 Thanks so much for having me. I know many of you have worked in related area, especially about two sided platforms. I really look forward to everybody's comments. Particularly grateful to Conrad for being willing to read a preliminary draft and discuss the paper and to entry for moderating the whole session here. So this work is joined with Genja and Liyada Webman at Iney Noise Institute of Technology. Just for disclosure, none of us have any substantial financial relationship with Airbnb. So this is a pure academic exercise. And the data we use on Airbnb is from a third party. So this really has nothing to do with Airbnb except that and they are the subject of the study. So we're motivated by the fact that platform is not only a batch maker but also a rule maker. We know many platforms are multi-sided on one side and they face consumers. This could be buyers, could be guests, could be viewers, could be writers, could be diners, gamers and the platform often sets some rules on these players. On the other side, we would have typically the supplier. They could be sellers, the hosts, advertisers, content app developers or drivers. So this is quite common for most platforms. It's also common that the platform would set different set of rules on those two sides. Typically the rule on the consumer side is, I will say, a light. For example, it could be very easy for them to register on the platform. They often face a low price or even zero price or even negative price. But on the other side, for the supply side, the rules are typically more explicit and sometimes heavier. This could be that they have to provide certain information for registration. They often have to pay substantial fee in order to be eligible on the platform. There might be some quality control like quality standard. And sometimes the suppliers even subject to sort of less ability to give feedback to buyers, for example, in eBay. So in this situation, although the two sides appear to be asymmetric in terms of rules, this asymmetry is making economic sense. For example, the suppliers would be willing to and subject to this relatively heavier rule because they know the platform are able to give them access to a lot of consumers that might be less costly to access as compared to alternative ways like having their own marketing team, for example, or reach out to potential consumers and so forth. So this works because of the positive network effects between the two sides that they're hoping when the platform are able to accommodate the consumer side that would help the supply side because of the positive network effects. Okay. However, recently we also see a lot of a lot of players, especially on the selling side complain about platforms. The typical complaint is, oh, the platform is sort of a bottleneck here. We are subject to pretty heavy rule from the platform which somewhat is unfair from the complaints perspective, but they feel like they have nowhere to escape because the platform already sort of control access to a lot of consumers. And they really wants to have somewhere to escape. So following this thought, the policymakers and academia people have been thinking about, okay, we probably should introduce competition at the platform level. If we have competing platforms, then the sellers or whatever the suppliers subject to a heavy rule from a platform will be able to switch to the competing platform. And that would be a constraint on the first platform. However, if we think a little more about this, we can see the economics is not that easy because for the competing platform to be attractive. It also must have the other side going forward. It must have enough concerns so that the supply side will be willing to at least choose the competing platform. So that means on one hand, we really want to understand the demand side. We want to know how the buyers, the guests actually choose to use which platform or maybe multi-homing on multiple platforms. And of course, we want to understand the positive network effects between the two sides. And we also want to know how the suppliers wants to consider their choice of platform and sometimes maybe multi-homing decisions. And in addition to all of that, we also want to be aware of that might be within platform competition. It could be even if all the players or most of the players decides to use one platform, they still need to consider how they compete against each other in order to appeal to consumers on the other side of the platform. So was that insight? In this paper, we want to ask the following research questions. We want to understand how Airbnb's 48-hour rule, which is a pro-guest policy. I'm going to give you more details about this. We want to understand how Airbnb's 48-hour rule would affect guests, hosts, and the platform as a whole. We want to ask question, pretty straightforward question, to what extent this rule would increase guest demand or decrease guest demand. Does it drive away a host or keep the host on the platform? Who is mostly affected and so forth? And we also want to know to what extent the platform competition will affect Airbnb's rule setting incentives. I think we have a pretty neat setting here that we know Airbnb is competing with verbal in the U.S. We have our data from 10 large U.S. cities. We can actually measure the local competition between Airbnb and verbal around each listing. So this empirical setting would allow us to understand the role of platform competition around this Airbnb 48-hour rule. And finally, we want to see what the insights from this empirical exercise would help us understand whether platform competition will really mitigate the related and the trust concerns, as I mentioned above. So just for the structure of the talk, I'm going to give you very brief literature review followed by empirical setting and some basic economics. Then I will show you the impact of the 48-hour rule and price, quantity and host dynamics. We're also going to examine the role of competition for the impact of the 48-hour rule. At the end, I want to show you some back of envelope calculation of this 48-hour rule for Airbnb's gross booking value. And if we have time, we can discuss the welfare implications. So in terms of literature, we know very well that economists have studied a lot about a symmetry pricing decision of a platform. The typical wisdom is that platform will divide and conquer and they typically would appeal to one side and give low price or even zero price to one side and then try to get an avenue from the other side. And the side they sort of try to appeal to typically will be more price elastic and probably more likely to single home, especially if there are competition at the platform level. And empirically, my previous work with Mark Reisman, we have looked at this pricing decision. We do find it's a metric pricing and it's very smart competition patterns in the context of sports car convention. So that's on pricing. Recently, we see a lot of literature on now pricing. Most of that literature is theoretical and some is also empirical. So for example, the platform may strengthen seller certificate and or impose minimum quality standard and platform sometimes may hide certain buyer information from sellers in order to improve match efficiency. And we also see series showing that platforms may have incentive to hide at on cost at noise in ratings reduced search quality or highlight high price offers in order to reduce seller competition on the platform and sometimes avoid repelling the sellers away from the platform. In many of these series, they make ambiguous prediction about the effect of platform competition. So hopefully our paper will buy some empirical evidence regarding the role of competition in platform competition. So just give you some empirical context, you know, Airbnb and has grown very fast as of December 2020 when Airbnb went public. It already has four million hosts, seven million listings across 220 countries, the total gross booking value in 2019 has reached 38 billion. And the Airbnb is pretty large in the absolute size. And if we just look at his whole unit listing, even as early as 2016, and it's only next to the largest hotel chain, which is mayor out international. And if we look at the shares of available rooms in the whole hospitality industry, Airbnb's share in that quote unquote market and has reached 4.5% in 2016. And in the US, we understand Airbnb is the largest digital platform for short term rental. And followed by verbal booking.com and flip key. It's very interesting to see that Airbnb actually was founded later than all these existing competitors. So Airbnb was founded in 2008, while verbal was founded more than 10 years ago, before Airbnb at 1995 and bookings was 1995 and flip key was 2006. So Airbnb was able to grow pretty fast and exceeding all these alternative platforms in terms of scale. In this paper, we noticed that there are two pro guest rules about cancellation on Airbnb. The first is that Airbnb actually would have automated mandatory and guest facing posting of host cancellation. So that means if you as a guest has a book of particular property, and the host later on cancel it, and that cancellation behavior will be automatically posted as a review by the platform on Airbnb. And this practice has been there for a long time. The other practice is a more recent after May the 1st of 2018, Airbnb adopts so called 48 hour rule. That means guests can cancel the reservation within 48 hours of booking if the reservation is at least 14 days away. I will give you more details about this particular policy. The important point is that both of these pro guest rules about cancellation, they are Airbnb specific verbal actually does not adopt these two rules. So that gives us an interesting comparison between Airbnb and verbal. So in this paper, we're going to focus on the effects of the 48 hour rule. So our data, our most of data data comes from air DNA, which is a third party that kind of scrape listing information from both Airbnb and verbal. So our data goes from 2017 January to December of 2019. And that's because January of 2017 is the earnest time that air DNA started to scrape listings on verbal. So we use that time and on to compare Airbnb and verbal. The data, and you can think of that as a listing month's level. We purchased the data from 10 US cities. So that's New York City, Los Angeles, Chicago, Boston, Austin, Atlanta, Washington DC, Seattle, Houston, and New And we focus on the whole unit rentals only because verbal only have whole unit rental while Airbnb have both whole unit rental and a private room or share room. So we're thinking that wholesale whole unit rental to be a separate market. So our observation at the listing month's level, we know the average price per night. We know the occupancy rate of this property in the particular months we know the total number of reservations in that month. We also know what kind of a guest cancellation policy day host adopts at that time. We know whether the host actually had any host cancellation reviews in the history of reviews and we know when this cancellation review had happened. We also know other regular information such as total number of reviews, average ratings, whether the host is a super host or not, whether the host offers instant booking or not. What's the response rate of the host, number of bedrooms, bathrooms, photos, minimum stay of the listing. More importantly, although Airbnb does not give the precise location of a listing, it gives proxy geo coordinates. So that allows us to define the local competition within a point three mile radius of each listing. We also can, I think AirDNA give us an indicator of whether this listing is a co-listing or not. We made extra effort to identify whether this listing is originally Airbnb or originally verbal. And if it's originally Airbnb when this listing started to co-list on verbal advice versa. So that's our main data from AirDNA. We also supplement this data using zip code business pattern from the US census. And that give us the number of hotels and number of restaurants per zip code year. And we use that to proxy the tourism activities in the particular zip code. In the earlier time, we also did some Google search to collect number of attractions, museums, seam parks and national parks per zip code. But because this is not time varying, it typically will be absorbed by zip code fix effects or listing fix effects. But the zip code business pattern will still survive because we do observe time varying number of hotels and restaurants. So that's our main data. Let me give you some sense of what I mean by host cancellation review. So this is an example of a particular listing. And you can see that's just a list of reviews. One that I circle in red here is a automated host cancellation review. The text reads the host canceled reservation two days before arrival. This is automatic posting. So this is a standard language from Airbnb. And this language allow us to identify whether this review is automatic Airbnb mandatory review on host cancellation. So just to give you some sense, about 11% of Airbnb listing months have any host cancellation review in the history. So that's a minority, but not a trivial fraction of minority. And if we follow Airbnb listing over time, we know having one more cancellation review typically would reduce the price by about 4% and reduce optimistic rate of the listing by 4%. So that means the host cancellation review is really regarded as a negative signal by guests on this platform. So we're going to keep this in the background later I will see this host cancellation actually become a tool that host can use in response to the 48 hour rule. So let me give you the example what I mean by the 48 hour rule. Suppose you book a reservation on Airbnb a month in advance. Okay. And the listing would subject to some guest cancellation policy. This is typically chosen by host, but Airbnb have some standardized definition of guest cancellation policy. And we can think of that as three types. It could be flexible, it could be moderate or strict. If it's a flexible that means the guests can freely cancel it until 24 hours before arrival. If it's a moderate, the guests can free cancel until five days before arrival. If it's a strict, then if the guests cancel seven plus days before arrival, they guess typically have to pay about 50% of the reservation price, and plus the cleaning fee. So, we're going to refer to flexible and moderate host at the loose host wall street ones will be the other group. And we have about 26% of Airbnb host to be strict host as of April of 2018. So, in terms of this 48 hour rule, before Airbnb adopt 48 hour rule, you can see that's the cancellation on flexible and moderate will be free. Within the 48 hours of booking, but the cancellation and still makes the guest to paid Airbnb service fee. Okay. But after the 48 hour rule, Airbnb has waived the service fee upon cancellation. And for the strict host before the 48 hour rule, guests can only get 50% of the refund, but still have to pay cleaning fee and Airbnb service fee. However, after the 48 hour rule, the platform basically require the strict host to honor free cancellation within 48 hours. So that means the guests can get 100% refund and does not need to pay the cleaning fee. And again, the Airbnb service fee will be waived. So you can see this change sort of have, if we can decompose this change, it has two changes. The first is, is directly affect the strict host, but not directly affect the loose host. For the strict host, this could mean less revenue and higher uncertainty and potentially higher operation cost. And it also makes the strict host and loose host less differentiated, at least in a dimension of guest cancellation policy. You may have a question of why the Airbnb host have to adopt this kind of cancellation policy. And that's because those hosts typically just run one or very few number of properties. So for the heritage hotels, hotels typically would have many rooms, they would have more flexibility to adjust if one guest have canceled, and, or if other guests wants the other rooms at the same time, they have a lot of room to juggle with that kind of uncertainty. And host on Airbnb because they manage very few number of properties, and this is harder to manage. So that's why most hosts actually adopt some type of cancellation policy. And in terms of why we think the strict host may potentially face higher operation costs after 48 hours rule. So then we give you example to suppose this is a calendar for a strict host. I use different colors to represent reservations by different parties. So some may have reserved property for relatively long time, and some may be quite short. If you look at the orange color here, so this is a party that only reserved for two days. Okay, and we have this property almost fully booked. Because that's typically, they typically the strict host would have a higher occupancy rate because they arrange more popular properties. So in this case, if this orange party canceled, then you can see that now you have those two days, empty and subject to future booking. However, if the yellow party already reserved this property before, and the blue party has reserved afterwards, the strict host must find another guest that just need this property for one or two days within these two days. So that could be harder to find, especially if this property is popular, and is sort of hard to move other guests around the same time. So that's why we think the strict host may face potentially higher operation costs after the 48 hour rule. So this differential change between strict host and loose host is the first major change by the 48 hour rule. Another major change is an Airbnb has waived Airbnb service fee for all cancellations within 48 hours. So that applies to all Airbnb host. So this is a absolute quality increase. If you think that way for all hosts. So this could raise consumers overall quality perception of Airbnb and make them more willing to book on Airbnb. So I want you to think about the 48 hour rule as two changes. One is reduce the distance between the strict host and loose host. And the other is an enhanced the quality of all host. So if we're thinking from the Airbnb's perspective, what are they trying to get from this 48 hour rule. We don't know exactly what's the objective function of Airbnb, especially how they trade off today's game versus tomorrow's game and so forth. So if you're thinking statically, the monthly Airbnb gross booking value per city, what depends on the number of listings on Airbnb would depends on average price per listing, what depends on average occupancy rate in the months and if the host resort to host cancellation would leave a bad mark on Airbnb and make consumers less willing to book on Airbnb. And that will be a loss for Airbnb. So that's kind of the very simple mathematical formula in our mind. And if you're thinking the economics. First, the guest may perceive Airbnb as a better quality after 48 hour rule. So that means they may be more willing to pay a higher price. And they may more willing to book on Airbnb so that could increase the price and increase occupancy rate. So if you think there's a positive network effect here, you have more guests on the guest side that could attract host to join Airbnb. So that could imply we have an increase in the number of listings. So that's the first economic force here. The second one is, and now we have potentially higher costs and higher uncertainty on the strict host, the strict host, some of them may be really upset by this role they did may decide to leave Airbnb so that could generate a decrease in the number of listing, or they stay on Airbnb but they now have to charge a higher price in order to compensate the higher costs and higher uncertainty so that could imply a higher price by the strict host, and accordingly the demand for their property might be going down. And if the strict host also use host cancellation more often that could imply the loss due to the host cancellation would increase. So that's the second economic force. The third one is that now we have less differentiation between strict host and loose host on Airbnb this could intensify the competition between them. And that could imply that the average price may be lower because the competition is more intensified and occupancy rate might be higher to some extent some listings on Airbnb may be squeezed by this competition and make the listing on Airbnb less profitable and some of them may decide to exit so that would imply a decrease in the number of listings. So I make this table just to give you some sense that we have different forces of economics going on here. And some may make the number of listings go up. Some may make the number of listings going down. So whether the net effect is up or down is empirical question, similarly for price and occupancy rate. Some forces may say the price should go up and some may say the price will go down. So we're going to sort of give you the net results from our empirical data. So to give you some sense of the data showing you the summary statistics of Airbnb and verbal. The price per night on Airbnb is slightly lower. The number of reservations is slightly higher in Airbnb but occupancy rate is lower on Airbnb. Host cancellation review is only available on Airbnb so we only provide the summary statistics for Airbnb about 11.5% of listings have any host cancellation review. In terms of the strict cancellation policy, Airbnb have about 26.5% with strict policy, while verbal has about 31.5%. In terms of co-listing you can see that's about 13% of Airbnb listings co-list on verbal, while about 39% of verbal listings co-list on Airbnb. We're able to compute the competition index for each Airbnb listing because we can say let's draw a circle around you about 0.3 miles radius. We can count how many verbal listings in this radius and divide that by the total number of listings that would give us the extent to which verbal listings is competing in this very local area and that average is about 0.124. So I want to say something about the strict host. The strict host in terms of price is actually charging a higher price. So despite they have more strict cancellation policy, they are able to charge higher price and that's typically because they manage more popular properties. And if you look at within the strict cancellation policy, the different bars here represent the number of host cancellation reviews they have here. So you may have expected a monotonic relationship here. However, we see a nonlinear relationship. Those that have one host cancellation actually have a higher price than others. And that's probably because they manage popular properties and they sometimes even if they cancel one, they still appeal to consumers. And because they're very popular, they probably sometimes will be more willing to use the host cancellation in order to manage the flow. So that's why we see this nonlinear relationship. You also see this nonlinear relationship in the occupancy rate that you see the strict host have higher occupancy rate and those that have one host cancellation in the strict host would have a higher occupancy rate. So more generally, the strict host are more likely to have popular properties, they are more likely to call list on Airbnb and verbal, they're more likely to face higher competition index. And many of them are more likely to operate in tourism area, and they're more likely to have cost cancellation reviews. So in other words, we can think of strict host product sometimes is higher quality because they manage more popular listings and their location might be better. But they are also lower quality other dimensions, especially in cancellation policy in guest flexibility and this kind of uncertainty guest must face. So that may give you some baseline results of the effect of the 48 hour rule. This is from the very straightforward the ID specification would compare Airbnb versus verbal and before and after 48 hour rule. So we can see that this 48 hour rule actually would push up the price in Airbnb by about 2.8% and at the same time, the occupancy rate would go up by about 2%. So that's the overall results. If you look at the loose host only loose host in Airbnb versus the loose host in verbal, we see that effect is even stronger. The price increase is about 3.6%. The occupancy rate increase is about 2.3%. If you compare the strict host on Airbnb versus the loose host on Airbnb, we can see that this demand expansion potentially demand expansion effect is less on strict host. Although the 48 hour rule has the most direct effect on the strict host. And still in the net, the strict host still enjoys slightly higher price and slightly higher occupancy rate after 48 hour rule. So that's the baseline results. We want to point to some caveats in this identification because we're using verbal as the control, but verbal obviously will be affected by the competition between the two platforms, and we can only identify the relative difference between the two. So that's probably a fundamental caveat in our method. We have a key assumption here that no other major policy change happening on Airbnb as of May of 2018, and that relies on assumption that the two platforms are pretty similar before this policy change. And we have tested this by pretreatment test. We also did some plausible tests and do not find any effect in this hypothetical plausible test. So if you want to see the raw data, this is raw data, you can make the judgment yourself. Okay, so the red line representing verbal, the dash line representing Airbnb, you can see that the two platforms are pretty tracking each other in listing price before the 48 hour rule. And then after 48 hour rule that gap between the two is narrowed, because Airbnb price is increased. So for occupancy rate in the below picture, you can see that the two track each other pretty closely, and then the gap is a little closer after 48 hour rule. So that's corresponding to our DID results. So that's just- Can I ask a question on the previous slide? Yep. So it seems like the gap only starts shrinking later on after the policy change. So I guess I have a question about why it's so gradual and it would actually be good to show separate coefficients for each of those to see when the effect stabilizes and how long it takes to get there. So I think that's kind of one kind of question. And the second related question I have here is there's other work that shows that Airbnb hosts are actually oftentimes not changing their prices. They don't even change their prices for an entire year. So then some understanding about why they're- What's driving the changes in the prices? Can you identify ex ante hosts that are changing their prices a lot? And is the effect coming from those hosts or is the effect coming from the hosts that are maybe more- have more rigid prices? But is this a big enough change that they've noticed it anyway? So those are coming to- Yeah, all very good questions. Yeah, we'll do all of that to our ability. Okay, so that's just on P&Q. You may also wonder about host dynamics. So here we're showing the results of a number of listings in a particular zip code by months, whether that changes before and after the 48-hour rule. Again, we're comparing Airbnb versus verbal. And this results show that the Airbnb listing goes down about 2.5% in a zip code. And we also ask who is likely to stay and who is likely to exit after the 48-hour rule. So we defined exit at not being active three plus months after the 48-hour rule. So these are just the summary statistics about the stares versus the exitors in their statistics before May of 2018. So you can see that those who exit seems to manage very non-popular properties. They have much lower number of reservations, much lower occupancy rate. They charge much lower price. They are less likely to be strict hosts, actually. They're more likely to have consolidation reviews, and they tend to be in the area of less competition. Okay, so that's on the number of listings. We can also ask the same question in terms of, okay, what's the number of cross listings? What about the likelihood of having any cross listing for a particular listing? And to what extent do they still use strict cancellation policy or not? So the results here is showing that the number of cross listings increased slightly on Airbnb after 48-hour rule. So that means Airbnb listing is slightly more likely to coalesce on verbal than vice versa. Okay, and the probability of cross listing is higher for Airbnb listing after 48-hour rule, and Airbnb listing is less likely to use strict cancellation policy. So this is a graph showing the adoption of strict or loose cancellation policy. This line is 48-hour rule. So you can see that after 48-hour rule, the above two lines are for Airbnb and strict and loose. So the strict one is going down over time. Well, the loose one is slightly increasing. So this suggests that some hosts may have switched from strict cancellation policy to loose cancellation policy. Interestingly, this trend is also appearing on verbal. So what the results I show you on the last slide is a relative difference. We don't know whether this verbal change is because of the Airbnb policy or just because of the natural change on verbal. So there are also some dynamics within Airbnb. So for example, if we ask whether the host will use host cancellation policy, we find the strict host after 48-hour rule actually more likely to use host cancellation policy. If we draw the picture for the loose host versus strict host, you can see the strict host is indeed much more likely to use the host cancellation policy. And recall that host cancellation policy is actually quite a negative signal to guests and could affect guests negatively. So this is sort of a very interesting finding, but something that we didn't expect a priority. So let me use the last few minutes to tell you what's the role of local competition we find in this effect of 48-hour rule. So these are the 10 cities we look at in our data. New York and New Orleans have the highest competition index and also New York has the highest number of listings. So the specification is very simple that we add a triple the ID here that's we interact the Airbnb post with whether the area is above median of competition as of April of 2018. Okay, so you can think of we divide every city into above median competition area versus below median competition area we want to see to what extent we have different effects of the 48-hour rule in these two types of competition areas. Okay, so this one I'm showing you say if we just look at the loose host on Airbnb and verbal, we look at the effects for above median competition versus below median competition. So you can see that the effect on PNQ, they're both larger in the less competitive area. Okay. If we look at the strict host on Airbnb versus strict host on verbal, again, we see the effects on PNQ to be larger at the less competitive area. So somehow the competition will sort of weaken the effects on PNQ. So we can see the effects are positive but less positive as compared to the less competitive area. So that's one finding. If you look at the competition's role in those dynamics for example in number of listings, we don't find any interaction effect on the number of listings, but we do find slightly positive effects on the number of cross listings. Okay, so we can see some of the cross listing effect is driven by the above median competition area. If we look at the cross listing likelihood or the likelihood of adopt strict policy, we do see the competition matters here above median competition area have more cross listing and increase and the same competition area have more decrease in strict policy. If we look at the cross listing and exit between strict host and loose host, actually we see the strict host is less likely to change their cross listing. And this is mostly because the strict host have been close listing, even before the 48 hour rule. And they are also less likely to exit as compared to the loose host in the above competition area. If you look at host cancellation in the above median competition area, we see that the strict host is more likely to cancel the host, I'm sorry, the strict host is more likely to use the host cancellation, especially if they're cross listing. Okay, in the above median competition area. Well, in the below median competition area, we see slightly an effect but less magnitude as compared to the above median competition area. So this seems to suggest that the strict host already cold listing. And after 48 hour rule, they may feel squeezed by the Airbnb platform because they're already cold listing. They may strategically use the host cancellation on Airbnb. Maybe they use the canceled reservation to appeal to consumers on verbal. So that's, that's one change we observe after 48 hour rule. So just to give you some welfare calculation here on consumers, we try to calculate the change of willingness to pay for Airbnb after 48 hour rule. We know less about the host cost so it's hard for us to compute a host property. It's hard to compute the host profitability. For the platforms, we can compute the number of listings. We know that's going down, average price is going up, average occupancy is going up. We know some increase in the host cancellation. So we can summarize this and change in gross booking values. In terms of consumers, we use a very simple utility framework. We think the utility may be a function of the price, maybe some listing characteristics and we allow this utility to increase or decrease for Airbnb listing after 48 hour rule. We separate Airbnb single listing versus cold listing. And according to the typical IO structure that we can estimate this utility function by looking at the market shares of the listing versus outside good. And we can use the parameters here divided by the parameter on price to sort of put a dollar's amount on the consumer willingness to pay for Airbnb quality increase after 48 hour rule. Okay, so I'm going to skip this. We use some private room price as IV for log price, but we acknowledge some attributes like cancellation policy or host cancellation reviews, and these could be changed by 48 hour rule, but we just use it as given we don't have instrument for them. Okay, to show you the results. These will be the coefficients, and that's basically imply that for Airbnb only listing they they perceived quality increase is about 1% of the price. They perceived quality increase for Airbnb cold listing is about two to 3% of the price. So that's sort of confirmed that consumers somehow have a better utility out of Airbnb listing after 48 hour rule. So for the Airbnb's gross booking view of gross booking value, we can compute the change of their gross booking value for each city after this 48 hour rule. And we show that this percentage change in the gross booking value is positive every city, but this is going down in the cities with more competition. So, so just to summarize everything here, we talked about platform competition before. And we can see that they, when the platform change their rules on the guest side. So in our case will be more pro guest and against host, we see one effect is the demands are going towards this platform. Okay, and this effect is weaker if there's more competition between platforms. The second effect we see is that on the supply side, we see more cold listing and some of the loose host actually started cold list on the competing platforms. So that's the second effect. We also see some of the strict host now use more host cancellation so that potentially lower their quality. So that's the third effect, both the second and third effect become stronger when there is more competition. And we don't find much effect about the, the positive network effects because we see the demand side is going this way, and we see a supply side is going the other way. So they're not going the same direction, at least on this margin. And it seems suggest that positive network effect is not as strong as maybe some of us have saw before. So just in summary, we think the evidence does not support the most stylized antitrust story based on the positive network effects. And we find the role of platform competition to be somewhat complicated. The supply increase does not always follow the demand increase. So this is against the positive network effects prediction. And we find the suppliers that mostly directly affected by the 48 hour rule actually they try to escape by host cancellation, rather than exit. So they stay on the platform, but they're sort of providing a relatively lower quality product than before. And we also see the within platform competition, although the 48 hour rule affects strict host directly, but most changes in exit and cross listing actually occurring on loose host, rather than in strict host. And the role of competition on the effect of 48 hour rule tends to weaken the effects on the consumer side on PNQ, but strengthen the effect on the supply side in terms of cold listing and host cancellation. And these competition effects could undermine Airbnb's incentive to adopt a 48 hour rule, according to our calculation. And lastly, there's a caution here that we do find a competition do not always undermine the progester rule. For example, for host cancellation and reviews, we find competition strengthen the value of host cancellation for loose host, but weaken the signal for strict host. So that means the role of competition is still probably rule specific. So that's all I have. Thank you so much.