 Welcome everyone. I would like to thank you already for the organization team for inviting me and also for setting up such a great moderation team which is very, very related to my research. So it's my pleasure to present to you some research that I do on the recommendation behavior of Amazon which I do together with Matthias and Tagium. So please let me know about any comments or feedback that you might have so you can send me an email or you can also just write in the chat. Okay, so what is the motivation for my talk? So we depart from that online marketplaces host a large number of suppliers and products and they facilitate consumers to find what they are looking for by giving recommendations. And when giving these recommendations it's natural that platforms also consider their own interest and however this raises some discussion. So especially in the context of hybrid sales platforms which are platforms which operate a marketplace but at the same time act as a retailer on these platforms on the same marketplace have become into an intensive discussion. So specifically there is the fear that the interest of consumers in hybrid platforms might not be well aligned and this fear is also reflected in current legislative initiatives such as the Digital Market Act or the American Innovation and Choice Online Act. And part of these initiatives is to prevent very large platforms from self-referencing their products at the expense of competitors. Of course as a very large platform which is hybrid Amazon is a case in point and there's also a growing literature studying potential self-referencing behavior with respect to how Amazon might favor its own sales department by the so-called buy box. So I think there will be also some people present here which study this question but in our paper what we want to do is we want to study a more subtle form of self-referencing and as I will explain to you this will consist in first not providing any supply recommendation for a specific product and then in the second step if then using distorted product recommendations just to consumers elsewhere. But what we have in mind is maybe better explained with a series of screenshots how Amazon works. So on Amazon once you are interested in a specific product so you end up at some product page. So here for instance we have a very popular card game called UNO and this product page then gives you a lot of information about the product characteristics and also user feedback in the form of ratings and stars. So what you see here. Furthermore as for this product here often there are plenty of suppliers available so for instance as you can see right here it's not not that big you can see that this product is offered by 39 distinct suppliers. Of course most consumers do not have the time or also not have the interest to look at all of them and therefore Amazon pre-selects one of them on the right hand side of the page. So this is here and this is what we call the buy box. So for instance in this buy box what you can see is there is one seller promoted or featured so which is called in prime time and this is also a seller which uses Amazon to do the fulfillment. And once Amazon features a supplier at this buy box it's very likely that a consumer interested in this product will choose to buy from it and there are different studies which claim that over 80 percent of transactions are done with suppliers that are inside this buy box and this should highlight to you the importance of this design element to see a consumer. Without surprise it's many suppliers as well as researchers have been wondering about how to get into that buy box because if you are one of the 39 you might imagine okay they have a high interest to be there given that all the most transactions are done over this design element. And a second question which is more related to the policy perspective is if this buy box is somehow biased in favor of Amazon's own state department and this is also what others for instance my discussion studies. In our paper we would like to study we would like to take a different route because we think there are many more ways to stick consumers towards other products and this is what I will explain in a minute. So what's remarkable is that in a significant number of cases Amazon does not feature any seller. So this can be seen here in this example on this slide so here we have a very similar product and it's just a different variety of the same card game you know that I have shown you before and what you can still see you have the consumer feedback you have the stars you have the rating and you can also see that this product is available from 21 different suppliers. However on the right hand side you don't see any buy box so Amazon decided not to feature any of these suppliers and hence something is different about this page design. So in case you still want to buy this product what you would need to do is you would need to click on see all buying options and then you could compare the 21 different offers available. It's debatable if this makes the purchase of the product more complicated but at the same time most people would agree that this deviation from the default design has the potential to create some confusion among consumers and this confusion may even lead to a situation where some consumer thinks something is wrong so maybe the quality of the supplier is not good or maybe there's another problem associated with this with this product and this might be dead to meant to fall the sales. Of course one might wonder why would Amazon do this because Amazon so if this is detrimental sales why would Amazon make it more difficult or make it create a situation where consumers are confused. So what Amazon claims is that they do that in the interest of consumers so specifically in particular they say they may remove the buy box when it be when they believe to observe harmful pricing practices and harmful pricing practices could be that the price of a product is higher than prices on or of Amazon. Of course this is of course a nice behavior so if we say Amazon is very interested in getting a lot of consumers then this could be a good strategy but we still believe that this is not straightforward and we believe this behavior is interesting for two reasons. So one thing is that as Amazon is a hybrid sales platform one might wonder if Amazon applies the same standards regarding harmful practices especially in the cases when they also supply the product. So the question would be is there a situation where Amazon thinks the price is too high when Amazon is also among the sellers if not this could hint an unequal treatment which would be in favor of its own sales department. The second aspect why we think this is interesting is that if consumers are indeed confused by the absence of the buy box as I said before it could have the potential to reduce the product sales and so even if this policy is only applied to third party suppliers Amazon potentially foregoes commission payment and if consumers turned to buy somewhere else then Amazon would even make losses in the short run. However the letter point may only matter if Amazon then really loses these consumers. In practice Amazon might achieve to increase the sales of other products by redirecting consumers to them and specifically consumers who search for a given product but do not see a buy box might be more attentive to other recommendations of Amazon for alternative products. And one might wonder then if this is the case that consumers get more attentive to recommendations for something else if Amazon is then using this channeled as a means to favor products where it's also selling itself. So these are essentially the aspects that we want to study in our paper and for this we have we do conduct an empirical analysis of web script data from Amazon in the US over a period of four months and then we proceed in two steps. So the first step we study we investigate the buy box oppression itself. So specifically what we are interested in is to analyze under which circumstances Amazon really removes the buy box. So for instance if this is really related to two high prices and also if these factors are different for Amazon's own office. Furthermore we want to quantify if the removal of the buy box has an effect on sales because then this would also give a price to the suspicion that you effectively steal consumers away from the stock of product. So this is what I will focus mostly in my talk today. In the second part of the paper what we want to do is to analyze potential follow on effects of the buy box removal. In particular we would like to know if consumers once the buy box is not shown are more attentive to alternative product recommendations and if this materializes also in more sales of other products. And for this exercise we will focus on one type of organic product recommendations which is called compare with similar items that I will explain later. And we will for this design element we will also study if we confirm that consumers are more attentive to these recommendations. If Amazon is leveraging this situation to steal consumers more towards products for which it is a supplier. Okay so for those including Jacques who will not have much time I do a preview of the results. First regarding the buy box oppression so we as I said we study frequency determinants and effects of buy box suppression and what we find is that Amazon always shows the buy box without telling itself a product which suggests or which could hint already an unequal treatment. However if Amazon is not among the suppliers we observe in our sample that it removes the buy box in 39% of the cases of our sample particularly if it considers prices to be too high. Now that we know that or that we found out that Amazon is not that Amazon is never removing a buy box in our sample when it's selling itself we try to apply the same policy for products when Amazon is among the suppliers and if we do this kind of simulation we find that Amazon would not be able to show a buy box in 13% of cases. Third what we also find is that indeed the buy box oppression has an effect on pay up. So we work with different approximations that we may discuss later and what we find is that if a product does not show a buy box then this reduces sales by 25 to 30%. Okay this is for the first part of the paper. On the second part of the paper whether there are follow-on effects essentially what we find is that products recommendations on pages where there's no buy box are associated with higher sales. So whenever you end up on a page where there's no buy box then apparently consumers are more attentive to recommendations to other products and this materializes in higher sales and the effectiveness is almost twice as high and regarding the question whether Amazon leverages this situation what we see is that Amazon tends to steer consumers more towards product itself if the buy box of the focus product and this is an indication of a very subtle step preference thing. Okay do you have any questions so far? If not then I would just move on. Sorry have you tried to look at the FBA versus FBN? So we will discuss this as well but so maybe we can come back to that when we in the analysis. Okay so our paper relates broadly to three streams of literature. So first there's a large theoretical literature in an industrial organization highlighting that intermediaries may have incentives to distort recommendations and it's clear that Amazon is an intermediary and of course because of the commission income may also distort recommendations towards something that consumers would not be their first choice. Second closely related there's an increasing theoretical literature about distortions that may emerge from hybrid sales platforms so hybrid sales firms that act both as a marketplace but also as a retailer and this literature shows that businesses which might have these businesses might have incentives to distort recommendation in favor of their own product and services. However the impact on welfare is not so clear because the degree of distortion might be less severe than in the case of a separated business. So naturally Amazon is a hybrid sales platform and we contribute to this literature by showing empirical evidence about this kind of steering behavior towards own products. Third on the empirical side there's an empirical large literature in marketing which highlights that in the online context the presentation of a product so whether it's made very prominent or whether it receives a very good ranking may drive higher sales and specifically for Amazon there's a growing literature also from marketing and as not only economics which mostly studies with how Amazon enters product markets and how Amazon makes the decision to who to feature in the buy box and as well as how other product recommendations may be used for that preference and we believe that we can contribute to this literature by providing like a new discussion on first the unequal treatment that Amazon may apply when deciding whether to show a buy box or not and second a more subtle form of steering which is by the combination of the absence of a supplier recommendation in order to make to make a product recommendation more visible to consumers. Okay I will describe now the data that we use in our analysis so what we do is we use web script data from the US version of Amazon for a period of about four months and our initial sample consists of the 50 best-selling products in 50 categories on Amazon. As we are somehow also interested in competition with other platforms specifically with prices on other platforms we have further expanded this sample to include the same set of products which are very popular on the biggest competitor in the US which is Walmart.com and by including the very popular products on Walmart.com we end up with a sample of almost 4,000 products and for these products what we have done is we have collected daily data on product characteristics on the decision to show a buy box and on the availability of distinct sellers and the prices that they set and we have done essentially the same collection also for the platform Walmart although we use mainly here the price level on Walmart for the same products which are available on Amazon. Finally we have also amended our data set with third-party suppliers like the Keeper and others. Okay for our analysis we don't exploit the full detail of the data that we have so in principle we have more detail but to simplify a little bit the exposition we use data a data set where one observation is one product on a specific date and by this we have then around we have more than 270,000 observations where yeah over the period of four months for these around 4,000 products. There are distinct things that I would like to highlight about our sample so in a bit more than half of the cases we see that Amazon is acting as a supplier so you see there are even among the top selling products there are a lot of products which where Amazon is not active. However it's relatively rare that a product is only sold by Amazon so for 92% of products in our sample there's at least one third-party seller so this in turn means that only in 8% of the cases only Amazon is available. For these third parties which are in the sample this can be FBA sellers so sellers which use the facilities of Amazon to do the fulfillment or they or sellers which are called FBM which use their own logistics to do the fulfillment. Okay most importantly for our study the bybox is visible in only 83% of cases of course this might at first sight look look like a very high number but what I will show you the following that if you condition the data set in a relevant way it's very frequent that the bybox is not visible. Okay finally anybody might be wondering indeed in 93% of the cases when bybox is shown and Amazon is available as a seller in our sample Amazon is getting the bybox in 93% of the cases. Okay for our analysis we also disposed of a lot of other controls like the sales rank which is a good proxy for sales for those which don't have data from Amazon itself or from a large merchant. We also have information about the number of sellers offering the product which can range to up to 280 different offers and then we have other pricing information so we have the manufacturer recommended sales price and also we have information whether the product is available at Walmart or eBay and we have also information about the respective prices. Okay in this slide I want to contract contrast a little bit first the differences of product day observations where the bybox is shown and where it's not so on the left hand side you see statistics for the product day observations when the bybox is visible and here on the right hand side on the middle columns you see the same information when the bybox is invisible. So one thing that we see first is that when the bybox is missing so these columns this never coincides with Amazon selling the product itself and if we take this information into account that the bybox is only missing or can only be missing if Amazon is not selling the product then we can calculate that the bybox is not visible in 39% of these cases where Amazon is not among the sellers. Second aspect that I want to highlight is the suppression of the bybox does not relate too much to the presence of suppliers which use the fulfillment by Amazon. So if we look at the relative frequency of an FBA seller being available FBA the sellers being those which use Amazon facilities so we see that the bybox is visible when the bybox is visible then in 48% of the cases these high quality sellers are available while it's only 44% when the bybox is in physical. So there is a difference but the difference is not relatively large or at least it's not that large that would as it could explain that the bybox is not shown. What is a bit surprising is even that the number of offers when the bybox is not shown is substantially larger so while we have here when the bybox is visible around eight offers on average this number increases to four teams so we have 14 different offers available when the bybox is invisible in our sample and this makes it even more questionable why Amazon thinks none of them should be promoted because among the 14 especially if we also see that the frequency or the presence of FBA sellers is not that low one could wonder why there's no one promoted. Finally what we can also see what we can also deduct a little bit also from the descriptives although it with these type of variables not always that clear the negative clock sales rank which could be a measure of our sales we can see that products which don't have a bybox apparently are less sold or are less often sold as reflected by this higher sales rank or by this more negative sales rank. Of course this could be related of because the product is not popular but it could also be an effect and it could be an effect of the bybox removal but this is of course something that we need to study econometrically. Okay what's more important probably is how the suppression of the bybox relates to prices and for this I have prepared this nice chart so what we see here is the fraction of pages or product day observations without bybox in relation to the price difference of the minimum amazon price compared to different bench numbers. So for instance we compare the minimum price on amazon to the price on walmart on ebay or to this manufacturer recommended sales price and what we can already see here is that there is an increasing pattern so the more expensive products are on amazon the more likely this bybox is removed and specifically for the case of walmart what we can see is if the amazon minimum price is 10 percent higher than the price on walmart we already see that in 80 percent of the cases there's no bybox and this should highlight that apparently amazon cares a lot about the relative price difference to other competing marketplace. Okay as these descriptive tables cannot account or this chart cannot account for the identity between products and sellers and their availability we also confirm these initial observations econometrically so specifically we estimate linear regression model to explain the decision of amazon to show the bybox or not. So the dependent variable is an indicator taking value one if the bybox is shown and the external variables briefly speaking are product market characteristics such that the availability of certain supplier types elect measure for demand and the minimums nymph price so the more we control for prices on competing products and sales channels and as well as for product and date fixed effect. Okay and when we do this then our result that we got from the graph are largely confirmed but let's maybe go one by one so first of all what we see is that the presence of a high quality data has some effect on the presence of the bybox so here depending on the specification we see that the likelihood that the bybox is shown increases by 7 to 9 percentage points and surprisingly we see that there is a negative correlation between the bybox presence and the number of sellers but yeah these effects are not that large what may be more important is the effects of the price here we see strong effects so for instance the higher the minimum price on amazon the less likely a bybox is shown um we can then also see how the comparison to some benchmarks also affects the likelihood of a bybox being shown for instance what we see here is that if the price on amazon is higher than the manufacturer recommended sales price this also decreases the likelihood of a bybox being shown the same goes for a comparison of the product price with an average of other products on amazon and finally what we see and where we also see the largest effect if a price if the price of a product on amazon is higher than the price of the same product on walmart or ebay then we see that the that the likelihood of the bybox being shown is reduced by 22 percentage points okay so it seems that that the competition between the marketplaces has the largest impact on the bybox operation compared to within platform orders okay there are no questions i would should we go on so now that we have seen before that the bybox is always shown when amazon is among the suppliers one might wonder whether this stems from that fact that amazon is always acting as a good supplier so amazon using its own fulfillment could be considered as high quality supplier and also amazon as a large platform or as a large seller could also offer very competitive prices so which in turn could justify that we observe in our sample that the bybox would never be removed however to see whether this is the case we want to do an exercise where we essentially proceed in the following steps so we first take the observations where amazon is not a seller to calibrate a model which predicts if the bybox should be shown or not and then we take this model in the second step to the data where amazon is among the sellers to do an auto sample prediction if a bybox should be shown or not and if we essentially see that the bybox could always be shown by this prediction then we know okay amazon or this could be an indication for an equal treatment of amazon's own sales department and third party suppliers but if we don't see that then this could be an indication of unequal treatment okay so step by step so if we so here in this table what we have is we have on the left column one to three we have the observations where amazon is acting as it's not acting as a seller and as i said before we see that the bybox is only visible in around 64 percent of the cases we use different models like ols or lotted in order which are very similar to the model that i have shown you before and we find also that the predictive power of these models are relatively good so we correctly predict whether there should be a bybox or not with 85 to 87 percent of the cases on column four and five we then have the data where amazon is is a seller where in the really existing data we have observed that amazon is always showing a bybox when it's among the sellers if we use then these different models from the data where amazon is not selling to predict in the cases where amazon is selling whether there should be a bybox then indeed we see there should be cases where amazon should not recommend themselves um specifically what we see so in the different models is that the bybox should not be shown in 13 to 14 percent of the cases why this is the case so on the one hand we also observe that amazon is sometimes not the cheapest phase channel available on the internet so even if um so for instance walmart or ebay might be cheaper and then even then amazon close the bybox possibly promoting its own offer and second what we have also done is we treated amazon as a seller that is as good as a fulfillment by fulfilled by amazon like an fba seller which also decreased this number however if we would treat amazon as a fbm seller for instance then this these numbers would be much closer to these numbers here okay i see in the chat there is a question okay i don't know that was okay um as a final step of this analysis of the bybox suppression we can wonder if consumers do not see a bybox whether this affects the likelihood of a purchase and unfortunately of course i don't have state data available as most researchers which do research about amazon and therefore i have to substitute this data by other data that is available and uh like in other studies i use the sales rank which is published regularly by amazon um which is information for each category of network which which are the most frequently sold product and the good thing about this variable is that it negatively correlates with the sales and um what we do is then we adopt a model similar as rimers and bald fogel last year an a or a study about book sale and estimate a conditional adjustment model which is uh we explained the minus log negative might also explain the negative log sales rank by the lack negative log sales rank an indicator of taking value one when the bybox is not available and then essentially controlling for the same factors as before so product market characteristics um product fixed effects and then combinations of the category and the day fixed effects okay if we do this uh what we can observe is that indeed when the bybox is invisible this decreases the the sales so um how good do we get to this interpretation so as i said here the dependent variable is the next sales rank so first what we need to do is we need to do an adjustment for the conditional adjustment model and if we do this then we would find we would be left with a message that the bybox depression worsens the sales rank by 50 to 60 percent um as this is very difficult to interpret what this means 50 to 60 percent of the sales rank uh we then in the second step apply a conversion uh which is based on previous studies which had access to to amazon data or amazon sales data which approximate the relationship of a sales rank change to to sales change and what comes out is that essentially a good approximation is that um that the effect has to be has to be divided by two such that we end up with a with a decrease of sales of 25 to 30 percent what's interesting is um that we get to this conclusion even when controlling for the fact that okay that the price might be higher so prices indeed have a uh decreasing effect on sales we also see that if a price is relatively high to the amazon category then this also decreases and specifically we also see that apparently uh there is we can also measure interplatform competition here so if the price on on amazon is higher than the price on walmart or ebay this also decreases thing and even then we still find a sizable effect uh of um 25 to 30 percent um when doing this okay good so i would be done with the first part of of the paper which was about the buy box itself so in the second part of the paper what we aim to study is the question whether there are follow-on effects of this buy box suppression why because as i have explained in the beginning um it is a bit surprising that amazon would make it more difficult or more complicated for a consumer to to buy the product specifically in the market for for third-party products and why is this puzzling because amazon is essentially making its income from commissions then and if a product is not sold uh then this of course decreases decreases this commission income and then could also lead to a decrease in profit however if amazon is able to to steer consumers to other products this might not be this might not be uh too relevant and this is essentially what we want to study in this second part um so our idea here is that when there is no buy box for a specific product i so remember the uno game like the the lila one without buy box then the purchase process is more costly for buyers or at least consumers are are somehow incentivized to think that something is wrong and therefore they for don't want to buy this product um it's possible that consumers will partly continue to search on amazon and then recommendations to other products might then become more relevant and especially what we have in mind is that on each product page so even the product pages without a buy box might have a lot of recommendations to other products and what we would ask now the first step is for the recommended products j which received more product purchases when being recommended on a product i page without a buy box okay and just minder you have finally needs left okay um okay and to see what we have in mind so um there are different recommendations to other products and one that we will focus on is this compare with similar items table and what is good about this one is that it's a non-sponsored form of recommendation so manufacturers cannot buy in order to be promoted there and it's also present for most of the products and it's directly accessible for from the top of the page and what it does it allows an easy comparison of with up to five other similar products and for the products that we have considered where we have were in the first part of the paper interested whether they have a buy box or not we have also collected this information about the comparison page table which is shown on this table and so first thing that we want to establish is whether recommendations of the on these tables have an effect on sales of these other products for this we have estimated a similar sales model then that we did before for the buy box oppression and what we indeed see is the more a product is recommended on other pages within this compare with similar items tables then we see that the sales are higher and the second step we distinguish whether these recommendations come from pages of products which have a buy box and those which don't have a buy box and what we observe is that essentially recommendations from tables which are below are on a page without buy box are more effective so what you can see here is that the effect of the recommendations of of another product page without buy boxes almost twice a time finally we would like to see whether amazon is using this higher visibility of the comparison table to promote more often its own products and our idea would be an independent marketplace operator should not make the composition of these comparison table products dependent on the buy box present or not and while an integrated operator so like amazon may enter the composition as the demand effect is higher without buy box so what you see here is yes does amazon say how it sets the comparison table because this is not through ads so i was wondering does it say anything about how it chooses those products um we have to look into this specifically but if it's not about uh if it's not about ads then usually they would say this is about prices and product quality again no okay um so what this chart to show you is that products that amazon sell are then more often in the comparison table especially when the buy box is not visible so what you see here is the proportion of products which are on this buy amazon which are on this comparison table compared to the proportion of compared to to the presence of these products in the product space and what you why you can see already here that even when the buy box is visible there is some kind of over representation of amazon products in this comparison table this over representation of amazon product is much more severe in the case that there was no buy box especially when the frequency or the relative frequency of amazon products in the product space is relatively low so if you have for instance 20 of amazon products in one category what you can see is that um amazon tends to steer consumers much more to awards its own products when there's no buy box okay we can also compare confirm this with the regression analysis but this one i will skip so let's get to my conclusion so i stay more less in time so the findings of our study are as follows so amazon tends to hide the buy box if it receives prices too high but only when it's not acting as a better second finding that i would like to highlight the buy box suppression decreases demand for a product on the amazon marketplace and amplifies the sales enhancing effect of product recommendations on the same page and third when in such a situation we also observe that amazon tends to steer more towards own products when the buy box is suppressed um what do we take from these findings so we would like to highlight some some aspects about this so first one is that the fact that we only observe that the buy boxes can be suppressed when amazon is not present is an indication of a different treatment of third party offers and is also a sign of intransparency of the of the behavior of amazon second the fact that we observe that purchases are less likely when there's no recommendation highlight that there's a search friction for consumers so consumers which would be interested in buying this good are somehow confused uh by this policy of amazon and this might also be detrimental for consumers third um another risk of this policy from our perspective is that if amazon conditions uh the prominence of products or the prominence of of uh supplier offers uh on prices on competing marketplaces this can provide similar incentives as a price very close so we don't say that amazon is replicating a price very close or something like this but we just want to highlight that this can be the risk that it provides similar incentives to to condition the prominence of products on prices somewhere and finally for the part that I have shown you very briefly we also provide subtle evidence for a subtle form of step referencing so we see that when amazon abstains from recommending a specific seller then this makes consumer more attentive to other product recommendations and amazon effectively uses this also by distorting the recommendations more towards its own product with this I'm gladly done in time so thank you very much for your attention and I'm happy about any of your comments yeah so thanks Urik for your great presentation and so we are happy to have DaVish to discuss this paper and meanwhile if you have any question welcome to type in the chat or you know talk after the discussion um hold on one second let's see all right thank you all for um for this so I really enjoyed this paper I thought it was very clear this is something I've actually wondered a lot about and I think I learned a lot from this paper but what amazon is doing so I want to start with some screenshots so these are screenshots that took about a year ago maybe a year and a half ago this is some kind of tire sealant and you can see in green the buy box is suppressed but if you look in red there are a bunch of sellers and if I go the next slide you know there's a FBA seller for $29.99 which is you know has a lot of ratings and a large percent positive there are other sellers as well so it's sort of a a mystery of why is it that amazon is not allowing you to just buy the product in the buy box so I also looked at the comparison table and what you find here is that all the comparisons are to different sizes of this product sold by amazon itself um so everything's exactly in line with Ulrich's paper um and so finally yesterday I looked back since the year and a half later and now there is a buy box but it is amazon selling the product and for maybe about a 10 higher price the third parties are selling a year and a half so this seems to be a pretty ubiquitous thing so the question is what's the what's the mechanism what's driving it um so I'm going to give you two different views and I think my my advice to Ulrich is to maybe talk a little bit more about the second view the first view is that amazon wants to prevent price gouging now Jeff Bezos has said well he doesn't care about price gouging but consumers care a lot about price gouging in congress now there are multiple bills about price gouging as you might imagine amazon doesn't want to get the appearance that its marketplace is price gouging so there are a couple of things it could do one is it could sell the product itself where it can set the price itself um it could potentially match other marketplaces the second is that it can reduce the sales of products that are being price gouged either by suppressing the buy box or by steering away through recommendations and Ulrich has shown evidence of both of these um now when I can make a couple comments here first of all there's a as Ulrich said a subtle form of self-preferencing only amazon is allowed to price goug and it's very clear here you know if you know anything about amazon's pricing amazon does price match competitors and as Ulrich showed showed us they are monitoring ebay and walmart but they seem to have made the decision not to price match ebay ebay and walmart marketplace which is why their product is often priced higher but they're punishing third parties that price higher so you could imagine a strategy of amazon to price match ebay and walmart and then this idea of self-preferencing would in some sense go away because amazon would have the same prices and there's another view of this um which is that this is about sort of a moose favorite nation clause to the back door um so the dc government recently found a lawsuit which i'd recommend the authors to look at um calling this an antitrust violation and if you look at the history of this amazon used to have price parity provisions which basically said you can only sell on our platform if you're not selling below on competing platforms so if you if you're a seller you sell on less on walmart or ebay you're not allowed to sell on amazon that was eliminated first in europe and then in the us because of regulator concerns antitrust concerns and instead they have what they call a fair pricing policy um but part of that is that they suppress the buy box if it's a higher pricing competitor websites it's not clear if this is different than the price parity provision or not but this could be a maybe a more broad way to do it where there's not an explicit agreement on the price to the third party sells but you can get the same thing um so what i was wondering is it is a dc in the dc lawsuit case they say they basically say their fair pricing policy is just the same as the mfn contracts as before this is the anti-competitive and a violation of the antitrust laws um so there's a question of is dc right or not and then what's going on with consumer welfare so if you think about the first mechanism of price gouging you would think that suppressing the buy box might lower prices because it's going to force the third parties to lower their price so they're going to lose a huge amount of sales but if this is about third parties selling for cheaper on ebay or walmart this might raise prices because you know they they get the message and they raise their prices of other platforms um so i think that the main thing that i would like from this paper is to talk a little bit more about who is selling on ebay and walmart is it the same merchants as amazon which would be sort of like a pp like the price provision thing or is it others so this is for example walmart is selling for less walmart the retailer is selling for less than amazon it looks a lot like the price gouging whereas if this is the third party amazon is also selling on walmart marketplace it looks a lot more like the second view um there's a lot of regulatory interest on the second view and i think um if i were the authors i'd play that up more but the more you can get at that you know so if you could have a statistic of when amazon suppresses the buy box x percentage of the time third parties on the amazon platform are selling for cheaper on walmart ebay um i think that would really inform a lot of the interest debate um so with that i have to thank uh i encourage all you to read this paper and i it uh it is very nice and clear so thank you