 I would move to the second paper. And so here we have Thomas Eisenbach, who's an advisor at the New York Fed, who will present a paper on the fragility of safe asset markets. And this will be followed by a discussion from Cyril Monet, who used to work at the ECB previously and who is now a professor at the University of Bern. So, Thomas, if you are online, I immediately give the floor to you. Thank you. Fantastic. I hope you can see my slides and hear me. Please interject otherwise. All right, so this is work jointly with Greg Feynland, who is at the Office of Financial Research and Williams College. So we have an unusually large disclaimer here disclaiming that these are not the views of any of the institutions we're working with, except perhaps Williams College, although I doubt that Williams College has strong views on safe asset markets. All right, so what do we do in this paper? I think safe asset markets and their fragility have come up a number of times. And we think it's a very interesting topic. Specifically, what got us started thinking about this were the events of March 2020 in the US Treasury markets. So the first piece of motivation here is when in March 2020, the traditional flight to safety turned into what has been called the dash for cash. So if we revisit briefly what happened in March 2020 or in early 2020, as news about the COVID pandemic started spreading, we saw a very standard phenomenon where safe asset prices, so I'm plotting here the Treasury yield on a reverse scale. So this is Treasury prices going up. Safe assets appreciate, standard flight to quality, risky assets here, this is the S&P 500, risky assets fall in price. So you see the standard negative correlation in times of stress. And then the period we're really interested in is in this period in mid-March, which is about nine days, where suddenly as risk assets continues, there was continuously bad news about COVID, risk assets continue to go down in value. And suddenly, Treasuries turned around and basically fell off a cliff over these couple of days. So this is really a completely unprecedented event, such a large crash in Treasury prices at a point in time of severe economic stress. Now, the next piece of evidence that got us really curious is what you see when you look at dealer balance sheets. In particular here, we're showing an orange Treasuries on dealer balance sheets, both their outright holdings and their reverse repo positions against Treasuries. And you can see that as throughout both the flight to qualities, the run-up in Treasury prices and the crash, dealers were loading up with Treasuries. So their inventory, their balance sheets were filling up more and more with Treasuries through the run-up and then the crash. And you can see that the turnaround in Treasury prices, the quick recovery and the normalization coincides exactly with the timing of dealer balance sheets getting relaxed again as the Fed. So this is the green line as the Fed started purchasing Treasuries and large incomes. So it really looks like the other balance sheets were playing an important role in this episode. And then the final piece of really interesting motivation for us has been the question of who was actually selling and why? And here we're relying heavily on the great paper by Annette Bissing-Yorkelson and the JME and we've adapted one of her figures that shows from the flow of funds, the three sectors that had the largest sales of Treasuries in the first quarter of 2020 which were foreign investors, mutual funds and households, which include hedge funds. So starting with households and hedge funds, you can sort of see that the series is fairly volatile and if I hadn't put the orange square on Q1 of 2020, it wouldn't have been obvious which of the quarters it is. In contrast, for both foreign investors and mutual funds, you can see that this downward spike, these sales in Q1 of 2020 are a complete outlier and entirely unprecedented to anything that had been done, that any amount of sales they had done historically. So you have these two investor groups that are just selling in completely unprecedented amounts. And then if you look at the details of what Annette reports in her paper, you'll notice that a lot of these sales look like they were precautionary in the sense that the agents or the investors selling the assets, they didn't actually need the cash immediately. So the evidence for that is that among the foreign investors, foreign officials only consumed about 25% of the treasury sales, where consume here means converted out of US dollar assets into whatever their local currency is, which is typically then because they're doing foreign exchange interventions. Similarly, among the mutual funds that are the second largest group of sellers, they only pay out about 65% of what they're selling. So only 65% of what they're selling is accounted for by withdrawals they were facing from their investors and the rest was entirely precautionary. So someone like me who spent a lot of time thinking about bank runs in the classic diamond dipfic sense, this is exactly what the late consumers, also in the diamond dipfic model, the consumers who don't actually need to consume but who run the bank look like, right? So they have huge outflows, but they don't actually consume everything that they're pulling out in contrast to the early consumers who really have a genuine need for liquidity who sell and then consume. So what do we do in this paper? So in a nutshell, we combine two key modeling ingredients. The first is that we distinguish two fundamental characteristics of safe assets. So the first one is as the name suggests their safety, which is their low credit risk, their low or potentially even negative beta. And this has been a large focus of the safe asset literature. And then the second one is their liquidity. So the fact that they are very easy to sell and money-like. And oftentimes in the literature, these two are lumped together. Some papers have distinguished them and here we're making a very explicit distinction between the two by assuming that there's two different types of investors that hold treasuries and they're holding them for these two separate reasons. So safety investors hold them for their safety and liquidity investors hold them for their revenue. And then we combine these assumptions about safe assets with constraints on dealers who are the main intermediaries in the safe asset markets. But you can think of limits to arbitrage more generally. So the main effect that this assumption has is that when you have large net sales in a particular period, these can lead to persistent price locations in future periods. So if there's large net sales today, they push down prices today, but they also push down prices too. And then what arises in our model is a strategic interaction among the liquidity investors. So the investors who value treasuries for their liquidity. And they have a very important choice. So they have a choice whether to sell preemptively today or risk having to sell at a future date. I'm just always going to talk about tomorrow. And so the result is that we get fragility in this market. We can get market runs in times of stress in the spirit of the original market run papers from 2004, this Bernardo-Welsch paper and the Morrison-Chin paper. And then what's really special in our paper is that we show there's an interaction of this strategic element among the liquidity investors with the flight to safety demand from the safety investors. And in particular, we can show that the flight to safety can actually trigger the dash for cash if a particular set of circumstances is satisfied. So let me preview the results a little bit. So as a reminder, usually investors in financial markets face strategic substitutability. So this means if other investors are selling and so the price is going down, then I want to buy all else equal, right? So this is strategic substitutability. Everybody else is selling, that makes me want to buy. In our paper, what's important is we show investors can face strategic complementarity. And to get this, we assume that there's investors who hold the safe asset as insurance against liquidity shocks. So very similar to the diamond dip agents who receive liquidity shocks in the sense that they suddenly need to consume. And then what happens is that as other investors are selling, the prices decrease today and tomorrow through this dealer inventory effect. And that can make me as the individual investor want to sell today. And the logic is that I'm trying to get out today rather than risk being forced to sell tomorrow at a worse price. And so this strategic complementarity, as is typical in games theory, generates self-fulfilling equilibrium where we can have a hold equilibrium where everybody holds because everybody holds and that's an equilibrium. And also a sell equilibrium where everyone sells because everyone sells. We then use standard global game techniques to get a unique equilibrium that is going to be a threshold equilibrium. And so then the fundamental that determines what happens in this market is the degree of liquidity risks that these agents are facing, which is going to be the probability of suffering the liquidity shock. If this probability is low, then the market is going to be stable and the only people selling are those with genuine liquidity needs, the only fundamental sales. However, if the probability of liquidity shocks, if the liquidity risk is high, then the market collapses. The equilibrium switches to everyone's selling and the market is flooded with panic sales. People are selling, even though they don't have genuine liquidity needs. As a result of this global game structure, we get a discontinuous equilibrium price. So you can observe prices that suddenly drop by large amounts when the equilibrium switches. We also get the effect that all the announcements, even just announcements about future purchases can have large effects on price today by switching from the sell equilibrium to the hold equilibrium. And then we show that increases in dealer balance sheet costs reduce both the market stability through a lower threshold and increase the price discontinuity. So if dealers face steeper balance sheet costs, the market is less stable and the crashes are going to be bigger when they happen. And then the final element of our paper is this interaction with safety investors. So you might think that while in times of stress where a lot of the agents holding treasuries for liquidity reasons are selling, there's also going to be the standard flight to quality with a lot of people who are holding them for diversification reasons rebalancing into treasuries. And doesn't this naturally offset the flows and lead to sort of a virtuous interaction? And the answer here is it depends. And the key element is that the flight to safety has an effect on the prices today and also on prices tomorrow through dealer inventory. And so what we show is that you get a feedback effect and the direction of the feedback depends on the inherent stability or instability of the market. If the market starts out relatively stable, then the safety investor demand is further stabilizing. So then it's good, it's a virtual cycle. But if the market is relatively unstable to begin with, then the safety investor demand turns out to actually be destabilizing. So in that sense, the flight to safety can then be the trigger for the dash for catch. All right, so some details. The model, very simple two-period model. Two assets, a risk and a safe asset. The safe one is going to be a focus. Three types of agents, safety investors who are risk averse and they hold a portfolio of a risky and a safe asset. And so they rebalance between the risk and the safe asset and they will be the driver of a standard flight to safety element. Second, the liquidity investors, they are risk-neutral, but face liquidity shocks like in dynamic. So they hold the safe asset as insurance so that they have something to sell when they need liquidity. And then third, dealers who are risk-neutral but face balance sheet costs. So they provide a residual demand for the safe assets. All right, so dealers are very simple. They value the safe asset as a fundamental value. They have conduit balance sheet costs for inventory. They are competitive. So the demand is going to be given by a zero-profit condition and that gives us very simple linear prices, linear and total sales and inventory price today as a function of total sales today. And then the key thing is that prices tomorrow and period one depend on sales tomorrow and period one but also sales today through this inventory. So sales today can affect prices tomorrow through inventory. The liquidity investors, they are endowed at the beginning with one unit of a safe asset and they face these IID liquidity shocks with probability S. So S is going to be a really important parameter. It's the degree of liquidity risk faced by these liquidity investors overall. And the focus of the analysis here are the investors at T equals zero that did not receive a shock, okay? And so those investors are going to act strategically. They know they did not receive a shock today. They can still sell for preemptively reasons if they want to sell, in which case, they get an expected payoff P zero E. But if they hold onto their asset, then tomorrow at period one, they may get a shock in which case they're going to be forced to sell at P one E or they may get lucky and not receive a liquidity shock. And in that case, they get a continuation value V and that's sort of the best possible option. Okay, so the key thing that's going to be driving the equilibrium here is the payoff gain, the difference between these two expected payoffs that captures the incentive to sell given a fraction alpha of others selling. What does this payoff gain look like and how does it affect equilibrium? So an important thing to note is that the payoff gain doesn't depend just on alpha. It also depends on S, the liquidity risk. It modulates how much weight is put on today's price versus tomorrow's price. And so the key thing here you can see is when the quitted risk is very low. So this blue line, then the payoff gain is uniformly negative and it's actually decreasing in other sales. Uniformly negative means even if everybody else is selling, I do not wanna sell. So the only equilibrium is nobody's selling. Vice versa, if the quitted risk is really high, the red line, then it's uniformly positive. Everybody's selling and I wanna sell. And in this intermediate region, we get multiplicity where if everybody is holding on, then I wanna hold on as well. And if everyone is selling, I wanna sell as well. Using global game techniques, we can derive a unique equilibrium here that's in switching strategies around a threshold. So for low liquidity risk, all investors are holding for high liquidity risk above the threshold all of selling. And you can see that the threshold is a proxy for market stability because as the threshold increases and gets larger, the size of the hold region where the whole equilibrium prevails, that's larger. So a market with a higher S star is more stable than a market with a lower S stone. What do equilibrium prices look like? So first, here I'm plotting the equilibrium price as a function of this degree of liquidity risk. And you can see, let's focus on the pool line first that as liquidity risk increases, the price slowly decreases. That's because there's fundamental sales, right? There's just more people, more investors getting a genuine liquidity shock and needing to sell. But then suddenly as the equilibrium switches, the price drops off a cliff and then stays constant. And then the second element here is as you increase dealer balance sheet costs, you're moving the threshold to the left. So remember that means lower market stability, the price crash happens earlier. And in addition, you can see the price crash is gonna be much larger. So there's an interaction between the threshold and the balance sheet costs that generates this large price crash. All right, so what's the role of safety investors? Safety investors is very simple in our model. At a risk averse, they hold a portfolio of the same safe asset and a risky asset with some expected payoff mu. So we're going to think of a bad state of the world for them is when there is news, bad news about the expected payoff mu. And in that case, they are going to buy more of the safe asset, which is going to be parameterized by the parameter A. This flight to safety demand has two effects. It increases the price to date directly because it offsets some of the sales. So that is destabilizing from the point of view of a strategic liquidity investor is if I actually get a better price today, that makes me more likely to sell preemptively today. But the flight to safety demand also increases prices tomorrow because it reduces dealer inventory. So then dealers are willing to offer better prices tomorrow and that is actually stabilized, right? Again, me as an individual, the liquidity investor, if I know tomorrow's price is better, then I don't have to worry so much about being hit by the footage shop tomorrow. So then I'm less likely to sell today. Now the question, which of these two effects that destabilize and the stabilizing effect which of the wins out is a question of how big is liquidity risk, right? So in the payoff gain that captures the intent of the sell, price to day and price tomorrow both show up, but price tomorrow is discounted by the likelihood that I'm facing the price tomorrow. So if liquidity risk is low, if there's a low likelihood that I'll be facing this price tomorrow, then the destabilizing effect of P0 dominates. If liquidity risk is high, then the stabilizing effect dominates. So you can see that there's this interaction. But how exactly does this play out? So if you think of a market, if you vary the balance sheet costs underlying the market, and so first here I'm showing you a situation with low balance sheet costs, like we think prevailed before the great financial crisis, pre-2008. So low balance sheet costs mean you're in an inherently stable market. The threshold is very high. It takes a very high shock, the equity risk shock for a run to appear. And now flight to safety going from AL to AH is stabilizing. It actually moves up the threshold even further. So in this situation, you have a virtual cycle, flight to safety coming in exactly at the same time that the equity investors need their equity is stabilizing and you're very unlikely to observe this switch in equilibrium and this market run. In contrast, if you're in a situation with high balance sheet costs, so this is what we think represents the post-2008 environment, the March 2020 environment, you have high balance sheet costs. So you're starting out at already very low stability. And in this case now, the flight to safety is actually destabilizing and can trigger the dashboard change. All right, let me quickly show you two slides, two or three slides on policy implications. So the first one is, are really these dealer constraints relevant? And so here I'm again showing you treasury prices and I'm showing you two elements of the Federal Reserve's intervention. First, repos they were offering. So repos against treasuries for dealers, which importantly do not help with the dealer's balance sheet constraints. And you can see that these repos were going up, but they were well below what the Federal Reserve was offering and they didn't help at all with stopping the price crash. In contrast, once the Federal Reserve switches to buying things, demand for repo just vanishes. So you can see that the repo is really not what was helping anyone and the purchases are what's helping dealers. The second thing is asset purchases. So here, what's interesting, if you look at the details, the Federal Reserve started purchasing assets on this Friday. And on that day, it was buying half the amount it had planned to buy over the entire following 30 days. So from the point of view of a market participant, this looked a lot like the flight to safety demand in our model. Someone coming in buying today, but most likely not buying tomorrow. And then only later the next week is when the FOMC switched its guidance and said the open market desk should buy as much as necessary. So they switched to this classic, whatever it takes language, and then the equilibrium switches. And you can see the sales here, the foreign official sales, they suddenly go back to normal after the city was moved. All right, so to conclude, we're showing here how safe assets because they're held for different reasons, safety and liquidity generally provide a subbiotic relationship and markets are generally stable, but you can get this fragility from the strategic interaction between the equity investors. This is going to be worse when dealers face tighter constraints. And in that situation, the strategic interaction is amplified by flight to safety. And so we think this captures very well what happened in March, 2020 where there was a perfect storm of low market depth because of high deal advance sheet costs and unusually large liquidity shock and an unusually large shock to risk assets. And so in that case, our model as our model would predict the flight to safety turned into a dash from catch. Thanks. Thank you very much, Thomas for this insightful talk. And I would immediately then hand over to Cyril for his discussion. Okay, I was trying to unmute myself. Hi to everybody. Thanks a lot for inviting me to discuss this paper. Thanks, Thomas, for his great presentation. So this is a very nice paper. I really like it first off, but because I really like it, I'm going to not be nice for once. So what is the topic of this paper? What is the question? It's essentially how could the market for safe asset and the safe asset here are US treasuries for God's sake. How can they like liquidity? How can they dry out? And that's essentially the question that the authors are after. And the idea here of the story is that it's a combination of a liquidity shock in the sense that people, I mean, investors wanted to sell their assets, a combination of dealers not willing to actually make market because of balance sheet costs and also a demand for safe assets. So you hear this story. I mean, you read this story in the press. As I do now, you sort of wave your hands, but you want to make the story consistent to know what's really going on. And this is, I think, the big, big contribution of this paper. And then the question is, what are the policy implications? So I'm going to present you with a simpler setup than what the term has presented. And it's going to be a setup with cash in market pricing. And it sort of think of it as balance sheet asset pricing. Balance sheet in market pricing. So this means that the prices are going to be determined by the balance sheet space on the dealer's balance sheet. So the dealers, they are going to have M units of cash in pure zero and one in my setup. And M here is going to be fixed. You can consider different variations of the same thing that wouldn't matter. And dealers, on top of the liquidity investors, the dealers will also face liquidity shock in this setup. And so at T equal one, they might have to sell some of their assets. And this lambda here is going to denote their liquidity shock. If they don't have... So dealers don't have to buy the assets here. And if they don't buy an asset, they can always save their cash at a gross return of row. This is not going to be important, but just to make the story complete. Let me skip that. This is essentially the incentive for dealers to buy the assets in pure zero and they will always have this incentive. So here is the timeline, essentially. And this is what happens to the liquidity investors. So remember the liquidity investors, they hold one unit of the asset here at T equal zero. And with royalty S, they will face a sales shock. That's where S comes from, you have to sale. And if you sale your asset, you get a price of P zero and P one zero. So of course, this price is endogenous and we'll have to work a little bit to find it. If you don't have to sell your asset, you can hold on to it to P and T equal one. And then here again, you might face a sales shock. Okay, so with royalty S, you might have to sell your asset and then the price here is going to be P one. Now, maybe you hold on to your asset. And if you hold on to your asset, then you get the fundamental value V out of that and V is pretty large. So you don't want to sell the asset in that case. Okay, so this is the liquidity investors. The dealers are going to intermediate, okay? They are going to make the market here in the sense that the dealers are going to use the cash they have M in order to purchase some assets. So given the price is P zero, the amount of assets they purchase is M over P zero. These dealers, they hold on to these assets into a period one. And then maybe they face a liquidity shock of their own and then they would sell these assets that they just bought. If they don't have this liquidity shock, then they are going to use, they are going to hold on to the assets they bought and they are going to use the cash they have to buy some more assets at time T equal one. Okay, so that's the basic setup here. And then there is this possibility of a run. Okay, so this is what Thomas presented. And so essentially here at time T equal zero, it could be that some of the liquidity investors who don't receive a liquidity shock, so they don't have to sell their asset, they still want to sell their assets. Okay, so here this is this fraction alpha of liquidity investors who, although they don't receive a shock, they still want to sell their assets. If they don't sell their assets, you know the same thing happened, they hold on to time T equal one and so on and so forth. Okay, now how are these prices, P zero and P one, and this is where cash in the market pricing come in. And so here you have the equations that essentially pin down prices. Okay, so the price at time zero is going to be such that the nominal supply of assets. So this is what I highlight here. Okay, this is the demand for assets. I'm sorry, this is the supply of assets is equal to the nominal demand for asset. Who demands asset at this stage is the dealers. How much do they have? They have just cash M. Okay, and so that's the cash in the market pricing. So this is going to determine P zero alpha. P one alpha is going to be determined in the same way, but now notice this expression here, which essentially is the dealers who bought some assets and who receive a liquidity shock. They have to sell some of these assets at time one. Okay, who buys these assets? It's essentially the dealers who did not receive the liquidity shock. Okay, so that's why you have the one minus loan down here. So here you have these two equations and that's determined P one alpha and P zero alpha. So pretty simple. When do you have a run? It's the same condition as in Thomas' paper. You have a run whenever the price at which you can sell the asset at time equals zero is high enough relative to the expected value of holding on to your asset at time one. Okay, then you have a run. What you can find is that you have a run whenever this condition is satisfied by doing a bit of algebra. So what you get out of that is essentially that you have a run if actually the cash that dealers have is large enough. So it has to be large enough so that the price effects dominate the fundamental value effect. So what you want is that this effect here sort of dominates the fundamental value effect which is here. So you need them to be large enough. What is sort of what I was, I thought working on that, I thought I would be able to get a run just out of the cash in the market pricing or just out of the dealer's balance sheet. And actually I can't. The only way that you can get a run is if LONDA here is, which is a liquidity shock that dealers receive if this LONDA is high enough. So here you need that the liquidity need of dealers have to be sufficiently strong at time key, call one, to sort of depreciate, to lower the price at time one, to trigger the run. Okay. Then, so Thomas adds these safety investors and I think this is the novelty in this paper. So the safety investors, what they will do is that they will increase the demand for the safe asset at time T equal one, okay, which implies here, sorry, it's time T equals zero, it should be. So essentially here you're gonna have that the safety investors, they will ask for more assets, more of these safe assets. So what is going to be the consequence of having this additional demand is going to put a pressure on the price. So it's going to increase P zero alpha, but that's gonna be destabilizing, okay. So that's this effect here. So here I could recompute the price. And so you have that having these safety investors is actually destabilizing in the sense that it induces, it increases incentive for the liquidity investors to actually sell their price early on, rather than later on. Now you have a stabilizing effect of having these safety investors, which is due to the fact that it sort of frees up the balance sheet space on the dealer's balance sheet. And then the dealers, they have more cash in order to purchase assets at time T equal one. So that's the second effect that you see here. And so this effect is gonna be stabilizing, okay. And so, you know, Thomas made a good, in this presentation, you could see how these two effects play in different directions, depending on the liquidity pressure, which I find is very nice. The policy implications is that the central bank should purchase the assets at time T equal one. The central bank is sort of this black box here. It's not clear how it should buy the asset, whether it should buy directly in the assets, whether it should open a facility, standing facility. But what is very important is that the asset purchase program should be announced at time T equal zero and should be active at time T equal one, okay. So essentially here, the takeaway from Thomas' paper is that you have to announce, so when there is no run, it has to be clear to the investors that in case there is a run, there is going to be an asset purchase program put in place. And so in that sense, it seems like the asset purchase program is sort of a standing facility, which is, so that's my point here. It seems to be, to me, it's a negative in the sense that central banks don't like to set up facility just for the sake of setting facilities. And here, this is what the central bank would have to do to prevent a run from happening. The good thing, however, is that the standing facility is only used in a run equilibrium. And because you put in a standing facility in place, there's not going to be a run in equilibrium. And so in equilibrium, it's not going to be used in this facility, and that's a positive thing. Okay, so what are my comments? My comments is that, it's very difficult to get a strategy complementarity when you are disciplined by the pricing mechanisms. Okay, so this is very hard to get. And so I praise Thomas to get this effect. Now here, you have to work very hard to get this pricing complementarity. So this pricing complementarity was in this paper by Bernardo and Welsh, on which this paper builds on. And Bernardo and Welsh, they have, again, to work a lot to get to their results. And in particular, and this is a critique of Bernardo and Welsh, this is not a critique of this paper that builds on Bernardo and Welsh, but they suffer from the same critique, is here, why is the dealer's problem static? So in Bernardo and Welsh, and Welsh, sorry, Welsh, the dealer's problem is static. And it seems to me that dealers should actually be the most rational of all investors. So they should foresee that the asset price is going to be cheaper at T equal one. And in that case, they won't actually to save versus then purchase asset at time T equals zero. So I would rather have a dynamic marginal pricing like Alla Diamond Veracca, where dealers would be rational. And one reason is that, given the pricing that Bernardo and Welsh use, dealers here might make negative profit on the last units that they purchase. And this is sort of not ringing true to my ear. So that's the main comment to Bernardo and Welsh paper. Now, it seems that another comment is that the market structure matters, or does the market structure matter? I think it does. And this is a nice graph from the BIS, where if you concentrate on the right hand graph, you see the bid-ask spread for 30-year treasury bonds. And the red line is the bid-ask spread for off-the-run treasuries, and the blue line is for on-the-run treasuries. And so you see that in March 2020, most of the liquidity actually dried out in the off-the-run market. And what is special about the off-the-run market is that the off-the-run market is mostly OTC, while the on-the-run market is mostly more like centrally traded, the on-the-run treasuries. So, and we know that the SLR affects most the OTC market. So I think that since Bernardo and Welsh, a lot of progress has been made in understanding how, theoretically speaking, how OTC markets work, and a lot of technology has been built in order to understand OTC markets. And I think it would be very nice to actually have a framework where because you trade OTC, the liquidity is gonna drive much more and much quicker than if you trade on a centralized platform. So I would encourage the author to think about that. And in particular, in their setup, it seems that there's a big missing market and it's the repo market, okay? And so how would the repo market here work? So it would be that if you are a liquidity investor and you think there's a run that is going on, then you might want to engage in the repo contract. And the repo contract would be that you sell at P0 and you repurchase at P1, okay? And we know that in a run equilibrium, you have that P0, the price at time zero is higher than the price at time one. That's why you have a run. But if you can engage in this type of repo contract, you can make a killing, okay? Why? Because again, you're a liquidity investor, but you don't need to sell at time zero. So, but you engage into this repo contract. So if you send the repo contract, you're gonna get P0. And then you need to repurchase at this price P1 alpha. So this is what you get as a profit from the repo contract. If it turns out that you have this liquidity shock at time one, well, you can sell again spot your asset because you just repurchased it. So this here is clearly positive. If you don't face a liquidity shock, well, it's great because you can keep your asset that you just repurchased and you get the fundamental value fee. So overall it seems that a repo contract would be used and it would actually prop up the demand at time one for these assets, these safe assets. So here, it's more like a question. So if you have this type of repo contract, would you actually kill the run equilibrium that you have? So final comments. So one of the competitor of Thomas Paper is He Nagol and Song, just published in your GFE where they also tackle this question of how can you get a lack of liquidity dry out in safe asset markets. And it seems to me that one of the main difference with He Nagol and Song is the role of safety investors that Thomas presented. And there I'm like, okay, but so Thomas, you have to correct me if I'm wrong but if this is your contribution, if this is the main difference with He Nagol and Song, then why do you delay the discussion to the later part of the paper? I mean, you have to bring this part much sooner in the paper. And then you have to explain what's the new insights and I think you have new insights, okay? Particularly about the timing, I mean, the stabilizing and the destabilizing part of this safety investor, I think it's great and you should really try to push that more, much more than you do. So I would, I mean, my recommendation would be sort of postpone or put the global game thing analysis to the appendix or later on, but concentrate on the safety investors. And my final comment, final, final I promise is, I mean, could you comment on Duffy's proposal? So in the paper, so the proposal to reform the treasury market. So you actually, I mean, you quote Duffy, but you just quote Duffy in some sense. So I would like to hear your views about Duffy's proposal to have central clearing of repo transactions that essentially would net all the repo transaction positions of the dealers in order to reduce their exposures and therefore free up some space on their palingines. That's it. So thanks a lot. And again, great paper. Thanks for, thanks for the opportunity to present to discuss it, sorry. Thank you very much, Cyril, for your discussion. So again, if there are questions from the floor, please raise your hand. And in the meantime, maybe if Thomas wants already to address a few of the points made by Cyril, please. Yeah, I'll try to keep it short. Thanks a lot, Cyril. Please send me your slides. Let me just make two points. So I fully agree. It's really hard to get these runs going. And I think we should be clear in the paper that it's actually very hard. You need very strong assumptions to get the runs going, which I think is okay because we just, we've only observed it once in the history of the treasury market. And yes, the key element is that the price tomorrow has to be lower than the price today because I may not have to sell tomorrow. Maybe I'm fine and I'll sale through tomorrow. So really it has to be a set of assumptions that generate a price tomorrow that's even worse than the price today. So if you had dealers who were not myopic, they would immediately equalize prices today and tomorrow and a run wouldn't happen. So I think one element we didn't mention is the fact that the Volcker rule, which is another important post-crisis regulation, takes away a lot of incentives of dealers of doing these equalizing prices over time. We saw that in September, 2019, when they didn't step into repo markets to arbitrage. And then in general, I think in terms of the facilities, the details of the regulations matter. So the SLR also affects their repo holdings. So a repo facility doesn't help the dealers, right? It really would have to be a purchase facility or you would have to exempt outright treasury holdings and the treasury repos from the SLR when we start. Thank you very much. So we have one question from the floor, please. Hi, Thomas, nice to see you. Nice presentation, very interesting paper. What I found particularly interesting is that you have this, that you strip out essentially these two functions of safe assets, namely the liquidity and the safety. And you highlight in particular the interaction between the safety and liquidity investors. And as it seems to me, the safety investors are quite mechanical, yet affect the liquidity investors. But there is not, this doesn't happen the other way around. And furthermore, kind of liquidity is endogenous, but safety of the asset is kind of something exogenous or doesn't affect the demand for this, from the safety investors, something like that. So it's completely reduced form. I think we try to keep it as simple as possible to get these linear structures that we can solve in closed form and show nice graphs. The safety investors are not affected at all by the liquidity investors. Well, no, so they are actually price sensitive. So the safety investors, their behavior results in a downward sloping demand for the safe asset. It's just that the main effect they have is that when there's bad news about the risky asset, their holding man just parallel shifts up. So the intercept is the key thing that affects things. But they're still price sensitive. So as more, as the safety investors, the liquidity investors dump all their assets and the price goes down, the safety investors actually buy more of it. And then vice versa, obviously the safety investors do affect a lot the interaction of liquidity investors. So we've tried to keep it as simple as possible, but there's still fairly rich interaction between the two. Thank you. Is there any further question at this moment? If not, then let me thank the four presenters and discussants and maybe we offer them a round of applause for a two very good presentation.