio so the two papers have a lot of different assumptions as a matter of fact. In Zenio that adopts a specific commission on sellers the hybrid marketplace is basically neutral does not change the commission applied to the party sellers and is ultimately neutral on consumer welfare. While Anderson and Bederefoly adopt ad valorem commissions as I will do and this is the critically relevant assumption because Amazon as others do adopt percentage commissions and they emphasize that there is a systematic incentive of these hybrid marketplaces to increase commissions on third-party sellers and through that to reduce consumer welfare and the mechanism is a demand shifting mechanism when they introduce their own products they have an incentive to rise the commissions on third-party rivals so that some demand is shifted to their own products where they make more profits. So I try to explore more this issue in a different kind of model going back to the traditional micro foundations of monopolistic competition models in the in the sense of Spence and Dixie Stiglitz so with represented agent micro foundation that can deliver more general demand systems including the logit one but also other as the iso elastic or linear demand systems and try to understand whether in this more general class of the demand systems there can be additional mechanisms. So let me anticipate the main theoretical this is a theoretical paper of course points. First I confirm something that is already present in those papers so that in this framework the hybrid marketplaces for a given commission so if the commission rates are not changed does not affect consumer welfare. There's an neutrality that is inherited by a class of models with free entry as we will see. So any impact of these hybrid platforms on consumer welfare and as a matter of fact also on user welfare because under free entry they are the same will happen through the changes of the commissions on sellers. If the commissions are increased this will make consumers who also offer to higher prices and less variety. If the commission rates are decreased consumer welfare will actually increase and actually in this more general environment what happens is that there are two mechanisms that are working in the opposite direction. One is the demand shifting effect as the one emphasized by Anderson embedded there fully and the other is actually a mechanism working in the opposite direction. I called it an extensive margin effect and the intuition here is that the hybrid marketplace introducing its own products is necessarily giving up to some sales of third party sellers and so some commission revenues from these the party sellers commission revenues that come for free in a sense to the platform and so it has an incentive to reduce its commission rates which can reduce the prices of these sellers increase the gross profits and recover entry of some of these sellers. It can also attract more purchases by consumers and in this way it can expand the commission revenues. The two mechanisms are working in opposite direction. Can I ask a clarifying question? Actually Amazon has got does free things it's a marketplace it sells it's a retailer which sells product advice from other people and it also sells Amazon branded products and when you when you're speaking about hybrid marketplace I assume you have a marketplace and you've got one of the two other activities which one are you focusing on? An hybrid marketplace is here defined as a marketplace that provides both own products and third party products on which it monetizes only through the commissions. So the fact that Amazon is a retailer doesn't play your role in your analysis. When you say when you say own product do you mean Amazon branded product? You should think as a private level product if you want but with slight modification you can think also as an activity as a first party retailer so but the primary interpretation if you want can be in terms of private labor. So products produced and commercialized by the same platform okay which is the same kind of assumption that as in these other papers. So I was saying so there are these two effects going in opposite direction so the result at the end is not obvious there is a class of demand functions that is quite prominent because it's the one with isoelastic demand function so the standard big stiglitz case where when the marketplace and the seller space the same constant demand elasticity for the products actually you have neutrality in the spirit of the Xenio papers as a matter of fact and the commission rates are not changed by in hybrid marketplace compared to a pure marketplace one that doesn't sell its own products and then you can have results going in one or the other direction. So I will come back to this let me mention some of the literature here there are some papers that came out recently quite interesting and focusing also on other aspects. This framework is related to work on stackable leadership with free entry. The leader here is the platform because it's set in the commission before all the others and it's also set in the price of its own products as a matter of fact and to the tiers of monopolistic competition recently there has been some movement in this literature with more general micro foundations that are usually applied in general equilibrium models of trade macro. Here I will borrow some of these methodologies that you will see from joint work with Paolo Bertoletti applied in a partial equilibrium contest which is the one we care about. There are also some empirical papers recently structural estimation models based on the logic framework which is quite tractable but they don't consider at the same time endogenous commission sent into your sellers so I think there is more interesting work to do in the empirical front. Ultimately here I will leave you with a clear empirical prediction of this framework that they already anticipated so in this framework basically the impact of the introduction of new products of the marketplace in its own platform will affect consumer welfare in a way that depends on whether the commission rates on others are increased or decreased so a very precise prediction and so one can empirically want to do that this is theoretical paper but the commission rates set by Amazon are publicly available especially for the last five years country by country and product category by product category if you look at the United States but the picture is similar as well you will find that the most most product categories the commission rates did not change there are some product categories where they were reduced and there are a couple of product categories where they were increased by I think 2% so overall there is not I don't see a systematic correlation between increasing commission rates and introduction of products by Amazon which happened in most of these categories as a matter of fact but there is empirical work there can be done trying to screen between the two different consequences okay so I will move to present the theoretical framework so the theoretical framework is relatively simple timing of the game also then while you go on with the analysis to find the optimal the commissions is like an optimal taxation exercise ultimately things becomes a little more complex and I will rely on intuitions but so let me start from saying that this is a model where consumers purchase a bunch of goods and goods the number will be actually endogenous and we will have monopolistic competition between the sellers so I will rely on a traditional tiers of monopolistic competitions but following recent micro foundation I will derive the man's system and not from the direct utility as in Spain's tickets tickets but from the indirect utility this is of course the two methods are equivalent in a sense but this will be quite convenient as you will see in a moment so let me first introduce a price aggregator which will be part of the indirect utility of consumers the price aggregator depends on the prices of all the products present on the marketplace and so you have to think of it as a measure of the quality if you want of the marketplace and the restriction that I will impose is that this is an additive price aggregator so each product available in the marketplace has a price and it provides an additional surplus to consumers which is this V function so it's a surplus function that is decreasing in the price and convex and and this applies to every good present in the marketplace H is an exogenous component from exogenous Jesus so if you have more goods or if you have lower prices of these goods the value of this price aggregator the quality let me call it the speaking of the marketplace increases and the other way around and given this let me define the indirect utility of consumers through a log linear version so these are log linear preferences there is an outside good where you spend the rest of your income and this specification is actually also used by a recent paper by knocking shoots in my paper I actually use a more general version but I find it convenient to use this because it's particularly tractable and the reason is that when you derive the demand system for all the goods you just apply the identity by basic properties of preferences and you get a demand system as this one so at the demand of which good depends on a ratio at the denominator you exactly find that the price aggregator now this which depends on all the prices but since we are going to consider monopolistic competition every single seller will take this price aggregator as given because every seller feels itself negligible and cannot affect with its price choice the overall quality of the marketplace however at the numerator you will find the you should the the slope of the surplus function which depends on the price and each seller understands that the price shapes the demand through this aspect let me give you some examples so you will be convinced that we are in a very standard world everything depends on the surplus function so if the surplus function is a power function like the price to the power of one minus epsilon you get isoelastic demands so this is the Dixie still it's a story basically if you have an exponential surplus function you get an exponential demand notice that here I am not writing down the denominator where you have the aggregator because it's going to be taken as given by the sellers but if you consider that you will see that the exponential surplus function generates a demand system which is basically the logit demand system within this micro foundation the last example is also useful sometimes because provides a generalization of the linear demand through the parameter gamma if it's one you get a linear demand or you can get a very rigid or elastic demand but these are examples in general you can take any surplus function satisfying certain properties and you get a different demand system with different characteristics and the elasticity of this surplus function are what will matter mainly so let me mention the elasticity of the surplus zeta tell us how price changes are reflected the changes in the surplus of consumers and the demand elasticity which is the second order elasticity will of course affect the pricing now notice that these two elasticity in general can take different shapes in case of power surplus function they are constant in the logit model they are increasing the price and they are identical and in general they can be increasing or decreasing and they determine the shape features of the demand system there is also relationship between the two that I report there so as I mentioned in the paper I have a more general micro foundation but I will not discuss it here in the paper I also have products that are heterogeneous on the marketplace today to make it a little simpler I will assume that every product as imagined at cost c everywhere okay and they as I was saying to Jacques I will have a number of products by the marketplace and this is exogenous for our purposes now if there is an extension to endogenize which products you want to introduce in the paper and all the others the total n minus m will be the products of the sellers and this number would be endogenized in the model so the timing the timing is the one you should expect there are four stages and there is a first stage in which the marketplace is setting at the commission rate tau between zero one so ad valorem commission in the paper I have the extensive to a specific commission to relate with the work of Zeno adopting that assumption but here I will focus on ad valorem commissions as Anderson embedded a fully there is a second stage in which the marketplace is setting its own prices please try to remember that when you see an upper bar on the price p upper bar for instance generating a surplus of the upper bar I'm referring to a product of the marketplace and here I'm assuming the same surplus function for all the product of the marketplace okay upper bar marketplace without the upper bar I'm talking about right products of the independent sellers so notice that I'm assuming a commitment also on the prices here this is not necessarily the right assumption to make it's giving a higher market power if you want to the marketplace in the paper that I have an extension where all prices are set at the end simultaneously and that increases the chances that an hybrid marketplace increases consumer services so here I will adopt this conservative if you want a sanction commitment on commission and prices third stage endogenous entry of sellers on the marketplace they pay fixed cost and they decide whether to enter or not and in the last stage of course I have the pricing of all the sellers under monopolistic competition in the paper I have an extension to Bertrand competition and of course the markups will be different will be higher in that case that applies if you have a limited number of course of sellers in some product category but the spirit of result doesn't change because the the neutrality that I mentioned before applies also there as a matter of fact it was derived in a number of applications of free-enter models with strategic interruptions and finally just to give you an overview of what you will find in the paper but I will not discuss today I have an extension to endogenous spending in advertising by the sellers so that's a word where each seller can also invest in ads for product discovery to attract clicks to its own product and possibly get it from some other potential sellers and there the focus is on how the hybrid marketplace will affect the total payment of the sellers in commissions and ad fees so it's a little richer but the spirit of the result doesn't change very much so I will focus on the baseline model so if there are questions this may be a right moment and I will of course start from the last stage looking at the pricing of the sellers so the sellers here face a certain commission Tao already chosen and the profits that you see in front of you at the denominator you have this price aggregator which affects the demand but is taken as given in the monopolistic competition of course if the value of the price aggregator the quality of the marketplace increases the profits of which single seller goes down because you will sell less but the price will be chosen to maximize the denominator essentially and the price rule is something very standard that where you have the demand elasticity and of course the cost and the commission with it are shifted into the final prices depending on the shape of the demand elasticity you can have full pass through incomplete pass through and so on and you can characterize exactly how the price depends on the commission here I show you just to give you an idea the close form solutions for the three examples that they provided before the first is the classic Dixie Stiglitz pricing rule and the pass through is full that's the case of extra power so those functions the other two cases the logic demand and the generalized demand imply incomplete pass through but in general for any surplus function you can characterize the relationship between the price and the commission if you know that you can understand at the previous stage how entry takes place so let me remind you that this aggregator the quality of the marketplace has two components mainly the surplus generated by the products of the marketplace and products with prices the upper bar already chosen at this stage and the surplus generated by all the all the sellers and clearly if you have more sellers you have higher quality the profits of each seller will go down so effectively under endogenous entry and the zero profit conditions what you will pin down is really the value of the aggregator which as you see here will depend on the commission but this is the key aspect that actually applies also in the other two papers and here is just extended it does not depend on the surplus generated by the marketplace with its own products whatever the marketplace done with does with its own products there will be an adjustment in the number of sellers entering in the marketplace but the overall value of the aggregator will not change this will actually depend only on the commission rate now with this micro foundation this neutrality which applies in aggregative games and I mean emerges and also in other application this neutrality is extended to consumer welfare because the utility is a function of the aggregator and so this implies that the pricing of the marketplace of its products does not affect in the smallest the consumer welfare percent the only channel through which a consumer welfare is affected is through the commission and as you see that the commission is actually is actually negatively related to the value of the aggregator of course if you increase the commission the profits of each seller goes down there will be space for less entry and the quality of the marketplace will go down and the other way around so this simplifies a lot the analysis in this class of the situations at the bottom you also find the endogenous number of sellers which of course you can derive it from the above conditions but if we know this we know the pricing we know the free entry we can go to the marketplace and try to see what happens there so here I will write the profit of the marketplace which is what we want to maximize choosing the prices and the commission so the first term is the commission revenues the commission rate multiplied by the number of sellers which have just endogenized and the revenues of each seller and the second term is of course the profits of the marketplace from its own products and the prices per bar is what we need to choose now if we use the free entry condition we can rewrite this this formula let me just say that the first term that you find in the profit function will be the the profit of a pure marketplace one that doesn't provide any of its own goods it's just relying on third party sellers so this is just commission revenues of a pure marketplace and the last term instead is the additional profits that the every marketplace gets from its own particular products where the key fact of this delta I called it an index of differential profits because it represents the difference between the profitability on its own products and the lost commission revenues on third party sales that are eliminated when you sell some of your products so this difference is what matters it determines which products you want to introduce and at which prices and now I will exactly choose the prices of the marketplace to maximize this this this aspect and what we remain we remain that these differential profits will be decreasing in the commission because if the commission is higher the incremental profit that you make on your own provision of products will be lower so what will be the prices chosen by the marketplace maximizing this animal you will get the price rule as the one you see in general here you will have a different demand elasticity for the products of the marketplace but even if it was the same if they were providing the same products basically marketplace and sellers here there are two effects going on first the marketplace is not paying the commission so this pushes towards a lower price choice second the marketplace has a lower opportunity cost of increasing its prices because the demand that is lost that goes to third party sellers where there is a monetization through the commission and this pushes for higher prices so in general we don't know whether the marketplace I mean depends on the particular micro foundation we set higher prices or lower prices this is an example with constant demand elasticity and actually everything depends on whether the marketplace faces a more elastic or less domestic demand if they face the same demand elasticity that will set the same prices in the paper there is a discussion of conditions under which the marketplace will set an higher or lower prices and depends on the shape of the two mentioned elasticities but I will not go deep in this because it can be interesting from a descriptive point of view but the neutrality that I mentioned before implies that the actual prices will not affect directly consumer welfare all we care about is to go back to the profit of the marketplace and find that the commission rates that maximizes them and then we need to understand whether a hybrid marketplace or one providing more of its own products will increase or decrease the commission rate and so this is the same expression of the profits that they have shown before the point is that we know how the price depends on the commission from the pricing of the sellers we know how the price aggregator depends on the commission from the free entry we also know how the delta depends on the on the commission so we just have to maximize this anymore in general there will be a first-day condition that I will not show you because it's a little big but the point is that if you are a pure market base you maximize only the first part because that's its profit and if you're in a hybrid marketplace you have to consider also the additional profits from your products depending on the products you are providing and here however both the numerator and the denominator the delta and the price aggregator as I have suggested before are decreasing the commission rate so in general we cannot really say whether you want to increase or reduce the commission you need to to look case by case or find some particular rule or example to understand better here I will show you given the limited time I will show you the case of the case the example that I have mentioned more times is the case with the elastic constant demand elasticity is the biggest case if you want and if you consider this case this is the commission rate that you will find for a pure marketplace it's an implicit condition but the right hand side is decreasing so pretty much you have always an interior solution within zero one and this commission is decreasing in the demand elasticity of the sellers as you should expect and if you are a hybrid platform so you provide your own additional products and here I assume that you have a possibly different demand elasticity epsilon bar the new optimal commission for the hybrid platform is given by the formula you find here now if the two elasticities are identical you get the same commission rate so neutrality as in Xenia as a matter of fact in a different environment however in this case if the marketplace is facing a more rigid demand for its own goods and this could be the case for instance because customers amazon has a in this example has a bigger reputation for shipping services or post-sale services if this is facing a more rigid demand it will reduce the commission compared to the party sellers and this will increase consumer welfare the other way around in the opposite case so what is the intuition that if the marketplace is facing for its own provision of goods and more rigid demand it will exploit this setting higher markup but they will compensate customers by reducing the commission rates on third-party products which will reduce their prices but will recover entry of some of these sellers and will promote attract more purchases by consumers and this ultimately will recover more commission revenues on the other side so this is an example that clearly shows that things can go in one or the other direction here you get the general rules for the platform so here this formula depends on other elasticity the demand the surplus elasticity the past through elasticities and this elasticity can be different between sellers and the marketplace so as you can imagine the comparison becomes a little more difficult and in the paper there is a discussion of that here I would like to mention first of all that I confirm of course the results by Anderson and better the following in the sense that in the case of a logic demand system which in my notation can be easily found because the surplus and the demand elasticity identical the second formula simplifies and you get the result by Simon and Oslem and that the hybrid platform always set higher commissions than a pure platform however in other cases you may have the opposite for instance in the third example that I mentioned with the generalized linear demand you can actually get the the opposite so the ultimate point is that depending on the preference if you want or the structure of the demand system a hybrid commission could have an incentive to increase the commission rate on third sellers to shift the mandate over itself or to decrease them to actually attract recover entry of some sellers and expand and expand the commission revenues through that channels and in the paper don't worry I will not go into details but in the paper I have a more general versions of the model one I would like to mention since I see I have still a couple of minutes and this is one where you see the micro foundation is not the log linear before is a little more general there is a parameter k that represents the elasticity of the demand of consumers with respect to the quality of the marketplace and here if this goes to zero you go back to the log linear model but when this parameter becomes bigger I still have that formula but I will not go into details but the point is that when the demand of consumers is highly elastic to the quality of the marketplace it is more likely that the hybrid commission will reduce the commissions because it has an incentive to attract the model chases by consumers so this is an additional version so I mentioned already the extension that you will find in the paper if there are questions we can talk further about them one is actually analyzed in a different paper and it's about competition between platforms platforms that are allowed also to have a tool that here I didn't consider which is a subscription fees so amazon prime is a typical example consumers pay to access a superior service the results that I idea are pretty much translated in that environment actually in a world where platforms are the vice funded or funded with also payments by consumers there is an higher incentive for the platforms to internalize the interest consumers in their choices and it is analyzed in another paper and so I will conclude by reminding that so what in this exploration I could find is that for a large class of demand systems with this micro foundation I can confirm the fact that the forgiven commission the commission does not change much the same fact that the market base introduce its own products does not affect the consumer welfare because there is an adjustment through the number of sellers that leaves overall the gains from variety because that's what we are talking about in the monopolistic competition unchanged this is a neutrality result that was found in models with strategic interactions and free entry and applies in this under this micro foundation and however welfare mechanism work through changes in the commissions but in theory there is not a systematic tendency to increase the commission rate because there are there is a mechanism that I have emphasized that the works in the opposite direction it could be compensating the first one or also dominating the first one ultimately this I think is an empirical issues are commission rates increased where market basis this case Amazon expands its activity expands its share of the market or they are decreased or there is nothing going on so they whether there is a systematic correlation across product categories so it's not about one particular product category that but across product categories that there should be the relevant question from an empirical point of view and with this I exhaust my time so I will give back the floor to the moderator Osrem. Thank you very much Federico it's very well on time so we have Yusuke Zenyo as a discussant five minutes for your discussion Yusuke and then I have already collected a couple of questions for Q&A. Thank you Osrem and the thank you organizers for having me I am Yusuke Zenyo from Kobe University Japan. I'm very pleased to be here as a discussant for Federico's very excellent paper. The paper provides a general model for consumer preference that enables to assess the welfare impact of the first party entry by platinum owner on consumer surplus because time is limited so let me ask my question I have two questions today two questions I think the first one is related to the platform's private incentives for the choice of price elasticity for its first party goods. An important message from Federico's study is that price elasticity matters when it comes to the assessment of first party selling by platinum. More specifically speaking if first party products have less elastic demand than third party goods, first party entry is more likely to improve consumer surplus. This is a nice result for conventional policy however at the same time in the present model most elasticity for first party goods and third party goods are considered as fixed exogenous parameters so epsilon and epsilon bar are considered as parameters. So I am wondering what happens if the hybrid platform is allowed to endogenously decide the level of price elasticity for first party goods. If the platform has incentives to sell first party goods with less elastic demand than third party goods its platform's private incentive is desirable in terms of consumer surplus. However if the hybrid platform has the opposite preference such first party entry by platform owner is expected to reduce consumer surplus. If this latter outcome prevails in equilibrium there is a need for some sort of regulations about first party entry by platform owner. So my question is the endogenous decision making by platinum on regarding price elasticity for first party goods. This is my first question and next let me move on to my second question. My second question is about the the seller side of two-sided platform. In the present model Fidelica assumes that all third party sellers are fringe so very small sellers so any changes in their pricing do not affect the aggregate and do not affect consumer surplus directly and because of the free entry assumption in equilibrium the seller surplus is always equal to zero in the present model setting. So I wonder if you can consider the presence of big sellers in addition to fringe small sellers. So for example the simplest way to do to address this issue is to add one big seller or killer seller into the current model setting and one can consider that that big seller is making pricing decision at the same time when the hybrid platform does subsequently and after observing those pricing decisions made by the big seller and the hybrid set hybrid platform fringe seller are going to make entry decision and pricing decision as in the current model setup. If this extension is possible mathematically and analytically we can make further discussions about the effect of first party entry on the seller side not only on consumer side. I think this second extension will make your policy implications much stronger than the present one.