 Thank you very much for being here today, Guji. Thank you very much for the invitation and with all the negatives of digital platforms, one of the positives is that we can connect throughout the world during a pandemic and discuss our work, so I'm very happy of that. And this is a joint work with Krishna Kampali, who is a student at Columbia, and of course, you know, Raghuvarshan. So the motivation is very simple. Kampali seems to be reluctant to fund investments in spaces that are close to large digital platforms. And one of them say, this is a real thing, the scale of these companies, and their impact on what can be funded is massive. And this is at first moment seems like strange because the prospect of being acquired should spur not stifle innovation and investment. And so before we go into trying to understand whether this statement is right or wrong, let's first look at the data and see what happens in this space. So if you look at VC early investments in the social media space, you see that the dollar amount has plummeted. Now, this might not be that surprising because there are waves in VCs and there was a moment in which this was the hot moment and that later on is not anymore. So we would like to do a more systematic possibly event study of what happens in term of investments around the time that you had this acquisition. So how do we go empirically? We identify the major acquisitions made by Google and Facebook. So acquisition above 500 million in the period 2006-2016. So these are the acquisitions and the events in a sense. And then we look at the treated firms as firms that are similar to the one acquired in the same rough space. And then we define a cycle adjusted measure of investment as a measure of how much VCs put into them versus what they put to them in the software industry in general. And we compute this cycle adjusted measure around acquisitions event. And then very much in the finance style of events studies, we aggregate this across events to try to to give a sense of what happens. And just in case you want to know this, these are the acquisitions. And as you know, Facebook, the two major acquisitions are WhatsApp and Instagram for Google. They go from YouTube to Waze and so on and so forth. And what you see is in before the period, you have a fairly constant level of investments. And then shortly after the acquisition, this is the blue line, you see that this level is going down. Now you say, OK, maybe this is not a phenomenon unique of Facebook and Google. Maybe this happens throughout the entire software industry. So what we've done and this is the red line, what we've done is we look at other acquisitions done by other firms, same magnitude, same period, do the same analysis. And what we get is the red line where you have a much bigger pre-trend. So it's a bit of the either beholder, whether the level after is above and below depends on what year you take as a reference. But clearly what you see and in the paper, we do that more systematically than just a graph. But what you see clearly is that there seem to be a unique phenomenon that in the space surrounding digital platforms, you see a decline, a relative decline in investments after the acquisition takes place. And so why that's the case and what is the difference about this world and the world of micro one-on-one that we all know so well where this effect will not happen. So first of all, we know for a fact that this is a sector where there are a few gigantic incumbents. This is a sector where there are very strong network externalities and a sector where there are some switching costs and where platforms are two-sided or multi-sided and one side is charge is your price. That's clearly the case for Facebook and Google. And so the question is whether the intuition that we have from micro one-on-one might not be there once we put these elements together. And so before I try to present quickly the model, let me give you the intuition and intuition is extremely simple. So in any acquisition, the price of the entrance depends on two facts, depends on the competition among bidders and depends on the entrance outside option to go it alone. Now one of the characteristics I just described above is that in this space, incumbents are very powerful and tend to be monopolistic in the sector. So competition among bidders is not really what drives the evaluation, which suggests that the outside option to go it alone becomes quite important. Now what this go it alone value depends very much of course on the quality of the entrance, but also on the number of customers that the new entrance can attract. And this is simply because of the network effect. And so now the customer decision to switch or not depends itself on some other parties. And now in the paper circulated, we call them techies. The paper is under a vision. So we updated that I recall app designer. So that this decision is swayed very much by the app designers, which have of course some cost of switching. So if I need to prepare my software not only for one platform for another, I need to adapt my software. And so entrant is already a disadvantage because many of the outdesign are with incumbent. And ironically, this disadvantage can be exacerbated if you see an acquisition down the line. Because if you see an acquisition down the line, you're not as an app, you're not willing to pay the cost of switching when you know that everything would be integrated pretty soon. And so the result is that high expectation of being acquired, depressed the number of app designer that switch. And in so doing, they depressed the attractiveness of the new platform for ordinary customers. And this is depressed the number for two reason. One is there is an inference about expectations. You see very few app designers. You think that the platform is less valuable because you think app designers no more than you do. But also because of the network similarity, you use a platform because of the app that are there. So the fewer app available, the less attractive is the platform. So that depressed the number of ordinary customers who choose the new platform. The pressing is standalone value. And for the reason I explained before, this depression is standalone value, the press also the acquisition prices. And then once you understand that the press acquisition prices is a very easy step to say this depressed the investment by potential entrants because if I cannot get out at an attractive valuation, I am not investing and that closes the circle. Can I intervene if you because there was a question about say you compare the situation of without network effects or or you showed this graph with the blue and red lines. Yeah. And there was a question from Yossi Spiget asking why the red one was also going up after the accusation. So what's the economic rationale for that to start with? Because you then move to the network effects and explain the intuition for that. I think the question is, why do we see the red going up after year zero? So I'm not so sure that you can make a big statement about one year. And if you look on average, the three years before and the three after, you see a slight decline in investment. So I'm not trying to make a strong statement that in the software industry in general, the investment go up after an acquisition. What we can do and we do it in the paper systematically with the right statistics is that there is a difference between acquisitions conducted by Facebook and Google and other acquisitions in the software industry where we don't expect those networks, the narratives and the platform of features to be so prevalent as they are in the Facebook and Google. So you should interpret the thing you should focus on is the difference between the blue and the red line, not necessarily the behavior or any single year in either line in particular of the red line. Okay, then there's a follow up question. I don't want to take so much time on that, but just to clarify. So then you see asks about the blue line does not change much either. So the same can be said about the blue line. I think that his question is about the difference between red and blue. It looks similar before and after. Yeah, this is why I refer to the paper. Where we do the difference statistically and they are statistically different. So that's the only thing I tell you. Okay, thank you. I think this is the only question. Okay, so thank you for the questions and please interrupt me. That was a good moment because now after giving an intuition, let me describe briefly how do we go about modeling and then the specific of the model itself. So as we know, networks and ideas lead to multiple equilibria. So if I think that many customers will switch, I will tend to switch to and vice versa. If I don't, and we know this situation can lead to sort of the sunspots equilibria, very similar to bank runs. And so that's the reason why we're going to go in the direction of global games in order to get a unique equilibrium, which is related to fundamentals. And that allows us to do some comparison. So let me go to the model, which is very simple. Sorry for interrupting. Just before you dive into model, there was another question I saw with a little bit later. What exactly are the app designers in the intuition that you provided? So what are those app designers? That's a question from Andre Hadjou. Sure. I think that we provide in the paper multiple explanation. One is literally the explanation of app designers. So people that provide some additional software that makes the use of the platform more valuable. So for example, there is an issue whether if you use Google Maps or Waze before, you can crowdsource the information about where the police cars are. And so there was an app on Waze that was providing that information. And when Waze was acquired by Google, there was an issue whether Google will integrate that particular application or not allowed and whether they will develop or not and so on and so forth. Now, in this particular case, there is some liability issue because, of course, you're trying to scam the police so that an extra layer of problem, but that's one interpretation. The second interpretation is some kind of influencer that are more informed than everybody else because they might be sort of a technician of the sector or they might have some positive utility out of searching and finding new applications. I am always the last one to switch to anything because I'm not a techie and I don't want to pay that cost, but my son or other people I know are really sort of active and they find that they're proud to be the first one to use an application and in doing so, they provide information and reassurance to a larger public and they're proud to be the front one and they're proud to be the influencer. So you can take either interpretation and the model I think goes to in both interpretation. Thank you. Okay, so there is an incumbent platform I, which is threatened by a new entrant E and without loss of generality, we assume that the quality of the incumbent is normalized to zero and the quality of the new entrant, the incremental qualities theta and as I described, there are two groups of customers, the app designers, which we assume have a major lambda and ordinary customer with a measure of one. Now, the app designers at date zero, they observe a public signal about their quality and their posterior belief of quality after they observe is distributed normally with a mean of Q and a precision, I'm sorry that is the other way around, is alpha and if the app designers will adapt the app to the new platform, they get a per-peered incremental utility, which is somewhat equal to the incremental technical quality of the platform. The idea, if you stick very closely to the interpretation of app designers, what you can think about is you're going to recover the cost of investment only if enough people eventually will use the platform and people eventually will use the platform, the quality is good enough. So this is directly a functional quality. We take a very simple linear function, but don't interpret that so narrowly. And then app designer will have to pay a cost to adapt and once they adapt, they can easily multi-home and this cost of rotation is uniformly distributed between zero and level as a par bar and is individual specific. So important, the signal they observe is public, but their cost of switching is private. And then there are the ordinary customers. The ordinary customer don't have sufficient information to switch right away. They wait to see what the app designers do and what do they observe? They observe two things. They observe how many app designers switch and they observe a private signal about the incremental quality of the entrant and this private signal is equal to the true value plus a random noise, where again, the precision here is better. And the difference is, while the app designers don't have a network externality, the ordinary customer do have some network externalities. So their switching decision is based not only on the expected technical characteristic of the new platform, but also the number of customers a platform will be able to attract or retain. And so I think that this is repeat, sorry. And then at date one, you have the two companies decide whether to merge or not. Now the share that each one gets is determined by a bargaining process that I will describe later. If they do merge, the superior technology will be adapted by the merge entity and all the customer will enjoy it regardless of whether they switch or not. And so the acquire and the merger ensures smooth transition to all customers without any problem. If the two companies don't merge, they will survive for a number of period independently with a different technology. Now, what is the value of the new entrant? The platform service is given for free to customers in the spirit of Google and Facebook in exchange for their data. So where do the platform gets a profit? There are other profits from the advertising side of the market. And what is their value there? Their value is a function of the number of customers a platform can obtain. And for truth-based simple reason one is that more customers on the platform means more eyeballs for the advertising. And also, of course, more customers also means better data, so better targeting add and so on and so forth. So you can think about the value of the entrant platform as V of P where P is the fraction of ordinary customers that switch to each. And of course, this is an increasing function. So this is the timing of the game. The app designers see the public signal and decide whether to switch or not based on the adaptation cost. Then at time half, the ordinary customers see if the app designers have adapted, observe their private signal and decide whether to switch or not. Then at date one, the incumbents and the entrant decide whether to merge and the terms of the mergers. And then basically, if they don't merge us, life will continue as it is for the future and then we'll discuss what happens there. Sorry, I have a question. Sorry. I have a question on the timing. I believe some other people also have. Could you go back to the timing? Absolutely. So basically, given that the app designer see a public signal. So the fact that as an ordinary customer, I observe an app designer adapting doesn't give me any information about the quality itself. Is this correct? Sorry. The answer is yes or no. So we observe that all the app designers observe the same signal, which is not observed by the customer. So there is some value provided by that. But as I will explain, if I have time in the robustness, this is not crucial. The crucial component of the app designers is that they provide some network externality to the ordinary customer. So that's the key ingredient of the model. Okay. So basically you're going to argue that it's not their information, private information, which is more valued, but it's more that they generate network effects early in the game. Absolutely. Okay. Okay. So there are, I think, other questions probably on the model. So Chiara from Agalia asks, so why app designers pay a cost to adapt to a new platform while adapting to the merge platform, which embodies the innovative features of the new platform, is costless? So I think it's very simple. Going back to an old platform that we all understand and know very well, think about the early, maybe this is too old for many of the participants, but when Microsoft was coming up with Windows, there was an issue of whether some products that were developed, for example, for the Mac base, should be also ported on Windows or not and what cost this would be. And so if I am the producer of Excel, and initially I had produced only for the Windows, only for the Mac environment, actually what happened with Excel, they first produced for the Mac, they first produced for DOS, and then there was an issue whether they will produce for the Windows, the new Windows, or they produce for OS2. Most people probably don't remember what OS2 was, but it's a different operating system that at the time was a potential competitor of Windows. So I think that they definitely have a cost of portability. Once the, let's say, Microsoft buys OS2, they make everything backward compatible, and so there is no need to pay that cost anymore. So that's kind of what this is all about, that's intuition for the cost. And by the way, we eliminate, because we wanted to add the model symbol, we eliminate a cost of the ordinary customer as well, I think that if we assume a cost C for customer to switch, everything goes through with a C, an additional C in the equation that I will describe in a second. And there was also another question from Giancarlo Spagnolo, he asked, I believe it is Spagnolo, yeah. So he asked whether the, is it important that the app designers do not care about network effects? I mean, would the results also go through? No, it's not important, it makes solving the model simpler. But I don't think it's crucial. Okay, thank you. I think that's all. Okay, so again, perfect timing, because now we're going to the incentives in the game, once I've described the game. So the app designers know that if they switch, they expect a game of Q, that's the expectation of theta, for a period that is only one, if the merger takes place and is equal to N, if the merger does not take place. So as a result, each app designers will decide to switch, very obviously when the benefits are beginning to cost, so the benefits are one plus M times Q, the cost is the individual cost SI, so that's the decision to switch. So remember that SI is uniformly distributed, so the measure of app designers who switch day zero is given by that expression there, if you are in the interior, and of course can be different if you are the boundaries, we proceed in the interior for simplicity, but nothing changes if you analyze the boundaries, it's simply a more complicated analysis. So think about, given a public signal Q, you have a mass of app designers switch that is given by that particular expression. Now they are the ordinary customers, they observe how many app designers invest in the platform, and now they know, or they have an expectation about M, so they can back out Q, so that's why observing or not observing theta doesn't really make a difference because there is no noise there, they can observe Q perfectly, and so they have an initial expectation of theta and they combine this initial prior of theta with a private signal that they observe at the period two, and then in this, they're going to have a posterior belief where the average of this posterior belief is given by raw, which is of course a weighted average of the two signal they receive and the precision is given by the sum of the precision of the two signals. So an ordinary customer will switch if and only if the network externality adjusted quality of the entrant is superior, so this is the equation, let me try to explain in steps what this is, so first of all, raw i is their expectation of theta, then they have their measure of the externality provided by ordinary customer, that's nothing else but P of raw, so is the fraction of ordinary customer, they switch as a function of their expectation of raw, and then there is the number of app designed in the switch and that's the 1 plus M lambda Q divided by S of a bar and this should be bigger than the fraction of ordinary customer that stay in the old platform plus the app design present in the old platform and here we assume multi-homing by the app designer, so if you are Excel and you produce a new version for Windows, you can do both Windows and OS2 at the same time in the paper where lies also the case in which you can't and basically nothing major changes in just a threshold of switching changes, so that's the decision rule and this is a typical global game, so we conjecture that ordinary customer will follow a switching strategy where they switch if their prior quality exceeds a certain threshold and then when a customer is the marginal switcher, he has to believe that the fraction P of customer will switch as well for this to be an equilibrium, so the fraction P should be at least as high as his own posterior and so we can calculate P at the switching point in the following way, at the switching point, you should have the rho star is plus 2, 1 minus the cumulative of a normal function calculated at the rho star plus the other terms that are there and so we first identify this function S of rho and if gamma, which I realize I didn't define here but it's basically a measure of the precision that is sufficiently small, so if there is not too much uncertainty then the function S is always increasing in rho and so there is a unique switching equilibrium that we can at least plot as a function, for example, of the period that you expect this stuff to continue and so the first and important result is that the optimal switching point decreases and the fraction of ordinary customer switching to the entrant increases in the number of periods that the tech he expects, this is the app designer, expects the entrant to remain independent, so the bottom line here is quite simple, is if you expect this new entrant to remain independent longer, you will have more app designers switching to the new entrant per given expectation the quality difference, this will increase the network's knowledge to the ordinary customer joining the entrant and in turn this will reduce the quality threshold at which you are going to switch and this enhances the expected value of the entrant as a standalone entity because remember the value of the new entrant is a function of our many customers at switch and what is important here is that price is not a factor so the customer cannot be attracted by offering a lower price because there is a zero lower bound and we can discuss how realistic this zero lower bound is but at the moment let's keep a zero lower bound so now we have the merger game where the merger game we assume a traditional bargaining where mu is the bargaining power of the incumbent take v of 1 as the discounted sum of profits of the merged platform now we know that v of 1 is going to be bigger than the sum of the value of the entrant of the incumbent if they operate separately the profits of the monopolies are bigger than the profits of two duopolis is always exposed efficient to merge and in our simple bargaining game where there is no friction you always do the efficient thing so you always end up emerging so the only question that is relevant is what is the price at which you merge and of course this price is driven by the incumbent bargaining power but is also given by this term here which is the standalone value of the new entrant now notice that this standalone value is a function of the customer's switching and this is indexed by M the stands for a merger will take place because as we just described this number of people switching is a function of whether you expect a merger or not and so imagine for a second award and I'm not saying this is how the world should be but imagine an award in which for whatever reason mergers are not allowed then the payoff of the entrant will be only the standalone value because there is no game, there is no merger, nothing and this standalone value will be V of E of P and M where not merger and now what is important is that the P of NM is larger than the P of M for the reasons that Corollary 1 suggests because if you don't expect a merger more designers will switch as a result more ordinary customer will switch and so the expectation of the value of the entrant is going to be higher if you don't expect a merger than if you expect a merger and if that's the case what you can say if they enter a bargaining power is either zero or small enough then the payoff of a new entrant is larger in a ward in which you cannot merge than in a ward that you can merge and it's larger simply because in very simple terms you can pre-commit what is interesting is once you think in those terms you start thinking about what are the other strategies to pre-commit not to sell out but one is to hire an egotistic CEO that is full of himself and I don't want to name names but a lot come to mind recently so I think that that could be an optimal strategy the other is you give an extreme amount of power to the incumbent so that if it's a little bit egotistic the founder so if it's a little bit egotistic doesn't sell so think about Mark Zuckerberg he has a dual class stock in Facebook and that's the reason why he turned down a very generous offer of Myspace back then everybody thought he was crazy eventually he's still laughing but at the moment was clearly any venture capitalist would have sold happily if he had the ability it was only his desire to be in charge that the block that now let's go back one period and now assume that the potential entrant does not get the theta as a mana from Evan but has to pay a cost so imagine that the new entrant grows a technology of quality theta and before deciding whether to invest will compare the expected profit given a cost of R&D so that she draws the technology knows the cost of the technology and looks at the expected valuation of entry and then of course only if the expectation is bigger than the cost you're paying and so in a world in which incumbents are prohibited from entering then of course you are going to see that the number of people entering will go up rather than down because they increase value of the investment so long story short is not the question if we take the fact that acquisition to place and was also allowed by the antitrust authority remember this effect is seen over years is not an event study in the financial that is a day this is over a certain number of years so if you see that acquisitions are clear by the antitrust authority the market updates that future acquisition will take place and so that the incentive to enter will be reduced because customers will not switch as a result of the game we just described so long story short I think this is what we think Wenger who is a venture capitalist from Union Ventures had in mind now what is interesting is that this might lead to a resurrection of the very old nation industry protection argument because you can say that if you prohibit in a certain economic area acquisition by an incumbent you're going to see development of local platform and we know local platform really did not develop very much by contrast China did develop a concurrent platforms in part as a result of prohibition of the existing platform to enter and the control experiment is what happens now in India with Tik Tok as a result of some border dispute with China the Indian government prohibits restriction on the use of Tik Tok and immediately you saw a nation industry in India trying to compete so I think that you see some elements of that now we don't want to come out of this paper and the paper is much more in detail on this and I'm happy to answer questions because I'm running out of time we don't want to come say it's optimal to prohibit all acquisitions because as you can see easily from this model acquisition are exposed efficient now the thing that clearly comes to mind is that you want to make switching costs as low as possible so mandating common standard and interoperability are certainly things that go in the direction of making the efficient decision so in if you don't go that route the route of emerges as a lot of problems both in theory and practice in theory because it prevents the industry from realizing exposed efficiency in practice because if you take an antitrust case by case approach it's hard for an antitrust authority saying you shouldn't do a merger because exposed the merger is efficient so if you every single merger would be approved exposed because it's efficient but exact might be not the right thing to do so conclusion I think that the only thing we're trying to do in this model is construct a simple model to rationalize the existence of this kill zone even if there's a kill in the title these are different from the killer acquisition the Florian and other developed this is more an area where nobody wants to invest because they don't get sufficient return and I think that it depends very much on free frictions and in the paper we do a better job that I could do in the presentation trying to see the robustness one is network externalities switching costs and lack of cross competition in the paper we discuss for example attempts by some browser like brain to pay people to use the browser these successes have not been these attempts are not being very successful so I think there is some evidence that the zero lower bound applies not only macro but also in micro and the evidence we find is consistent now is not obvious how to address the problem what it is important is that kind of digital platforms do provide different intuition that the obvious 21st 20th century economics but I'm very scared saying that to people especially to lose given that they are the forefront of this analysis and with this I'm done excellent timing thank you very much Luigi for that so we have discussion by Luis Cabral so for five minutes and then we will open for Q&A okay thank you let me know when I'm past five minutes and I'll try to stop then so thank you for allowing me to discuss this I think the best part of this paper in my opinion is that it establishes the existence of a so-called kill zone and I think it does so in a very compelling way so just to quote from the paper normalized VC investments in startups in the same space as the company acquired by Google and Facebook dropped by over 40% that's a very large number 40% in number of deals and 20% in the three years following an acquisition so that's big in terms of economic effect that's big in terms of statistical significance and I'm not an empirical economist but from what I could read and understand in the paper the data seems pretty clear in that point and I think this is a strong empirical point you know what follows I'm going to be somewhat critical about other aspects of the paper so I think it's important to understand that to me at least if there are nothing else in the paper and there's a lot but if there is nothing else in paper this would justify its existence because the kill zone has been in the law and sort of industry law for a long time but to the best of my knowledge there has been no convincing and systematic evidence of its size and prevalence so I think the paper does that and does that well so the part of the paper that I'm less excited about is is the model I believe the behavioral premise that techies are unlikely to adopt a new platform because they anticipated it will be acquired I personally don't find that as compelling as the empirical evidence I don't have any data to support my skepticism by the way let me be very clear but my guess is that it's not a very good description of the main forces at stake I believe that techies are very proud to be the first to adopt like Luigi's kid yes of course definitely do I believe that those early adopters are afraid that the platform will be acquired and therefore that their efforts will be meaningless I'm very skeptical about that had I predicted the future would I have been less likely to invest in learning how to work with Instagram knowing that it would be acquired by Facebook I don't know in fact I'm not sure how many people here know but before the acquisition the market for photo sharing social networks was dominated by a hipstomatic and Instagram so question number one who had heard the word it's thematic before if I could run a poll I doubt that more than 5 or 10% of people in here would Andre would certainly but I don't know about many other people so and I cannot stress this point enough I don't think it's I'm not saying that the doctors are not rational so when I present an example of this it's not that they're not rational I believe that you know that's my job I have to believe that people are rational it's that the business model of incumbent firms and entrants is extremely difficult to understand and predict and I've been repeating that in a variety of contexts if you want to understand the digital space how it differs from pharma, from cement from breakfast cereal what have you it's precisely the extraordinary difficulty in predicting business models even the incumbents themselves have no idea what they're doing most of the time they just encounter success as it comes yes they may see that this may have some potential competitor who knows in terms of modeling this leads to very diffuse forward reasoning and in this sense I think the digital space is very different from other spaces and to be go technically to what the paper does I'm not an expert in global games but I'm very skeptical of applying global games in a situation when you know priors and posterior are so diffuse as they are in this context I'm not criticizing my opinion of the type of application where I wouldn't use global games as a selection criteria and generally speaking and related to that one problem that is related to diffuse nature of this industry is the distinction between preemption and substitution between substitutes and complements the paper is focused on new platforms that having strong network effects will be a substitute but in practice this distinction is not at all obvious in fact I mean I hate to go back to the Facebook example but that's the one that everyone goes to yes she could have been in some ways it's a substitute for Facebook but it's also a compliment in fact it was clearly not a killer acquisition in the Florian et al kind of sense we know that Facebook did not kill it and in fact it has made a ton of money off of it more generally I believe the vast majority of the acquisitions in a digital space have more of a complimentary than a substitution effect and also for that reason I think the policy prescription of forbidding mergers it's very highly controversial not just because of the explosive efficiency as Luigi mentioned rightly so but because it's so difficult to determine whether a certain acquisition is a substitute or a compliment I simply cannot see a practical way of short of the nuclear option from now until the rest of the world Google and Facebook are hereby notified that it cannot make any acquisition other than the nuclear option I find it very difficult to have to implement that sort of policy now universal declaration of discussion rights clearly states in its article 13 that I discussed and has the right to include a shout out to his own or her own work and so based on that article I would like to mention that you know I've been also working on in this space as it were and I believe that my paper standing on the shoulders of dwarves deals well pretty much with the same problem but in a different way and it's kind of interesting that in some ways I would it would agree with Luigi's paper that banning mergers may actually have a positive effect on radical innovation so we do agree on that for different reasons by the way but I think that's kind of an interesting point of agreement and I think it's a useful thing to think about the question of course is the opportunity cost of that and the opportunity causes a huge in my opinion a huge decline in incentives for incremental innovation and I don't have any empirical evidence but I've been working on with the venture capital I mean with the startup a bunch of startups in this space and I would say that you know one out of two maybe two out of three their business model is very clear I want to be acquired innovation for buy out is a super important phenomenon and so I think that's the opportunity cost of a policy of forbidding acquisitions because this ecosystem is very different from other industries that they're more used to precisely because it's so difficult to protect and to transfer IP without just simply outright acquisitions and so I know that this is not the core of Luigi's paper but I think this policy is extremely problematic you have to understand radical and incremental innovation and I could talk more about that but I want thank you for that I think if we have time at the end of course we will talk more about this I think this is an important point on the policy implications the first one is from can I respond to Luis for one minute sure but if you want to get more questions that's why I was otherwise I forget so first of all thank you a lot for your comments I think they're very interesting two things one I would like to say rightly so that the complication of analyzing this space is how complicated the consumer choice is and now depends on the forward expectation of what happens and so I think that's exactly why Global Games is the way to go now you might disagree with our specific it depends a lot on what you think in the case of the Microsoft application in the past I think that this is clearly a good assumption in my view because there is a cost of moving excel from one environment to the other and we saw that that was a relevant factor today might be different but today maybe the cost is in individuals so I personally don't want to learn during the two weekends after what's output out a new privacy restriction a lot of people started to switch to signal and so I have a group of old high school friends in Italy and they it was a discussion about whether you move to signal or not and we still are not converging we are kind of multi-homing and it's a nightmare because you don't know now you have double cost of looking at the two things and part of it is people do not want to learn a new interface and so if you know that eventually signal would be integrated with what's up and it would be the same sort of things you don't want to pay that cost so I think that the idea that switching costs by a merger are real and the fact that the expectations imply I have implication for pricing are real and once you put the two together certainly it's not the only model maybe it's not the perfect model to capture this intuition but I think that it is a model to capture those two intuitions that in my view and it seems also in your view are crucial on the Saucidum compliment I cannot agree more with you in a sense we have a discussion in the paper in the presentation about how slippery the distinction between Saucidum compliment is in this space and but even if it is and so I'm sympathetic to the application part if you are a judge and you have to decide it's very difficult to decide but I think that in reality many applications have a little bit of both so think about the acquisition by Google of Maze which is an alternative way to do Google map and is Maze really challenging the search algorithm of Google probably not but is it complimenting and making stronger the market control of Google by merging absolutely so I think that it's a tricky business I agree thank you very much so I would like to just raise the other questions so that we have now open Q&A for the last five minutes so Heski Barisak would like to ask two questions so the first one was about indeed to follow up question on Louis's point he's saying that you know what about subsidizing innovations and entry rather than maybe a more radical policy like banning mergers yeah I think that the more I present this paper the more I think about this issue the more I think that clearly a policy of banning acquisition is not the way to go so the question is what is the best alternative to me the best alternative for this particular problem but in general is to increase the interoperability because at the end of the day the reason why the decision of my friends and I to be on signal or on WhatsApp is so complicated is because it's costly to be on both at the same time and very often at least I don't know if you who are more sophisticated but I take the switching cost as exogenous something technological etc but they are not and to me it was very interesting I listen and I strongly encourage all of you to listen there is an episode of Planet Money a podcast dedicated to power ventures this was a little start up in California that 2008-2009 was trying to kind of disintermediate social media by aggregating that so if you wanted to pause your picture contemporaneously on Instagram if thematic and Facebook you had power venture that will allow you to do that and will allow you to capture the messages from all this and see them in a interface of your choice so that was the perfect solution to interoperability now this didn't work no it worked perfectly from a technical point of view Facebook threw the hell out of them and was able to establish in the court of law that if I give Luis my login and password for Facebook and Luis logs into my Facebook account with my permission he commits a federal crime called hacking and he can go to jail and the irony is that Facebook had done that for the better part of their early existence but of course like every firm they are a pirate when they enter and they become conservative when they are established it happens to people too and I think that in doing so I think that they made interoperability very difficult interoperability is the way to go sorry for the long answers sorry we have another question sorry from Stefan about this interoperability point he basically raised whether the long term outcome would differ significantly with interoperability because there are also network effects playing a role so would that solve everything we have not worked out across all the fees and dollars but my intuition is that once you have interoperability then the major friction that drive this model that you are concerned about how many customers are going to get disappears and so if you have a higher technology all your customer with all the customer would switch to you almost instantaneously increase the return to technology now of course you're going to say oh but the comments are going to put additional friction are going to try to but that's in a sense what we're trying to fight I think that the fight should be in trying to reduce this entrenchment cost because this what these switching costs are rather than allow them because they exacerbate the problem so let me just say that very quickly I do completely agree that the key is going to be interoperability switching costs more generally behavioral regulation I mean I did the recent efforts in European commission but also in North America I think going that direction I think that's way more important than regulating M&A that and I'm glad that we agree on that because not everybody agrees on that by the way there was another question from Andre had you given that we are nearly ending our one hour slot so I would like to just raise it myself to make it quicker so Andre was asking that I think that point you already answered whether it's important that the network effects are not something that the app developers care I think you said that it's not very important for the main results but that it just gives simplified analysis the second question is the quality of the entrenchment they perceive is assumed to be wiped out by a merger what if that quality persists which is certainly the case with Instagram you know it's true that Instagram continue as an independent entity but not I should qualify not really independent in a sense in the moment in which in the famous weekend in which Facebook kick out Trump for example a lot of Trump supporters were questioning where to go and clearly they couldn't go to Instagram because they knew what will happen so the fact that even if the alternative exists the fact that is not pushed that hard but most importantly does not have the independence of movement is really important in this expectation of switching so I think that the reason why Facebook could do that could kick out Trump so easily is because at the end of the day many customers have nowhere to go and I think that the example gave of my friends were not by the way Trump supporter but they want to switch to a different platform to suggest how difficult that is okay so given that it's 3pm I keep receiving more questions I think there is more interest on this topic so let me just officially end the seminar here and for those of you who have raised questions on the chat I already told you would like to ask your questions yourself so you can stay here with us and then we can have further discussion for a couple of more minutes but I would like to just tell everyone thank you and the recording of the session at this point so that we have the official part ended but we keep discussing further