 So good afternoon also from my side and Christoph comes from the ECB's monetary policy department and thank you very much Philip for this excellent introduction. So I have the pleasure to share session one on monetary policy credit and financial crisis and let me congratulate the organizers of the conference for the perfect foresight. They had to put a session like that on a day when the Nobel Prize in economics was awarded to three outstanding economists for their work on the very theme of this session. I would also like to thank them and everyone behind the scenes for making this event run smoothly. So we are glad to have four very distinguished speakers and discussants in the first session today. Let me briefly introduce our speakers for the first session who are well known of course. So we're glad to have Thomas Philippon who is a regular guest at ECB events and has spoken previously for example at our Sintra forum. He is professor of finance at New York University and today he will present his recent work on the too big to fail problem that as we all know came from the center with the collapse of Lehman Brothers 14 years ago and has preoccupied researchers since then. We are also happy that Michaela Pagel has accepted to act as discussant of Thomas paper. So Michaela is associate professor of finance at Columbia Business School. So I very much look forward to Thomas presentation and the discussion. Let me just give some housekeeping announcements. So the organizers have allocated 25 minutes for the presentations 10 minutes for the discussion and then we will have 10 minutes for a question and answers including also Thomas reaction to Michaela's discussion. What I would kindly ask the audience to do is to submit any questions for the Q&A via the WebEx chat function. The presenter will be able to see those but I will then also read them out so that the entire audience knows about the questions. So without further ado Thomas please the floor is yours. Thank you very much for organizing this conference and for inviting our paper. This is John Warke with my colleague at NYU Olivier Wang and yeah we'll try to propose a kind of a new way to think about systemic risk and too big to fail which is not as pessimistic as the usual view. So the context of the paper is the regulation of the financial system post Lehman brother post the global financial crisis and broadly speaking we've done three things. We've increased capital requirements. We've beefed off supervision and we and we make we try to make progress on resolution. So there's been lots of work I was involved in the BIS review of all systemic risk regulation that was published last year and so my broad assessment is I think in terms of capital requirements you could argue whether we went far enough in the sense that you know maybe you would choose 15 percent or 17 or 19 or 22 and then there's some it's not clear exactly which is the perfect number but there is no doubt that we've made very significant progress. So broadly speaking capital ratios are twice what they used to be before the crisis and that's clearly a success now for supervision I think it's to some extent quite similar. I think we've improved a lot of supervision and again just like for capital requirements you could argue we could do more nobody is thinking it's perfect but I think the broad assessment is that there's been very very significant progress. The third pillar the resolution is a lot more uncertain it's a lot less clear you know whether we really made significant progress and the reason is is very deep it's because you know you can improve the coordination of regulator you can you know create single point of entry resolution regime you can work on cross border resolution and then on all of these points separately there's been a lot of progress but the very basic idea that the basic issue that's at the core of to big to fail is that no matter what you do it's still going to be essentially impossible to resolve all the banks at the same time during a systemic crisis. I think that is the fundamental tension and there are even people who like me think that we've made actually very significant progress on the resolution I think the living wheels are a very significant progress even people like me who are cautiously optimistic in that dimension even these people remain to some extent skeptical that this fundamental tension that you know no matter what happens if there is a big crisis you will not be able to resolve all the banks at the same time I think that remains to some extent unaddressed and so therefore from this from this line of reasoning came a somewhat pessimistic conclusion which is you know we're still going to have to bail out the banks therefore the expectations of bailouts will remain and will continue to distort funding costs and and feed more hazard and so what's our paper going to do well we're going to agree with the premise which is yes this fundamental remains I think it is unlikely to ever be credible for any financial authority to commit to let many banks fail and be resolved at the same time I think we agree with that premise okay but we're going to disagree with the conclusion we're going to argue that the conclusion comes from a very limited set of contracts and in fact models that people have used so the main point of the paper is that even if you are going to do a bailout you can still get rid of more hazard okay so we agree with the premise but we're going to show you that the the conclusion does not follow so we're going to be able in some cases of course there is like tons of cadets in the paper we're going to show that in the benchmark model we can implement the first based allocation with time consistent policies so to be clear what that means is that if you end up in a situation where it is exposed clearly Pareto efficient to do a bailout because the alternative is a price is a meltdown of the system and a very big recession so presented with this alternative any sensible regulator or policy maker or central bank will always decide to save the system so that's what we make time consistent so you cannot commit to let the house burn even in this world we're going to be able to implement the first based allocation and the reason is that we are going to step away from the one-size-fits-all bailout so in the model we're going to have bailouts and it turns out they are exposed efficient in large crisis so exposed you always want to implement some some bailout to avoid in very large crisis to avoid a big recession the banks are going to fully anticipate that they're going to understand that if they all got into trouble together then a significant fraction of them will be bailed out they fully understand that part and yet we get the first base so what's the trick well we're not going to bail out all the banks in the same way we're going to implement a tournament or it could be the result is more general than that but it could be implemented as a tournament and so that's the main that's the function of the paper now it turns out that in in the process of doing that we get interesting new definitions of systemic rig the difference between size and inter connections the the importance of sustainability and stuff like that and so for those of you who are maybe less researchers purely academic like me but who followed more like the debate about how do you define too big to fail for instance you know so if you look at the definitions usually we say well it's not just size it's interconnection these lack of institutions of like that so that's kind of the theory of it but then in practice when you look at what's done it's pretty much size only so I always thought that this debate was a little bit like semantic at the end of the day we just would just mean size well I changed my mind a little bit because in this model I will have for very precise reason I will tell you why it does matter the degree of substitution and what interconnection does matter independent of size but that only appears once you understand the the key time consistency issue so let's start with the baseline model so it's quite simple you have two periods and you have a bunch of banks so we are interested in too big to fail so that means the number of banks and must be more than two or three but less than 100 right so we don't have the low of the odd number or anything like that we have just a few large banks and the banks have assets and liabilities assets are standard you have the book value a and then return our i which is random which is the source of risk on the liability side we're gonna just split between anything you could potentially bail out sorry bail in like so t lack and deposit so it's slightly different from the usual definition where you would separate equity and then maybe debt short or long or subordinated and then deposit we're gonna lump all the things that you can potentially bail in as equivalent to equity here and all the deposit on the other side and just for simplicity and also to emphasize that we are not about we don't really want to work on the issue of capital requirement we want to say given the balance sheet what's the best way to act exposed there's going to be a capital shortfall if the banks equity or t like falls below some threshold which is proportional to asset value so that's classic and then in most of the paper i'll show you how actually you're like a general welfare function and with this assumption that the financial distress happens when e falls below some threshold we provide some micro foundation using exactly funnily enough a diamond deep big model of runs and order of fire cells to motivate the key things to show you that our results applies to a broad set of environments okay so what's the key issue of moral hazard bank i is going to choose its safety investment x i so x i is the key moral hazard viable right the bank chooses safety x i to maximize its expected value but not of course taking into account the negative externalities that distress in finance can post the rest of the economy so formally the returns that the bank is going to get is going to be most of the time so in normal times with priority p0 so think of p0 as like 0.95 you know like 95% of the time you're in a normal state in a normal state um you know old safety investment is useless so this is just a cost for you so f is decreasing and concave in x so all the things you do to prepare yourself against a crisis are cost and then when the crisis doesn't happen it's it's a waste so if you only look at the good state of the world then your incentive is to make x as low as possible um the return of the bank in the normal state is just f of x decreasing plus some noise that we add for just for the sake of completeness of contracts in the crisis states x becomes important okay so with priority ps you're going to be in a crisis state and the number of states with a crisis could be any number you want so it's indexed by the state um and you know state one is not that's bad and then the state big s is the the really big the really bad one and the key thing there is that um if you choose a higher level of safety x then you're going to have a better outcome for the risk distribution so formally it's like first order stochastic dominates x is higher than your returns are going to be better in this crisis states okay so this distribution function is increasing in the stochastic dominant sense in in x um all right so that's it that's that's the bank uh that's the bank problem okay so they choose their safety and they understand returns are random and then some day might get into a financial distress now when they maximize of course they take into account their expected bailout so m is a bailout from the from the government from the central bank and to be clear in again here we're going to assume that m takes the form of a cash transfer but i've worked of course with these models many times and we show equivalence to equity injections you know debt guarantees and things like that but here for simplicity imagine m is just a cash transfer that you use to increase the equity ratio of the bank so the bank know that the banks anticipate the fact that they might get some transfer from the government now the tension of course is because the government has a welfare function that is different from the sum of the value of the banks so the welfare from the perspective of society is the first part is of course the same so that's where they agree with the banks so the sum of the equity value of all the banks the debt is going to be fairly priced so the net value here doesn't show up and then the tension appears here so there's going to be a an externality function that says that if many banks are under distress then there will be a negative externality on the rest of the economy and that's where i'm using this welfare function v because to show you that the model is quite general and we provide micro foundation for v in the appendix so if the banks are under distress then there will be a loss for the government or for society so the government would like them to intervene by beefing up the equity of the banks that's this m transfers but of course there's a fiscal cost and the fiscal cost depends on the aggregate bailout m which is the sum of the little m's and we index it with a parameter gamma which captures the funding cost for the government so that we can compare say what a government like the us would do to what a government like the greek government would do so Greece would have a high gamma so a high cost of funds so may not be able to do a large m okay then the first best allocation here is just to maximize with respect to x and m this welfare function okay and just to be clear this is the first best in the sense that this is assuming that the government would choose the level of safety for the banks themselves okay so this is assuming a way more hazard so that's not realistic it's not it's it's the first best benchmarks we can compare later okay you just simplify that you just maximize welfare subject to all the constraints and then that's how it defines the first best so to make progress it's useful to actually specialize this value function here so in most of the in the benchmark model we use the one in the paper with a virile encode that we did on systemic risk and that's a welfare function that says that really the the cost of financial crisis comes from the fact that the entire system is under capitalized so we don't really care which bank it is and we don't care the specific allocation of capital shortfall across banks what matters is the aggregate capital for for the entire banking system taken as a whole so in that case the welfare loss or the the key metric is equity of the total system so total equity of the bank plus total bailout minus total capital requirements aggregated across all banks okay so in this simple model we can compute the exposed optimal aggregate bailout okay and it's just going to be the race between the cost of fund for the government and the value of beefing up the equity for the banks and in this simple case you can show that something is pretty intuitive you know there's going to be no bailout so m equal zero if the crisis is moderate and then if the crisis becomes large enough that you get into the big systemic risk region then you start to provide a bailout this the shape of the bailout depends a little bit on your cost function so if you take a country let's say like the us where this cost function is linear so gamma prime is constant and the government can it pretty much you know increase its level of bailout without affecting its cost of fund then the the first the the best thing you can do is to provide a safety flow so you say you know if losses are moderate the banks should bear the losses but there would be a threshold there would be an excess level of losses that we will not tolerate so you put a flow on the equity so formally the total bank return plus the money from the government it's going to be stabilized at some level n zero of course that's only if you have a constant cost of fund if instead you have a very steep marginal cost of fund so when you start bailing out the rest of the market thinks the government might not be solvent and the cost of fund grows up so that would be more relevant for say a country like Greece then you provide a partial bailout to partially stabilize the system okay so just to show you that we get exactly like very simple and intuitive predictions okay so in the welfare loss function that we have again is very simple which is if if the aggregate bank capitalization is is negative or is below what it should be so if the return is such that post transfer you're below some kind of cutoff then there is an externity that kicks in okay and then the optimal m is to provide a bailout of that that shape okay all right so that's kind of the the baseline framework now this is the framework people have used we discussed many papers in the literature review this is pretty much a framework that many many people have used to think about more hazard because obviously the banks anticipate that they will be bailed out and so they know that they don't need to ensure themselves very much against these really bad states because in these bad states they will be bailed out at least to some extent so that's the fundamental reason why there is more hazard and we show that you know benchmark model we get exactly the same result but the thing that's important here is the qualifier in all equilibrium with symmetric bailouts okay as it turns out the literature has always assumed symmetric bailout which is why they get their results and we're going to break that assumption and show you that everything changes so formally we get all the standard results okay so first strategic complementities that's kind of intuitive so if the other banks in the system misbehave your incentive to misbehave goes up why well because if they don't invest in safety then a big crisis is more likely in a big crisis everybody is going to get a transfer from the government so your own incentives to be safe go down okay so this this is more hazard at the bank level but also more hazard in teams second point there is more systemic risk than without the government okay so why is that well because of the anticipation of the bailout makes the banks less safe and because of that that's going to reduce the investment in the end we're going to have more systemic decrease than if you just competition down the government and third prediction which is also the one that people have emphasized is the deeper the pocket of the government exposed the bigger is the more hazard gain intuitive so if gamma decrease the cost of another government decreases there will be the government will do more bailout exposed therefore there's going to create more more hazard exxon t okay so in that in that system if you believe that's the right model of the world immediate prediction is you want to bind the hands of the regulators okay you want to decrease the degree of discretion as much as possible because if they have discretion exposed and if they have money to spend they will do too much bailout that will discord and create too much more hazard so this line of reasoning calls for very strong restrictions to limit the discretion of the regulators and I'm going to show you that none of that actually follows from the model all of that follows from the assumption of symmetric bailouts so let's do tournament no no just to be clear I'm taking exactly the same model that everybody has used I'm not changing anything to the underlying structure of the economy all I'm doing is I'm saying you don't have to do the same bailouts for everyone so let's do with two banks so we're going to start the following we're going to do so first of all we're going to be exposed efficient so time consistent okay the government is not contemplating the idea of letting the house burn at all it's very clear to everyone that no matter what happens we will save the system but we don't have to save it in and save everybody at the same time so formally here we give a bonus delta to whichever banks happens to do better in the crisis it could be that they both do quite poorly but one of them is likely to do better than the other okay in case of a tie we just break it it's very amazing okay but exposed even in a crisis where the average bank does poorly some banks are going to do slightly better and some are going to do worse okay so this is saying a bank that does better is going to get a bonus a bank that does worse is going to get the opposite okay so you can think of the previous model and therefore the previous literature assuming delta must be zero and here we're just relaxing that so with two banks in that it's easy to show that so with n equal to banks there exists a unique delta star so wage between the good and bad performing banks that would implement the social optimum and that is not only of course exposed it's efficient but that's by construction because I'm forcing my government to do the exposed uh efficient thing okay that's the root of time consistency so I will do x star total ballot exposed sorry I will do m star exposed I will do the optimal exposed ballot m which depends on the on the state of the world but more importantly I will get the right safety investment by the banks x anti in fact I will get exactly the first best safety investment okay um so this is just for two banks you can do the same with any number of banks in fact it's if anything is slightly easier because you have a finer grid when you have many banks and then you just this you just decide whether you're above or below the median and uh you cannot you can do it with bank that are different sizes okay so here we assume the banks are the same size to make the the world simple you can do a distribution of bank size and it works of course as long as it's not too asymmetric okay if you only have one extremely large bank that has 99% of the system and everybody has as one person then de facto you're in a one bank world and that doesn't work okay um but in the benchmark model just freeing up this uh delta degrees of freedom is enough to bring you back uh to the to first best and no more hazard okay so how come well remember that um we are constrained exposed to do an efficient uh transfer so that means that the total size of the bailout is fixed because it it's the one that stabilizes the economy um conditional on the cost for the taxpayer so on average each bank is going to get half of it in this example okay but you know it doesn't say you have to give half to everybody so what we are doing is we are giving half plus a bonus to the bank that does well in the crisis and half minus a bonus minus or with a minus to the one that does poorly now what does that mean for the banks it means that you have an incentive you as a bank to be sure that you're not the worst performing one in the crisis okay so you look at the other bank you think if it both go into trouble um one of us is going to be saved and with a bonus the one is going to be punished i don't want to be the one who ends up being last in the crisis okay so that's going to increase my incentive to be a bit safer so i can doesn't that does not necessarily push me to towards the first best it just stays that i should be i try to be a bit safer than my neighbor because if i'm a bit safer than my neighbor i will get the bonus but of course the neighbor is not stupid they understand that they do exactly the same and that creates a rest to the top where if you calibrate delta properly you get at the first best level of safety so this is just one example but in the interest of time and since we have a discussion maybe maybe i will skip that part and we can leave it for the discussion okay so in the remaining five or ten minutes i just want to give you some some caveats because of course you know like this is there is still a bit of simplification in that model so i want to show you the caveats in terms of what are the critical assumptions and you know when is it that always holds or when is it that it holds only approximately so first issue of course limited punishment all right so if you go back to that figure here the problem is that if you take if you take a malice away from a bank it could be the case that you are net and taking away money from the bank in a crisis state and maybe you don't want to do that maybe we're going to say that well the worst you can do is not give anything to a bank but it's not credible to say that you're going to take away some of their funding so if what happens if you have limited punishment so the transfer has to be weekly positive so of course that decreases the set of equilibrium that you you can implement but what's super interesting here is that it totally reverses let me go back to this remember this prediction of the standard model number three here that more hazard is worse when the government has more fund that actually is flipped okay and that's flipped even if it was limited punishment so formally the maximum implementable safety is decreasing in the cost of public fund okay so if you have more public fund exposed you can implement more safety which is the opposite of the more hazard result and why is that well because you have more money to play with exposed and so you're less constrained by the limited punishment constraint which is that m must be positive and therefore you have more actually you have more rather than less ways to create the right incentive okay so essentially if the government is rich exposed then the government will be able to sell at least five banks okay imagine you have a war with seven banks the government is rich enough exposed you know that it will be able to save five out of seven okay well you really want to be one of these five then okay um and so the fact that the government has these fiscal slack exposed actually is good for incentive provided that you use this tournament approach so in fact here we have a complementarity between fiscal capacity and incentives and by the way I think that's also important because in the real world I don't think the the the other prediction I don't think holds like we don't necessarily see more more hazard in countries that have more fiscal slack okay and it's not because they don't do any better it's because they spread the better um another thing that comes immediately from from this limited punishment is that if you hit the constraint that m must be positive then it gives you a rationale to try to increase the punishment beyond that and so any kind of clawback would actually be useful in that in that world okay so clawback causes compensation what's interesting here is that they become systematically important not because you care about the nitty gritty of compensation the bank the central bank in the regular they don't give a damn about that but they understand that if it limits the the incentives exposed then clawbacks are useful because they relax the limited punishment constraint so that uh first thing I want to emphasize second is differentiated bank and what's interesting here is that this notion of substitution kicks in relatively naturally so the key of the welfare function that comes from the paper I wrote with virile and matz and las or lasso on systemic risk was this idea that what really matters is the aggregate health of the banking system now I think it's a very natural assumption but hidden behind is a quite strong assumption which is perfect sustainability across banks that's what gives the tournament so much power because the size of the bailout is fixed by efficiency exposed but if the banks are perfect substitute you have almost infinite leeway in allocating the funds across banks okay and that does not alter efficiency because we assume that what matters is the total equity in the banking system so exposed one dollar of equity is a perfect substitute across all banks now is that a good assumption or not I think it depends on the model of course it depends on the application but you could imagine a world in which the banks are localized either they are localized in terms of the product space so they do they do sometimes activities where they are the only one or they are localized in terms of geography they serve of demographics of firms or household they serve a couple of a pool of customers that are only served by this bank so in this case you there is a sense in which you would like to spread a bit across all the banks it would be costly to just kill some of them okay so that breaks the perfect substitution um yeah next one two minutes this minutes okay perfect um and so what happened in this case well many things happen in this case but um um you know so the first of all is we get you know the benchmark result that you know you can still implement useful policies yes it's true and it's limited by the degree of substitution so in fact they would define a slightly different version of epsilon commitment so when the runner can commit a tiny bit just to get the continuity and we can show that you can implement degrees of safety and the upper bound is given by the the substitution elasticity across banks okay so if you're worried about more hazard then that means that this comes in and this constraint is binding then you have a strong incentive to increase the degree of substitution that means what that means making sure that at least two banks are active in each of the markets you care about um so it's a rationale for redundancy regulation of uh low substitution activities and it gives you a rationale for if you really can't do it then that's when you hit the limit and you should use a more like a utility approach to regulation um two minutes so I think let me discuss maybe two things we can also go back to um to that in the discussion but I just want to flag a few things so we have a discussion in the paper about you know can we reinterpret what happened in 2009 in the light of the mullet can we say which part of the bailout slash crisis response looked like efficient from the perspective of the mail which funds do not and um yeah we can and and it turns out that I think it's pretty it's pretty insightful the one thing I want to emphasize is the of course what happened in 2009 was not what the model was saying the key the key thing in the model is everybody understand exactly that these are the rules of the game okay and so if you want to think in your mind what is the telltale sign that this is correctly implemented well the day the government decides to let imagine there are 10 banks the day the government decides to let two banks fail the stock price of all the other banks should go up okay if that happens then you are in the equilibrium world describing because everybody understand that for incentive reason you have to kill two you're going to kill the two that perform the best signaling that price they didn't do the right thing the other eight are going to be saved for systematic reason so you kill two former hazards you said eight for systemic reason so as soon as you announce which ones are the ones are going to fail the other eight are revealed to be in the safe camp their price will go up okay of course that's not what we saw in 2009 and the reason is because the game was not explained properly it was not even played properly by the government so but that's all view of the world and I think you know I think you've heard everything so let's in the interest of time we stop here and let's give the floor to michela thank you very much tomah for this inspiring talk then we move to michela's discussion please the floor is yours you can upload your slides thank you so much for asking me to discuss the paper let the worst one fail a credible solution to the too big to fail conundrum this was a very interesting paper I really enjoyed reading it and the basic premise tomah explained really well in this paper is the principle when the government ends up bailing out the bad banks then they end up taking more risk ex ante so we have this moral hazard problem and now tomah and its course as idea or yes idea was to basically not bail out all banks or all banks that failed but instead to have a rule to reward good bank behavior and punish bad bank behavior and if I can credibly commit to that rule ex ante then the banks are incentivized to behave well instead of badly and that may overcome the moral hazard problem so this is the idea in the paper and I thought about it for a little while I really liked the idea and I thought I focused my discussion on sort of discussing why this may not work say in in the real world and there are three things that I think matter and that we have to think about for the feasibility of of this rule to be implemented and followed through and the first one and tomah mentioned that in his last slide is this idea that good banks may say no and that is just right because they do not want to take donations from the government if they don't need to but we need to commit to that sort of rule in order to change the banks behavior ex ante so it's essential for us that the good banks take the money and get rewarded for the good behavior and there's one thing which is stigma but then there's other things that is just if I force a bank to take a loan it basically also just reduces their their P&L if you if you will so there's something beyond stigma and then I wanted to think a little bit about this idea that it is not the bank is not one entity that or it's not the CEO that is making all the decisions but ultimately the people that are taking too much risk individuals within the bank and there may be agency problems within the bank that I think we have to think a little bit about and then the final point is sort of this idea that any policy needs to be agreeable enough for the people that live in the country so and here I want to you know think a little bit about how we would sell this to the people okay let me just there was a there the paper fits into a literature on the moral hazard too big to fail I'm not going to spend too much time here but the papers really were written and has a nice literature review the model right is pretty simple it's two periods two banks and then the government commits to a policy rule in period zero and then banks can choose to do the right thing and this is called x in the model it's an x anti-safe investment and so x has the feature of decreasing bank return in normal times but increasing their returns in crisis times so exit is a bad thing to do so to say in normal times but it's a good thing to do in crisis times and then there's like an idiosyncratic bank shock and an aggregate shock we can either be in normal times or crisis time and then in period two banks survive fail or bail out and so in principle increasing x the good thing the x anti-safe investment is not in the bank's interest because it is basically helping them in crisis times but in crisis times they may be bailed out so what's the point and now what we do is we basically have a government rule that rewards doing more of x which is the good thing and this basically just overcomes the morally hazardous incentive to lower x anti so this is how the the model works and now I just want to kind of comment on what what I think are the next steps to think about why we may not be able to implement this in the real world so and the first point and to my mention that is that good banks may not be going to be forced to take a loan when they are healthy and here you can think a little bit about JP Morgan was a healthy bank after the 2008 financial crisis and they were forced to buy uh bas-johns or you know the same happened with Bank of America so banks may not want to do that because it is there's some stigma associated with this that Thomas mentioned but then also simply because it lowers their P&R right if I like add an asset to my balance sheet that is not as good as the other assets I hold then this makes my balance sheet look worse so and the problem is that the moment that the good banks say no to the money we end up again only bailing out the bad banks so the whole system breaks down and now I want to think a little bit about ultimately who decides all of this like who's the one taking money from the government and then also who are the people that are inside the bank that are taking too much risk and I want to think about this in the context of what's currently happening with Credit Suisse so there was a lot of attention to Credit Suisse balance sheet these days right Forbes is titling as Credit Suisse going bust people are talking about Credit Suisse being the new Lehman brothers um what happened to in Credit Suisse basically is that there was this like family office hedge fund uh Arcadas and Arcadas making it was making a big bet and Credit Suisse was providing was their prime broker and providing the a lot of leverage to this hedge fund because Credit Suisse the people that were making those decisions were profiting from Arcadas endeavors right because they get the margin interest they get the fees so they are getting big bonuses because of the deals that they were making and now what happened basically is that the credit they had massive losses right that is then when people start paying attention when I think when things are going well and people are making a lot of money then I guess it's easy to close an eye on what's actually happening but the moment that then there are massive losses that is when everybody is like paying attention of course so now what happened is and I want to point this out I just read this somewhere that you know some people some bankers are now resigning um so this person resigned to join HSBC um and then there's so this article is basically talking about okay bankers have now the option to maybe leave before things go bust right or they wait until their year and bonuses and this is what people inside the bank think about right they think about their horizon is basically until December when they get their bonuses so and this is I think the problem that we have to think a little bit more about that the bank is not an entity um deciding what they do or don't do but rather they are the people within the banks that are taking the risk right that is the risk manager the prime broker unit at Credit Suisse and the risk management unit that let arcadeers make the bats and then only when they had massive losses um people were paying attention to it so the individuals within the bank are making decisions and their incentives may not be the same as the ones of the CEO or the bank shareholders so uh in fact the prime broker team at Credit Suisse or the risk management team they were sort of you know profiting and what they what they are concerned about are their year and bonuses um and you know now like things went bust so um you know they got fired but then they just may move on to the next bank HSBC which then right might be the next Credit Suisse and who were they coming from I don't know maybe they came from from Riemann processes they pure speculation at this point um so this is the first main comment I think that we have to think a little bit about when that the incentives for the whole bank are not the same as the people within the bank and then the second bank and this is nicely summarized in this graph trust me I'm a banker so you know ultimately we also when we implement policies we have to think about how they are perceived by the people and the question is whether we could convince the public that rewarding good banks is the right thing to do um or having this uh government policy rule and I just want to cite a few statistics that I came across so 77% of Americans believe that bankers would harm consumers if they thought they would they could make a lot of money doing so 65% say that any bank and financial institution should be allowed to fail 64% think Wall Street bankers get paid huge amount of money for essentially tricking people and 72% of Americans don't believe that new regulations on Wall Street think Frank um I actually helpful in preventing future financial crises so we also have to think a little bit about how the policy rule so to say can be sold to people to make them agree to the plan so these are these are my my three main comments so I really enjoyed reading the paper I think it's a very interesting new perspective it's a very interesting theoretical twist that allows us to get around the bank bailout moral hazard problem I think it's a very timely topic given that Credit Suisse may be the next Lehman brothers and I think there are three things that we need to think about one is the good banks have to say yes to this the second one is the banks are made of people that take on the take risks and follow incentives and then the people the public I mean need to be on board with the central bank policy so that is one other thing like how can we add this to and I'm going to stop right here thank you so much thank you very much for this very interesting discussion um as I said earlier uh also the audience is invited to submit questions via the Q&A button in the Webex I would invite you to do so we can take a couple of questions but maybe in the meantime I would ask Tomas if you would like to immediately react to Michela's discussion yeah well first of all Michela thank you very much for the for the discussion and I'm going to take our three points in reverse order so the third one is um is it particularly feasible to help the good banks in bad times I agree to me that's the that's the most tricky question uh because you need to convince the public that uh you are not that you are not helping all the banks in a way which is not discriminated and that's somehow the ones you are helping deserve it and I agree that's a tricky question I think it depends on the credibility and how you explain what you're doing um but to me that's the that's probably the part that I find the most difficult um agency inside banks yeah I agree so it could it could work both ways I mean it's like um I think it's important to take into account the agency inside the banks whether it makes it easier or worse to implement exposed I think that depends a lot on the model in fact you in some of these cases I would argue you could find that some of the employees could have been natural allies of the regulators and not even the result the result blowing is one dimension but the other like the inside of the CEO might be much more on the on the loading on systemic risk side than the one the employees would prefer so um I think that um I think we do we should take into account the agency inside banks but I'm not sure it would actually make our job harder in this case and on the first one uh on the fact that the good banks may not take the money so that I think is very important I think on that one we do have a complete solution so the first point is that remember the reason why we have the stigma and everything is because most of the government interactions where we are constructed under the view that you are helping the banks that are in trouble okay but in my world is the opposite and the second thing is and that gives me the chance to mention one thing I did not during the talk for like a time which is we do a lot of the implementation of the incentive using acquisitions and there is one thing that's for sure is the bank never refused loans and asset guarantees to make an acquisition so you know JP Morgan didn't want the equity injection for the government where they were very happy to get the subsidized loans and very happy to get the subsidies to take over or verston in the form of guarantees on the on the asset side or whatever buying so the banks never refuse the asset guarantees they really love it even the good ones and of course it makes sense because there it has the right shape for them for for their point and so in if you think about the acquisition game what we're saying is suppose you again suppose you have your four banks and then expose two are kind of mediocre one is doing very poorly one is doing relatively well in the crisis what can you do well what you can do is you make the best performing bank acquire the worst performing bank okay and then the more punitive the terms of this acquisition the better it is it's it's a double win because if the terms are punitive then that means the bank performing valley is going to be punished that's exactly what we want for more hazard exxon if the terms are punitive for the bad bank by definition they are attractive for the good bank if and and by the way that's the point where if they are not attractive enough the government can provide an asset guarantee for instance to make it even more attractive for the good bank and the keys that does not create more hazard because you only do that for the bank that performs well exposed so in fact in that case this acquisition solve all the problem and reduce the small hazard so one of the thing we discussed is the fact that the one place where I would be happy to have the regulators have a bit more discussion of power is you know to inflict terms that are pretty punitive to the shareholders of failing banks I think that's good also against small hazard and on that one I have no doubt that the good banks would be very happy to be the subsidy thank you very much we have one question from the floor from Carlo Alta Villa let me read it for everyone I hope Tomar can see it on the screen so the question is do I understand correctly that the distribution of bank bailouts matters for shaping incentives whereas the government cares only about the aggregate capital of the banking sector and if yes what happens if this assumption is relaxed and also when calibrating the price does the government know the distribution of shocks and then maybe let me take a second question which came just from Oresta Chastani a major concern was letting a big bank fail is contagion can you elaborate on the model yeah great question so yeah we discussed the we discussed the information requirement for the government so it's not as bad because I essentially what the government does expose is relative performance evaluation and that turns out to be a lot more predictable than so like we know in the data it's very hard to predict a systemic crisis in the aggregate but predicting who's going to do relatively better is a lot easier actually and so that's something we show in the paper with Viral Lasser and Elmat so and the key for the government you know the key for incentive in our world is the relative performance evaluation which is the part which is a bit less expensive in terms of information costs but otherwise yeah so the first question is yes the benchmark model assumes that the government cares about the aggregate distribution and then there is a distribution across banks okay imagine the go back to my suppose we have four banks and exposed two are okay but mediocre one is doing very poorly what is doing great so I'm going to take the the the best performing one and I'm going to make it acquire the the list the the one doing poorly and I don't care that the first one disappears I don't care that it's being acquired by the first bank because I'm assuming that the activities of these banks are quite substituted okay so that's the key assumption you could imagine cases where that would be a stretch okay if the bank that is being acquired there's something very unique that is hard to transfer to another bank that would limit this activity okay so this this is the key that that's where you hear the constraint about the substitution of activities um the contagion uh well so again so contagion depends on many factors okay in in the key normal there is that you explain that the game you're playing if the if the market understands what the government is doing the key the key prediction is as soon as you announce that these are the banks that are not going to be supported or punished and these are the ones that are going to be supported the when you fail the first bank the first the stock price of the other one goes up no down okay so that's the key difference with contagion because the market never thinks oh this is the beginning of failing all of them this is no this is oh the government has has looked at the data and this and figured out which ones are the worst performing that's going to deal with these ones and by definition I mean the flip side of it is all the other ones are going to be saved so in that sense it does not it's anti contagion now one caveat here is the timing okay and so of course in the real world it's not too period prices take take a bit of time and so the the risk is more before when you know there's something going that's that's going wrong but you don't know exactly who is more exposed that's the tricky part yeah but that would be true in any model that's not particular to us okay so thank you very much again to tomah and and to michael for the excellent and discussion