 My name is Fen Zhu and I'm here to be the timekeeper. And so our presenter today is Uri Leitenberg from Telecom Paris. And he will speak for the first 45 minutes. And then we will have our discussant, Mark Ivaldi from Turu School of Economics. And Mark will spend like five minutes discussing the paper. And then we will have 10 minutes for Q&A. So for the first 45 minutes, let's give our presenter as much time as possible. But if you have any clarification questions, feel free to still raise your hand or chat in. So Uri, go ahead. Okay, so thank you very much for the introduction thing and also thank you very much for anybody else for joining. So my name is Uri Leitenberger and it's my pleasure to present to you ongoing research, which is entitled Vertical Integration of Platforms and Product Remnants, which is co-authored by my French and German colleagues, Morgan, Matthias, Reinhard and Thomas. So online platforms are an essential part of e-commerce and a core function of them is to provide recommendations to products. At the same time, they also provide very often recommendations to suppliers offering these products. For instance, if you go to the Amazon marketplace, Amazon does not recommend you only which product to buy, but also gives you some kind of advice, which seller you should conduct the purchase. So at the same time, it's very common that online businesses have ownership links between each other. So online ownership links can exist, for instance, between the marketplace operator and the suppliers, and such ownership links are a major concern in competition policy, as has been seen, for example, in the Google Shopping case. So against this background, we want to contribute to this debate by starting the link between meta-search platforms and vertical integration in the hotel booking industry. So the starting point of my talk is the observation that in 2013, Prysline acquired the meta-search platform Kayak for 1.8 billion US dollars. And Prysline, which is nowadays the booking holding, consisted already back then of important online travel agents like booking.com or Aguda. So observers raised concerns regarding the utility of the ranking algorithm of Kayak after the integration, as Kayak may have the incentive to promote booking holding online travel agents rather than cheapest or equally popular ones of competitors. So against these concerns, the CEO of the booking holding replied that Kayak will of course not bias the search results, and this is what we want to study in our paper. So to get a better idea of the concerns that we have in mind, let's maybe have a look at the website of Kayak, how it was some time ago. And let's assume that we are interested in finding a hotel room next week in Berlin for two persons, and we just rely on the recommended search results. So the recommended search results is what you see here, so it's sorted by recommendation of Kayak. And we do that because apparently still the vast majority of people which use these comparison platforms rely on the recommendation. So when you search Kayak then delivers us with a list of hotels and presents us some sales channels for these hotels. We have here the hotels. In our example, this is Citystay Hostel Berlin and the Redison Blue Hotel in Berlin. And then you see here the different sales channels. So for the first total Kayak shows most prominently here, Aguda, which is a platform and online travel agent belonging to the booking holding. And also Kayak shows less prominently some other sales channels for this hotel like Expedia, hotels.com and ebookers. And then the other sales channels are hidden in some kind of drop down menu. And even that Aguda here is offering the lowest price among others. One could say that this horizontal ranking or so this positioning of sales channels is in the best interest of consumers. So if we look at the presentation of the second hotel, the Redison Blue Hotel, then things are slightly different. So Kayak shows here the OTA booking.com most prominently with a price of 107 euros while Expedia, which has a lower price, is shown less prominently. So it's still visible, but it's not as prominent as booking. However, this should not be the only thing that you should put your attention on. So in our stylus example, what we can also see that apparently Kayak shows a hotel on the first position where it has the best price on a booking holding platform, Aguda. While it shows the hotel with a lower price on an unaffiliated platform, which is Expedia, it shows later. So by this you can see that Kayak has potentially two means to steer consumers towards their own affiliated platforms. So this would be the position of the sales channel for a specific hotel. So what we call here the horizontal ranking, but also the position of the hotel itself in the vertical ranking. So here, so if you go down the list of hotels, then you see that hotels which are cheap on competing sales channels might be put later and motivated by this. We address in our article the question whether integrated meta search platforms favor their own intermediaries. So our application, as explained, is the online booking industry where the meta search platform Kayak belongs to the same company group as various other online travel agents such as booking.com. In our empirical analysis, we use web script data from Kayak for Paris. This gives us information on the ranking and prices of sales channels of a given hotel, and we further collect data on hotels and sales channel popularity as they are also important for the determination of the ranking. So in our empirical investigation, we conduct then regression analysis to explain how Kayak decides about the horizontal and vertical ranking. So this is we study the probability of sales channel to be visible or prominent, and also second the position of a hotel in the search results, depending on prices. And our findings indicate that the sales channels of booking.com are more visible on Kayak and that hotels with cheaper prices on Expedia OTAs are shown indeed less prominently. Does anybody have a question at this point in time. I was wondering actually, whether this would be these would be exactly the same deal so when you go back to that page. And so when when I looked at some point myself, when I booked a hotel, I discovered that it was not the exact same thing. Right. So, you know, for instance, the room type was different, or there was breakfast or no breakfast or the cancellation policy was a slightly different one. So, could that explain why the 107 is the recommended one while the 102 has the lowest price. So, basically asking is it maybe a little bit more multi dimensional than just, you know, about the price. Well, in fact, there could be many factors affecting it. So it could be that booking is the most popular channel in Europe. It could also be that what you say that there could be some deviation in the in the sales conditions. However, normally when kayak sees that so kayak also collects the information about the availability of of these commodities like good cancellation policy. And usually this is also then taking into account when highlighting the third results. So, I cannot tell for this example because this is only a stylus example, but to to some to some extent it might also there might also be different perceptions of consumers regarding what they might have when they go to a specific stage. And what's peculiar about the example is that it shows to discounted price. So, they don't show discount in the other for Expedia or for. So, you mean here in the top result you mean or Yeah. So they could use discounted as a selection criteria also. I mean, the assumption is a little bit that people go to this meta search websites also to to find to find the best price. So that rather you have on these pages like people that are very price sensitive and therefore the recommendation should be more towards providing recommendation of states terms with a low price, but we will come back to this a bit later. So I will reach an one last question. I don't know if it's another question here you've stressed the discrimination given the hotels you are choosing but it could also be that they discriminate by only choosing the to show you the hotels in which people from the booking sites from the booking group have a best deal so even if Tobias is right and they show you a booking.com because it's better than Expedia. One other way in which you could discriminate is by only showing or giving prominent place to the hotels where the booking group is better. So exactly. So that's indeed what we that what we study then in the vertical ranking. Okay, okay, sir. Okay, so I will briefly talk about the literature that we contribute to. So the first friend of literature which is the well developed theoretical literature related to our paper, which shows that intermediaries can have incentives to bias recommendation. And the paper closest to us here is the concierge and Taylor who study the incentives of integrated search engines to bias search results in favor of their affiliates. Then, second another strand of literature that we contribute to is there is mostly in marketing and this literature is empirical and highlighting the importance of rankings for the booking choices of consumers in the market. So specifically, it is shown that purchase decisions depend on very much on whether hotels are visible in the search result or not, which gives rise to why these recommendations matter in this market anyway. The third part of literature that we contribute to is the growing literature studying price parity clauses, which is both theoretically and empirically. So this price parity clauses have been employed by many online travel agents such as booking.com and Expedia have been forbidden in many European countries. And the literature shows here that the price parity clauses can have anti competitive effects. And so why we think that our paper relates to this literature is that if like a platform like the meta search engine kayak is conditioning recommendations on prices on other sales channels. This might have similar effects as price parity clauses. So our contribution is to the literature is that first we document empirically that integrated platforms may favor their affiliates by distorting the search results. Specifically, we see that rankings are conditional prices of non integrated sales channels, such which can have be similar to price parity clauses. And the following I will provide you with some background about the online hotel booking industry. So as you know, consumers can book hotels through a numerous distribution channels both offline and online. And while traditional sales channels like the phone or even direct walk in stay important online bookings distinctively gained popularity over the last years. So around 45% of all bookings are done over some internet website and an additional 20% are done by email. And in what follows I will briefly describe now the two main business models in the online booking industry making use of this simplified illustration here on the right. So the most important business model in this industry is the one of an online travel agent. So this is booking.com or Expedia. And these platforms so offers for different hotels to consumers and they derive their revenues by hotels if consumers book rooms on the OTA website. The payment which you see here in red is red on so which is here in red is really a commission. And this is typically a fraction of the room price. So commission rates are usually around 15 to 20%, but they can also be substantially higher if a hotel wants to be shown more prominently on such a platform. In addition to these OTAs, how we call them, there are these meta search platforms like Kayak or Trivago. And these meta search platforms allow comparing offers of OTAs and also the hotel service channel. So for instance, if you go to Kayak, as you have seen in the example before, you can see for the same hotel, you can see the prices on Expedia on booking.com or the hotel's direct website. Here we took Evis just for the example. And meta search platforms do not derive revenues if somebody books, but they derive revenues by ascending referrals to these sales terms. So the revenues for the meta search website is generated once somebody clicks on an offer of Expedia or booking.com or Evis. And this is usually intermediate by, or is this usually done by a cost per click payment, which is the most frequent one. And the cost for such a click may range between a fraction of a cent if it's very unpopular and also to several euros. In our example, the platform Kayak and the online travel agent booking.com belong to the same company group as highlighted by the blue box, but to this we will get back later again. So for you, it's only important that most of these website transmitted bookings are somehow facilitated by online travel agents and meta search platforms. What's maybe interesting to know is that the OTA market is quite concentrated so booking.com is the leading OTA in Europe accounting for roughly two thirds of all bookings that are done on these OTAs and Expedia is the closest competitor. So Expedia is much stronger in North America, while booking is the market leader in Europe. And while booking.com and Kayak belong to the same company group, this also applies for others meta search platforms that they belong to some company group. For instance, you have that Expedia and Vivago are both in the Expedia group. So to learn a little bit about how important these meta search platforms are, I will present you with some numbers. So the German Competition Authority contacted a sector inquiry in 2017, which was generally about comparison sites. And interestingly, the report also features results of a survey of 14 OTA meta search platforms, which are active in Germany, but also in other European countries. So while the following numbers are only for Germany, they are likely to be comparable in their relative magnitude at least for other European countries. Therefore, I thought it could be interesting for you to show them to see them. So according to the sector inquiry, the meta search platforms received in 2017, 360 million visits over one year. So in comparison, OTAs received 1.2 billion visits per year. And knowing that every visit to a meta search platform leads to a redirect to a sales channel, at most one quarter of the OTA visits could be originated by meta search platforms. Of course, sometimes a visit to a meta search platform can also lead to a booking on the direct channel. So this even shows that the visits are still relatively important to meta search websites. Regarding the revenues, one can note that these meta search platforms gain revenues in one year of 200 million euros, while the online travel agents receive more than 800 million euros. Therefore, we can conclude that meta search platforms are able to obtain one-fifths of the expenses hotels have for being intermediated by third parties. So by this, given that this is just about a recommendation where you should book, I would say that this is a relatively high share revenues that these platforms gain, which also might explain why company groups having OTAs might be interested in buying them up at very high prices. Okay, we get now to the most important question. So how should the meta search platform kayak optimally rank hotels and sales channels? And when acting as an independent company kayak should rank hotels and sales channels, according to their expected profitability. So that is the product of the expected through rate of a hotel and sales channels and the payment that kayak will receive. However, when we're talking about an integrated firm, so when taking the interest of other booking holding companies into account, then kayak, of course, faces a trade-off. And this trade-off comes from through sources, so when promoting OTAs of the same holding, it increases their conversion rate, leading to higher revenues for these OTAs. At the same time, by not promoting competing OTAs as it would have done probably if not being integrated, kayak has potentially to sacrifice value able clicks, thereby reducing revenues from competing sales channels by referral fees, anything else equal. So the solution to this trade-off depends literally on the levels of payments and the conversion rates. But one might remember that the expected payments for booking commissions on OTAs are usually higher. So we said, okay, these are 15 to 20%. And why the cost per click is can be at most one or two euros. From this perspective, it might be the case that losses from not promoting competing OTAs might not be that important. And it might be more profitable for the company group as a whole to promote their own sales channels. However, to which extent this happens is of course an empirical question. Kayak itself does not make any statement, clear statement here, but they just declare that the search results are ordered by an internal algorithm that balances between the prices and the revenue for the results shown. Revenues, of course, might also include revenues of affiliated platforms, but of course this information is not given here on the website. Okay, so in order to study if Kayak is favoring its own affiliates and following up on our motivational example with a screenshot I had, we test two hypotheses in our empirical analysis. So first, regarding the horizontal ranking, we test whether sales channels which are affiliated with the booking holding have a higher probability to be visible and are more likely to be the position leader. And second, we would like to test if hotels which have higher prices on booking holding channels than on other sales channels are more likely to be ranked worse than the Kayak search results. So to do this, we conduct both descriptive analysis first always and then a regression analysis where we employ a lot of fixed effects. And for hypothesis one on the horizontal ranking, our dependent variable are the probability of a sales channel being visible at all and the probability to be the position leader. And then controlling for offer characteristics and various fixed effects we test whether the group affiliation has a significant effect on the sales channel visibility. For the second type of this on the vertical ranking, our dependent variable is in the locked rank of the hotel and the Kayak search results. And here we test if the company groups, having a low price on for a hotel, these two are different ranking position of this hotel in the search results. Of course, there are some challenges that we have to overcome in our analysis, and namely the first channels, and the first channel at first challenge, sorry, is that the ranking decisions might be affected by unobserved heterogeneity in demand and hotel popularity, and it also channel popularity. So first to deal with the concern regarding demand and hotel quality that we do not perfectly observe. We use in our regression hotel fixed effects, which control for the average hotel quality, and request fixed effects to incorporate potential demand fluctuations. And we focus only on one city so we think that demand fluctuations for this city might be captured already by looking at the combination on when is somebody searching, and when wants somebody to travel. The quality of hotels, of course, might vary over time. So and as we have data from three years we have collected in addition, customer ratings from Cliff advisor, because we know that trip advisors an important source for many people to assess whether hotel is providing quality, and especially people are looking at the most recent customer ratings and these most recent customer ratings we take as control, so that we can also see that maybe a kayak detects okay this hotel is getting less popular that we cover for this as well. Another concern could of course be that we do not take into account that sales channels might be more popular than others, and that this might also change over time. So for instance, the popularity of booking.com has been growing in our observation period. And also many hotels engaged in operating a more professional website and to account for variations in the popularity of these sales channels. We also collected the search volumes for all sales channels, so not only for the OTAs but also for the hotel website. So as you remember, for instance, ebus with its part of our core has a very, very sophisticated booking engine, where people can can go also to start booking on OTAs. And we also have to collect the search volumes for this to account for the different popularity. A more severe challenge is our second challenge is that we do not observe the cost per click. And these payments might vary between hotels and therefore affect both the pricing decision of hotels across sales channels, and the ranking decision of kayak. And if these payments would not vary over time, then we would be able to control for these unobserved commissions by controlling for prices and hotel fix effects. However, in practice, we know that it might not be the case that these commissions are constant, and therefore we have to deal with this concern in different ways. And the first way, the first approach to mitigate this concern is that we conduct a comparable analysis using data for the same city and for the same observation period of another meta search platform, which is Google hotels. And what is good about Google hotels is that at this point in time, it also allowed comparing different sales channels for different hotels, but not being integrated with any OTA. At the same time, so because the business model of Google hotels is similar to that one of kayak, it's very likely to be affected by similar heterogeneity in promotional behavior of hotels over time. So we think that if hotels want to spend more in order to be more prominent, then they do it at the same time on kayak as well as on Google hotels, and the same goes for the sales channels. And we only find that offers of the booking holding are given more prominence on kayak but not on Google hotels. This is consistent with our hypothesis of joint profit maximization in the booking holding as we don't serve this behavior as we have the benchmark in addition, we also apply a various other complementary analysis using the kayak data and exploit here regulatory changes, but also the chain affiliation because of the affiliation we know that from past studies that this is correlated very much on with the commission that that hotels pay and also that chains are usually more sophisticated in their promotional behavior. Okay. So, if there are no questions I will get now to the description of the data so we have collected for our study with my web scraping we've collected data from the French version of kayak for hotels in Paris with in a period of the years. And so what we did is we repeatedly conducted daily requests for overnight stays for two persons with different booking horizons so we sometimes look for the same day we looked for what is in seven days in one month and even when half a year. And the resulting data set consists of about 17 million price observations and ranking observations. And these stem from about 1800 hotels, which have offers on up to 22 different sales channels. And more than two thirds of these search results belong to an OTA that belongs either to the booking holding or the expedient group. So, which reflects as well that this market is very concentrated so this is also reflected in the availability of offers. To contrast our results found on kayak with those often non integrated meta search platform we also web scraped data from Google hotels, only for a period of six months. And as mentioned as controls for the unobserved changes in popularity, we add data from Google trends and also data from the advisor. So if you're looking back at our initial screenshot. Let me recall you what we have in mind before looking at the results so first we define a sales channels to be the price leader. When it has the lowest price. So here, for instance, Agoda is price leader for the city stay hostel. While Expedia is a price leader for the resident medicine blue hotel. And second we define a position leader if the sales channels is shown very prominently so this would be here booking for the second total and for the first total would be Agoda. And then we say a hotel is visible. If either it is the position leader, or if, if the channel also appears on here in this in this part of the, of the offer so that you can at least read the name and see the price. Okay, I will come now to the first descriptive evidence that we have regarding horizontal ranking. So for the definitions we can have a look at some statistics so we focus here on offers where there's a unique price leader, because otherwise, the explanation would be a little complicated so. And indeed for 72% of all hotel offers which are available on kayak we observe there's one sales channels which is offering a strictly lowest price. The low shows that the percentage of offers where group is the unique price leader. And so this is conditional on this sales channels being available. So if you if you are very attentive then you would see that this doesn't add up to 100, because sometimes of course hotels do not have a direct channel or not. So this is relative to be being present. For instance, when one of the OTAs of the booking holding is available, then it is the unique price data and 10% of cases. And similarly, what we can see is if an expedient OTA is available, then one of the expedient OTAs has the lowest price in 7% of the cases. Very interestingly, when a hotel has a direct channel then this direct channel offers the lowest price in more than half of the cases. So if we look now how often each group is the position leader conditional on being available. We can also see an interesting pattern. So why the booking holding is as I said only in 10% of offers to the unique price leader. It is the position leader and 22% of offers. The difference in the relative frequency of being position leader by 12%. Firstly, if we look at the direct channel, then we see that it's less often the position leader than it's the price leader. Here the difference amounts even to 60%. Of course, we did not take into account that maybe some sales channels are more popular than others, explaining the difference. The difference it could be that why some websites have a direct channel, have a known website with booking functionality, then this website might not be very popular among customers. And therefore, Kayak then predicts, okay, this will not be popular anyway, and then this shows this sales channel less prominently, even more. Furthermore, there could be other things like other preferences which we do not take into account, like that for specific hotels, customers might have preferences for specific channels and also customers with different preferences regarding how much they book in advance might arrive differently. And the control for this compounding effects, we conduct our regression analysis and I show you the results on the following slide. So on this table here, I show you the regression results regarding the horizontal ranking. So in column one, the dependent variable is whether a sales channel is visible in the search results or not. And what we can see here is that first, higher prices of our sales channels is are associated with decrease in the visibility of the sales channel. So in addition, being one of the price leaders increases the probability of being visible by about 23 percentage point. And if we have a sales channel, which is even the unique price leader, then this increases its probability of being shown visibly by 48 percentage points. What's maybe more important regarding the group affiliation. We see that anything else equal sales channels of the Expedia group or other OTAs have a lower probability of being visible than sales channels of the booking holding which serves here as a reference category. So if you, if you are Expedia, then it's 10 percentage point less, almost 11 percentage point less likely that you are even even shown in the search results. And know that we obtain this result even when we control for the popularity of sales channels by our good friend measures. In column two report then the results of the regression explaining whether kayak chooses a specific channel as the position leader and results are mostly as before. The main difference is that the hotel side channel has anything else equal a higher probability of becoming the position leader. At the same time, or other pattern persists so competing OTAs have a lower probability to be shown prominently than booking OTAs, even for equal prices and for when controlling for popularity measures. I will go now to the test of our second hypothesis regarding the vertical ranking. And remember, we said that kayak could favor its own subset own affiliates by assigning hotels with which are cheaper or non affiliated sales channels I was ranking position, the vertical sales search results. And in this table, I will once again provide you with some descriptive evidence before we go to the regression analysis. So specifically what I do here is I provide you with the average position of the first total in the search result which has the strictly lowest price on a specific sales standard group. So one can see for example here that the first hotel which has its lowest price on a booking holding platform is on average on position 18. However, if you look for the first total which has its lowest price on an expedient group OTA, this appears on average on position 38 so 20 positions later. And if you know that the number of offers which are shown anyway on the first page, then this is very likely that that hotels which have lowest price on Expedia are not shown on the first page. So note however again that these are descriptive statistics which do not take into account that sales channels or hotels might not have the same popularity. And for the more part of the descriptive result might be driven by the fact that booking.com is more often the unique price data as we had it in the slide before. And to deal with that conforming factors once again we do our regression analysis. And in this table are the regression results where we explained the vertical ranking and the dependent variable here is the log rank of the hotel in the search result or the log position. So this means when we have whenever we have a positive coefficient. For instance for price this means a higher price leads to the worst is to a worse hotel position because higher numbers mean that you're later in the search results and this is bad. So note first here that surprisingly having the lowest the lowest price of our sales channels does not affect how kayak is sorting the hotels. However, what we can see more interestingly here is that hotel offers were an OTA of the Expedia group as the lowest price, then these hotels receive a significantly worse ranking position the search results. So here you see that for all possible booking Expedia directional and other OTAs, we see that even when you're when you're the price leader, and you are from an Expedia group OTA, then then your this hotel is shown much later. This effect becomes even larger when we control for variations in time of the popularity of hotels. So this would be here. So the second column where we include the trip advisor data. And it becomes further larger when we control for the channel popularity which is measured by the Google trends. Then indeed in column three what we see is that hotels which are cheapest on it on the Expedia OTA are shown on average eight positions later. Okay, so this was it for the main results. As I said before, the main challenge for our investigation is that we do not observe the cost per click payments that kayak also of course takes into account when deciding its ranking. And these cost per click payments might vary between hotels as well as across time and might affect both the pricing decision of hotels on channels and the ranking position of the hotel at the Metasurge website. And to deal with this problem, we analyze data of another Metasurge platform which is not affiliated with other state channels, which is Google hotels, which is at the point of our study or at the point of our data collection, a pure Metasurge platform, comparison of hotels and channel prices. And as I said, so our assumption is that Google hotels is likely to be affected by similar heterogeneity and promotional activity of hotels and state channels over time, such as kayak. The data that we use for this analysis is collected over half a year, and then we manually matched hotels between the two Metasurge platforms with an overlap of 77%. We also compared the samples so we saw that the composition of hotels is similar in terms of hotel's quality, so we have the same share of hotels of a distinctive number of stars or we have a very similar average hotel size measured by the number of rooms and the share of hotels with which are part of a chain is also relatively similar. Unfortunately, this data was only collected for the vertical rankings or the hotel position. And for our robustness check what we do is then we use this data to study the question whether we also see the unfavorable positioning of hotels where Expedia has the lowest price also on Google hotels. Okay, and as I mentioned, okay, so overall in our kayak data set, which is collected over data collected over three years, we have, of course, much more observations so I repeat here the results that you have already seen before three slides ago. So I show you here in the second column, what happens if we restrict the IAC sample to the observation period and the hotel sample and the booking horizon that we have also collected equally for Google hotels. And what you see here is, yeah, the number of observation decreases substantially, but the main result stays the same. So what we still see is that having a lower price on an Expedia group OTA worsens the ranking position, the coefficient is here a little bit larger than in the pool data set, and the significance also decreases slightly. So in column three, then we do the same regression analysis using data from Google hotels. And interestingly, we do not observe the pattern that hotels with the lowest prices on Expedia OTAs receive was ranking positions for here the coefficient is smaller and insignificant. And even more if we compare the difference of the ranking of the hotel on kayak and Google for the same search request regarding this travel date and the search date and the city, then we see even that the difference between, between both ranks is significantly positive. So, which means that indeed there's a significant difference of how hotels which lower prices on Expedia OTAs are positioned in the vertical search results on kayak compared to how they are positioned on Google hotels. Okay, I come now to my final slide, which gives our preliminary conclusions because this is ongoing research and we are happy about any comment that you might have. So, one, our conjecture was that kayak takes other factors than hotels and sales channel popularity into account when deciding about the ranking. And while this is of course not surprising for a profit maximizing firm, it raises the question of whether kayak optimizes its ranking only with respect to its own revenues, as it claims also on this website, or whether it takes the joint revenues of the integrated firm into account. And as I said before, if the cost per click revenues for all sales channels on hotels would be the same and constant across time. One could directly interpret our results such that kayak favors its affiliated sales channels in the ranking decision. However, as we know that in practice cost per click revenues might vary across sales channels in time. Yeah, so, so we have to exert more caution regarding our results. However, if our results would be solely driven by higher payments of affiliated sales channels. So if the whole result just comes from the fact that booking.com is paying a lot of money to kayak, then this could still be consistent with kayak favoring its own affiliates because as it's the same company. There might be also other ways for within an integrated company to do to compensate for these payments. And then in addition, we also do not observe these patterns in the hotel rankings of Google hotels which is not affiliated to do any other OTA. From our perspective this finding raises two concerns. So first, we find that the optimization of an integrated meta search platform may lead to worse ranking of hotels with lower prices on non integrated sales channels. And this can have similar effects as price parity clauses, which have been prohibited in many countries because they could be increasing commission rates and final prices. And second, our other concern is that deviations from a ranking that produces the highest match value for consumers made it only reduce consumer surplus, but also a locational efficiency. Regarding the letter point of course a limitation of our analysis is that we cannot provide a definite conclusion on what sort of ranking of hotels and channels is socially optimal. So it's unclear in this regard whether favoring affiliated channels achieves worse results than the ranking optimization if the meta search platform and the OTA were not integrated. So indeed, a separate meta search platform could have incentives to buy a search results, even more towards less popular sales channels. On the contrary, a meta search platform that is integrated with a popular OTA might improve consumers' search quality as it internalizes the profits of a subsidiary. So here in our application it could mean that from a consumer perspective if booking.com is the most popular platform in Europe and consumers really want to book only over this platform, then this integration might even lead to better search results because then it also pays off for Kayak as member of this company group to favor booking.com. However, empirically distinguishing between these outcomes is unfortunately beyond the scope of this present article and therefore left for further research. So thank you very much for all your attention and for participating in my presentation and I'm looking very much forward to all of your comments and questions you might have. So thank you very much.