 Thank you very much for coming back on time. I I'm amazed by that discipline. So fantastic So which allows to keep time. So we have a session with a pretty dry title called banking. Okay, but But I can tell you we have some two really interesting papers that will wake you up pretty soon. Okay, so So I mean I don't take much more time. So just introducing one of my colleagues David a porquelaccio Latia who has a paper on and then just bank fragility in a macro economy He is one of our prolific contributors from the DG research with very nice interest in a research agenda on putting financial Frictions and instabilities into macro models. David. Do you have 25 minutes? So here it is So let me start by thanking so much the organizers for including our paper I'm very excited to present here and I want to also start with an apology So the paper has been growing very quickly lately And so some of the things that quite a few of the things that I'll be presenting weren't in the in the draft that we sent Ansgar, so yeah, there's a little ambush for him Yeah, so this is joint work with Kevin who's at the LSE and obviously the paper is our views It's not the views of the institution. We I work for so Okay, so what am I gonna do? I'm gonna sort of think about endogenous runs and find a way to insert them in Hopefully by the end of this talk, I'll convince you to take the friction we have and put it in your favorite DSGE model That's the idea And so what research questions can we can we ask so they're gonna be thinking about run risk? So well sort of at the micro level I can think of how does run risk affect banks behavior at a more macro level So how does run run risk affect you know macroeconomic outcomes that we all care about? So about so let me give you sort of short answers to warm up So on the on the side of bank behavior, you know banks are gonna be worried about being too fragile so they're gonna leverage they're gonna limit their leverage they're gonna demand the liquid assets and Well if they're fragile they may end up having to pay funding spreads and For macroeconomic outcomes these spreads will generate amplification to shocks So things there's a bad shock to the economy funding spreads go up that you know further reduces investment Okay, so and then you know once we have this model set up Basically, we have real effects of liquidity and we can ask ourselves So what are what are the macroeconomic effects of government supply liquid assets and in particular how should you know we can Talk about optimal policy. How should the government supply liquid assets? Well the macroeconomic effects is that you know liquid assets Once they're supplied some of them end up on cushions on banks balance sheets and this supports lending in our model And then we can ask her a question. So how and how much liquidity should be supplied? Okay, so let's start with some motivating evidence to so The first one is that yeah bank funding spreads gonna be at the heart of the mechanism of our model They fluctuate they you know I don't have to tell this audience that they are sometimes thought of as barometers of financial stability and because they don't have to explain this to this audience I'm just gonna skip over that and Something is said that is I think a little bit newer is a Correlation that we find in the data between bank funding spreads that is the library So the rate at which banks can borrow at a uncollateralized basis minus the repo rate that is a safe rate These are correlated positively with liquidity premium. So it seems like and so let me point your attention to the scale on the axis I mean that that regression line that you see that is steeper than it looks And so this this Presentation is gonna be about coming up with a theory for that and then we're gonna do some empirical exercise And this is just a correlation. We're gonna try to identify that relationship Okay, so let me just show you that the red dots those are recessions the black dots those are expansions So this relationship still seems to hold even in expansions Yeah, maybe I should tell you what's what's the story that we have in mind The story is that in periods when liquidity is expensive Banks are gonna reduce their liquidity buffers and that's gonna make The credit risk of banks seem higher to the two investors people are gonna be worried that these banks have lower liquidity buffers And so they're gonna be they're gonna demand a higher funding spread Okay, so this is just again the blue line is what I've shown you at the beginning. This is the regression line I mean that's just a correlation and At the end of the paper, I'm gonna try to persuade you that we have some exogenous liquidity shocks And so the IV is the red line So I'm gonna try to persuade you that there's actually a structural relationship between these two things and it's yeah pretty strong Okay, so what's the literature of obviously there's quite a few macro banking papers Yeah, just quickly we have a different friction. So here's we're gonna have run risk those papers. They're focused more about solvency The run risk obviously there's there's a literature in banking theory they were gonna try to do this in general equilibrium and A key outcome of the model is gonna be this demand for liquid assets demand for liquid assets You know, you want to hold a liquidity buffer just to reassure your creditors that you're not fragile There's other papers that sort of study demand for reserves and have models that give this as an outcome They just tend to you know, it's demand for reserves tends to come from payment motives or Ours comes from this run risk and dodge honestly Okay Now what's the roadmap? So the first thing I'll do I'm gonna give you one slide on micro theory and that's Took quite a bit of the of reducing it's now only in one slide and please forgive me for that Then I'm gonna I'm gonna have the friction and I'm gonna introduce it in the macro model I'll show you how to do that then We're gonna do a kind of a quantitative exercise and see you know whether this friction matters And then I'll move on to the identification to the empirical evidence Okay, so here's the micro theory, please bear with me So in each period, we're gonna have banks banks have you know, the state The key state variable is the net worth of banks here and they're gonna make two decisions They're gonna choose their liquidity ratio. So how much government paper to hold relative to their total assets? I mean the alternative to government told in government paper is to hold physical capital and I'm gonna call that the illiquid asset And they're gonna also choose how much to leverage that's gonna That's gonna end up giving you a capital ratio. So net worth over total assets and You know banks make these decisions the households on the other side there they have to decide whether they want to hold or not Deposits now I'm gonna say banks and deposits here I think the deposits need you should have in mind Uninsured deposits are short-term debt and banks maybe we can think about a wider category It's just easier to say banks instead of financial intermediaries each time Okay, so just usual diamond and divic stuff You know illiquid assets if you have to sell them on short notice There's some liquidation cost and that means that if too many households decide they don't want to hold your debt So then you go under as a bank, okay? And you'll be thinking well diamond David. There's multiple equilibria here. That's true We get multiple equilibria. So we need another friction. So we're gonna put in a coordination friction This comes from the global games. So the global games literature has taught us how to do it All you need You don't really need an information friction all I mean you people can know the fundamentals of the economy arbitrarily Well, what's key here is that you need some? Uncertainty about what information others have so you need creditors to be unsure what other people are thinking So a departure for common knowledge if you do that then you get a unique equilibrium So the unique equilibrium is nice to do policy analysis and to study the dynamics of the economy Okay, and the unique equilibrium that ends up giving boy. It boils down to this no-run condition Okay, so what this no-run condition is is telling us is that? Households will decide not to run on the bank as long as they're sufficiently remunerated for holding the debt that is the left-hand side the funding spread Jay you can think of as the library so return on on Banks short-term uncollateralized debt and brawl you can think of the risk-free rate, okay? and Sort of what does remunerated enough mean so the term to the right that is the first theta That's a parameter. It's the loss given default if you're caught up in the bank's default then you're gonna lose that much of your principal and The rest that is a term that we call bank fragility it basically functions from the point of view of households as The probability of default of the bank and you can see that the bank can do something about that the bank can Reduce its bank fragility by having more cap a higher capital ratio. That's a small n or by having a higher liquidity ratio That's a small m Okay, and so that's gonna be the key And that is the friction and now let's try to put it in our macro model And let's keep it as simple as we can so the simplest macro model that we could think of was an RBC model But this is just for illustration purposes. You can take that that Inequality constraint and put it in whichever UK engine model you prefer But yeah, let me just quickly go over what happens in an RBC model. There's gonna be these households They save in bank debt. They're gonna supply labor and they consume There will be competitive firms they rent the physical capital that the banks invest in and they're higher the labor and there's gonna be a Government that supplies liquid assets you can think of them as treasuries or reserves sort of equally and In the background, there's gonna be lump sum taxes and transfers to sort of support the the debt Okay, now out of the way the mac the rest of the macro model Let's think about how does a bank behave within this macro model. Okay, so the bank is An entity that that is gonna maximize its lifetime dividends Has an infinite horizon so present discounted value of dividends is subject to budget constraints and to that no run condition that I That I specified earlier There's there's an additional thing that you always need in these macro finance model a minimum dividend payout That is just for the these banks not to be able to sort of run out of escape the no run condition I won't talk about about that today Let's so what is the key trade-off that this bank? Faces when it takes its both its liquidity and its leverage decision so the thing so let's talk about the liquidity decision so Liquidity has a lower return than other assets So it's gonna accept lower return if it holds more liquidity on the other hand It reduces fragility as we seen earlier. So it reduces funding cost. That's a trade-off Instead on the leverage. Well, you know if you lever up You can get more of those returns in the market and the physical capital has extra returns on the other hand If you lever up we've seen it earlier fragility goes up and then your funding spread goes up So those are trade-offs and let's see what the first order conditions look like Okay, so the liquidity choice. It's this equation. It's linking the funding spread with the liquidity premium So what that is saying is if the liquid if liquidity is expensive the liquidity premium is high You will economize on your liquidity ratio and then that will mean high paying a higher funding spread That's why they're positively related in in equilibrium And then the other thing we get is we get this supply of physical capital that comes from the decision of these banks and All I want you to notice is that M M is the government paper that is supplied to the economy And so you can think of it like I said reserves and treasuries. They are gonna crowd in capital Okay, because you know if there's more Treasuries or reserves then these liquidity ratios are larger funding spread or smaller and it's easier to invest Okay, so that is That is it for the macro model now. We have to close it with policy So we have a model with real effects of liquidity, right as we've seen earlier the liquidity reduces spreads and sort of promotes investment in the economy so so we can think about optimal policy here and Actually, the up the there's a simple optimal policy that is to just Print so or create so much liquidity that spreads becomes zero. Okay And but there's a buck there We're not gonna do that. We're not gonna go with this and there's some problems with it So what do you need if you want to implement that is you're gonna need to produce infinite amounts of Reserves of treasuries in this economy so that that seems tricky You're also gonna have some fiscal cost So of course it means also that a liquidity premium becomes zero So it's gonna be more expensive to issue debt in that economy And another problem is that you know these banks are making money out of those spreads If there is no spreads there's no money for banks to make so you're gonna end up in a steady state with zero bank net worth So we said we think this is this is not reasonable to to to think of this model in this way So the way we're gonna write policy is more modest. What the what policy is gonna do is they're gonna stabilize spreads that is and So you're gonna if there's a shock to this economy you'll respond by supplying liquidity And What's the point of that the point is to mitigate mitigate that amplification mechanism that we have in mind Okay Right so now we're moving into the quantitative part So the left table those are moments in the data that we're trying to target The only one that is sort of different from the usual macro banking literature is the fourth the liquidity premium So what we're saying we're trying to target sort of the average from 1986 to 2008 liquidity premium that is defined as the difference between the repo rate Safe risk-free rate minus the t-bill rate. Okay, and it was 24 base points So that is how we that is what we calibrate the model to that is it's gonna pin down sort of our liquidity policies Obviously things have changed since 2008 that's gonna be smaller now But we want to study sort of the world how it was before 2008 and see sort of I guess one of the question that we can answer is do we want to go back to that Okay, and it gives us I mean as you can see to the right-hand side the only sort of novel parameters that we need over sort of our usual RBC parameters are the first three and that those we can pin down with With the targets that we have on the left side Okay, so let's jump into the IRS So so I'm gonna study I mean this is a shock that is often used to think about financial crisis because it Directly impacts the net worth of banks. So first thing I would like you to focus on the dashed line That is what happens after a capital destruction shock in an RBC model So an RBC model you see capital stock goes down 5% because for some reason housing in the US lost value, right? But net worth doesn't matter Neither does liquidity in an RBC model and you see spreads. They don't they don't change at all There's they're not really I mean they're pinned down at zero and you see I mean you think that you think I mean GDP falls and the banking sector tries to invest to Build back up the capital stock Now what we have now that is the that is the red line You see the capital stock is the capital shock is going to destroy some net worth So what happens is that spreads go up? Okay, that is the financial friction in action and the financial friction in action means that investment goes down It becomes much harder for banks to finance the recovery Okay, and you see that GDP goes down quite a lot more and the distance is persistent, right? Another thing that happens in this economy is that After the shock Banks profits go down considerably right because it becomes harder for them to fund themselves So their intermediation margins go down and that's why the shock is so incredibly persistent You see like for years these spreads remain high Okay, so this is instead a thing that I can't even show you what an RBC model a shock that I can't even show you What an RBC model would do because there's no point in because there is recurrent equivalence in an RBC model There's no no action following an increase in in the supply liquid assets But I can talk about it in our model. Okay, so what happens you see liquid assets go up a lot. That's kind of the shock So what happens is that all these premiums go down, right? There's more liquidity so liquidity premium goes down and then all this liquidity ends up on banks balance sheet So their funding spread goes down lower funding spread also means that it's easier for banks to lend So the credit spread goes down and you see GDP goes up and investment goes up a lot. Okay, so that's the mechanism of the model Okay, so we've gone through all the theory. I showed you some correlations in the beginning now Let me try to convince you hopefully that you know, there's actually something in the data So the question then is does liquidity actually reduce these spreads? Okay, so Let's see what we do. So we're gonna run an OLS on daily data It's gonna be very important that the data is daily and we're regressing spreads on treasuries on outstanding treasuries in the US There's data on outstanding treasuries day by day Okay, I mean we're gonna have a bunch of controls. We're gonna have a lot of lags I mean this stuff is incredibly persistent, right? So you need a lot of lags there and we're gonna have dummies to these To the season allies, I mean within the year so weekday day month and We're gonna have dummies for recessions and then we also have a linear trend to the trend Right. So what's the identification here? So the identification? Well, the graph is kind of small So I don't know how much you can see of it. So the graph the point of this graph is to tell you that The government in the US this is an institutional feature they they the auction for Debt takes place a few days before the issuance We're gonna be focusing on issuance but so what this means is that the government the Treasury cannot react to shocks today by issuing more More treasuries today, which is what the model would say them would have to do but for institutional features It takes a few days. Okay, so so what how does that help us? so the first thing it does is it sort of It rules out reverse causality right because it cannot possibly react to a daily shock to spreads Okay, within the same day The other thing that it does it clears out our results of the information effects So why is that because if you identify, you know, the daily innovation to treasuries That is sort of an innovation to me as an econometrician But everybody in markets knows with at least a day of advance how many treasuries they will be out tomorrow, right? So there is to the issuance there is no information there the information is out on the day of auction Which happened a few days before And yeah, then the third is that you know, we try our best to have a VAR setup or all the legs That are gonna capture the persistence. Yeah, that is we hope that that does a good job Okay, so this is just telling you how you should think another way of thinking of the of the regression So these are I argue that we're gonna have these innovation to treasuries and that those are exogenous That is the regressor and on the on the y-side on the left side. You're gonna have So the outcome variable that we're interested in I'll tell you what they are and the legs and the dummies are there to give you Sort of a benchmark against which to calculate the residuals Okay, so let's look at these so it looks indeed like Increases in treasuries on a day-to-day basis like these innovations. They do reduce spreads So we find that the well the way you read this is a 1% increase in treasuries is Gonna reduce the funding spread by 1.7 basis points And it's gonna reduce the liquidity premium by 1.7 basis points and there's no reason why these numbers have to be the same they just so happen to be and Another thing that goes in the direction of our model is that we find that there is a positive effect on the risk-free rate The risk-free rate to mind you as the repo rate And so as in our economy sort of more liquidity crowds in investment Right, and if there's more demand for saving as economists, we think that the interest rates would tend to go up okay, so Well been Early, so let me conclude so what have I done today? So I've presented a macro model and I put in a new ingredient that is bank fragility I told you about the coordination game that gives us the friction so The thing the reason why banks want to avoid being fragile is that it drives up their funding costs And they have two ways to do that They can either lever down or sort of try to control their level of their leverage or you know Horde a lot of liquidity that is gonna bring down fragility and therefore funding cost and that's where the trade-offs are gonna come in The macro model so what it delivers that I think is novel is a demand for liquid assets by banks that are worried of being fragile And it gives us this amplification and propagation of shocks via the spreads and yeah, so We can think about in terms of policy, so what does liquidity do in this economy? Well, it supports bank lending and therefore supports economic activities. That's I think that's that's what we want Then then we do a quantitative Quantitative exercise the effects we find are pretty big so this capital destruction shock is gonna be amplified by 40 percent I mean much more persistent and in the end we did some empirical evidence that it looks like indeed. Yeah liquidity shocks reduce spreads Thank you Thank you very much. I think we should have a price for people who finish their presentations before time So you would have won it by wide margin So the discussant is Ansgar Walter from the Imperial College in London Are you know Ansgar by now also who has impressed us with various contributions to financial economics? In the last few years in Prudential field in several other fields Thank you very much for the kind introduction. Thank you for having me. So I've been ambushed The correct response to an ambush is normally to to run away, right? But I will I'm an academic so I'll respond by by talking so I Have my discussion. I'll point out the things that I'm going to say on the fly in response to some of the new facts Okay, so but the facts in the version that I saw that motivated the exercise were basically that in financial crises we see dramatic increases in credit spreads and Afterwards large and persistent real effects. These are sort of very well known facts Lots of I'm not even citing people here. Lots of people showed those after 2008 now The extra fact is that the supply of liquidity matters, right? So the these spreads as well as You know the the Treasury rate are Affected by supply and liquidity that sort of fact number three that you should add in your mind over there from the new version Right now we have lots of explanations of this stuff as well, right? So why would credit risk? Why would credit spread spike? It could be just even if you hold You know the risk a version of the market constant They might spike because there's just been an increase in the likelihood of distress or default And so that that's just increased credit risk out there now. There's also lots of evidence that Risk a version or risk premium in the market account to cyclical, right? So for various reasons for example habits or intermediary based asset pricing or fire sales or that sort of thing Risk premium might go up for a given level of distress risk in a crisis, right? So both of them would speak for higher spreads in a crisis And I think we have a pretty good understanding of of what's happening there from various perspective There's also behavioral perspective that people just psychologically become more risk averse in a crisis and that's also probably valid In terms of the large and persistence real effects There is any number of macro finance papers looking at amplification You know flight to safety is a thing that was studied first in the 80s by Bernanke and co-authors We've since had a lot of development there, but basically through Bank or expert more generally net worth a lot of these shocks get amplified when people face financial constraints And have to fire sell their capital and then they're even poorer and then they have to fire sell even more and so on You have all these Spirals going on right so we kind of have a story for this I'm wondering if we have a story for the link between Treasury supply and and prices. I was thinking about sort of the older literature on you know inside and outside liquidity There's a a lot of studies by Homstrom and Tyrol for example on that Topic where you know liquidity supply combined with financial constraints in the spirit of some of these other papers Does give you a link between liquidity supply and prices? You know, it's kind of not shocking that in a world where people face financial constraints Supplying something very safe that people can hold and used to pledge as collateral Changes the way the assets are priced and might alleviate a crisis But what's new here is that you can sort of rationalize all that stuff at the same time using the threat of debt runs or bank runs Okay, I should be clear here that this model is kind of about banking But a lot of the people that ultimately lend to banks, especially the uninsured Lenders to banks tend to be also institutions, right? They tend to also be People that are subject to balance sheet constraints of some kind So I think some of these explanations still apply, but here's a new one based on the threat of runs Okay, so something very quick on on the theory of runs This is in diamond and divvig runs are just a sunspot, right? They can happen whenever Wherever for no reason you switch from the good equilibrium to panic Global gains are attractive because they're more in line with history history tells you bank runs don't just happen out of the blue They tend to happen when there's some bad news Right, so the simplest version of the bank run global games version Was in a in a paper by Morrison shinnon or one But basically just says that you're gonna get a good equilibrium i.e. no run whenever the fundamental value of your assets Exceeds some threshold the NPV has to be not just positive But it has to be above some threshold that measures how strong the strategic complementarity between depositors is and so basically, you know that gives you Runs in bad times, but I think the literature has had a love-hate relationship with this because as soon as you put this into more realistic Models global games tend to become a mess. It's very hard to solve this stuff and you you hardly ever get clean characterizations Enter Davide and Kevin who have a very clean characterization and they go through a lot of effort to hide this really clean equation Which I think summarizes what happens. Okay, so You have capital K, which is illiquid. So let's discount that by by a factor lambda which measures the illiquidity you have money M and So that together lambda K plus M is sort of your amount of liquid assets roughly speaking in a weighted way, right? Now, let's say only a fraction f of your debt gets rolled over You have to keep you have to make sure that your liquid assets are greater than 1 minus f times your debt and That's exactly the condition in the paper. That doesn't hold you get a run Right. So basically when you face is the risk of a run you have a flight to safety You prefer M over K and you get amplification again via the net worth, right? Once you have to do that you hold more M and less K and you earn lower returns on your investments And then you you can invest less in the future and you're even more close to the run constraint So this is kind of neat. I thought this is kind of the main selling point is to to get something so tractable and so realistic This is exactly what we think about when we think about liquidity regulation, etc. I think it's very nice I would compare it a little bit to the literature on fire sales that I've worked on a bit more, you know Suppose you're a person who needs to roll over 1 minus f times your debt next to next period so basically only a Fraction f of your debt is is long term and the rest you have to roll over and you can borrow up to a fraction lambda K If your capital, this is more like a cute tacky more style model where you can borrow up to a fraction of your of your assets Right and you also have some M Now what is the constraint that you have to satisfy to avoid having to sell your capital to unproductive outsiders? Exactly the same thing right so Sort of observationally equivalent you can get in the fire sale models. You get a flight to safety. You get amplification via the end What you don't get in the bank run version and you do get with fire sales Is that the distress actually sometimes happens in equilibrium, right? You do see fire sales in these models and they generate an externality a pecuniary externality That's not properly priced. So you have too many fire sales. Okay, whereas in David is model For a given stock of government supply of liquidity Everything's fine. All the banks make sure they satisfy the no-run constraint and we never actually see any distress That's what the fire sales model are a little bit different And I want to push you to sort of you know test your models against alternatives a little bit more And this would be one of them that that you might consider Another thing that's nice This is sort of just selling the paper for you a little bit more if I look at the model constraint and I look at an LCR You know and that told us yesterday. Don't worry too much about the LCR. Just think about demand for money But here's the LCR anyway, right? You have sort of like a weight times your illiquid assets plus your cash Right, that's your high-quality liquid assets in an LCR has to be at least as big as a net outflows Your experience in a stress test scenario, right? Which is essentially some weight some other weight times your long-term debt short-term debt Excuse me. So it's the same equation again, right? Like this is K. This is M. This is D so you kind of Have a model that can gives you the correct scenarios and weightings to calibrate an LCR Which is which is kind of neat, you know that it comes out so directly It doesn't usually come out so directly definitely from global games models and also not from New Keynesian models with financial frictions. So I think this is actually quite a nice thing Now let me talk a bit about the spread so this condition you can unpack because the the run rate the f Or the one minus f depends on the spread Okay, when you have a higher spread when you're giving your short-term lenders a higher spread on their investment in you They're less likely to withdraw right so you increase the funding spread in equilibrium to ease your constraint Right if you're on this binding constraint where you're about to face a run You can promise people a slightly higher rate on their deposit and you get away from the run constraint Right, so you're increasing the spread which is here called J. I guess to avoid distress in more standard models The market increases J for you when they anticipate distress, right? This is a model where stress never happens, but you promise people money so to make sure it never happens Right and normally we would think of spreads is going up because people are pricing in the possibility of distress In a heavier way and also maybe applying a higher risk premium to that Right so again, this might be a testable prediction of the model That's very different from other models and that you might want to you might want to consider to sell your story Is something we need on top of all the other microfinance literature. Okay, so Okay, so I'm probably also going to finish before time which Which is nice because I got two minutes head start But this is my last slide, but I'll talk a little bit more broadly around this about the contribution of the paper Okay, and this is a little bit Intricate I think it's written like a Vanilla macro paper at least the version I have right it's kind of you you have a theory you plug it into a general Equilibrium model you draw some impulse responses Right, so that's a crowded space Okay, like that. There's a lot of impulse responses of that have been drawn since 2008 around things like capital destruction shocks net worth shocks all the rest of it, right and and The impulse responses you drew were yours versus a completely vanilla RBC model That seems like a straw man to me. Okay, I think that you know At least run a Gertler and Kiyotaki or something like that, you know see that's that's who you have to win Against I think to make sure that people realize this is a new story something that adds to a quantitative understanding okay, also, I wouldn't necessarily go for a capital destruction shock because I Don't know what you saw on the impulse responses was the increased Investment after a bad shock because if all you do is just randomly blow up some capital The first thing that the agents in the model do is really quickly rebuild capital Right and that's completely counterfactual investments is one of the things that tanks the most during finance after financial crisis Right, so I think you should be hitting that and I think you should be blowing up something else Not just not just a capital stock, but let's see Okay, so I would say a little bit about the contribution in terms of timing and nature of debt runs I think I cannot stress enough that there's a very elegant solution to the conundrum that global games kind of feel Like the right way to study bank runs, but have so far been very intractable. Okay, and you've done a very good job the trick there is to basically Define a very clever variable of a very clever state variable about which people have heterogeneous information Okay, that's kind of cheating but to be honest because we let the heterogeneous information go to zero anyway When we solve the thing it's fine because you're cheating about something that you're letting go to zero So I don't care right so I think this is very elegant, you know And you have this this nice equation with liquidity and leverage choice in the same place It relates directly to the LCR, etc. I think that's that's very nice I think you can sell that much more as the core contribution of the paper Just make sure you don't get any macro referees in that case, but I think that's You know, I think that's a way to to sell this slightly differently and maybe to a more general interest audience So you can either go down the you know The quantitative route is is important especially because you've also added the liquidity thing at the end Right, so what you need to then do is maybe convince me that the liquidity thing your IV at the end and the rest of the story You need to be in the same paper and I think that Again, maybe they're pictured against something like home strumming to roll picture against someone that's actually thought about this before because They have this is just a different story based on bank runs and that's very attractive, but you know I would think a bit about that Okay The other thing I would think about if you're pushing more that you have a much better way to analyze bank runs than other people Which I think you do is Can you get to a point where runs actually happen in equilibrium? Can there be some sudden stops, you know You experience a shock that you just couldn't hedge against and the run actually happened some of the time and So you'd now have both a very clean equation for what banks do to prevent the run at the same time as some Characterizations of run when actually runs happen in the data and that again makes the model more testable and and it Let's the model say something that other papers aren't able to say so I think I'll stop there and Thanks very much again for putting me on the program Thank you very much Another price one You may not have realized I did all this to squeeze them with time that you have more time to ask your question So please doesn't don't disappoint me and are wonderful So I see I think the first hand was there in the front Please state your name and affiliation nevertheless, even though some have may have heard it already When she do Columbia Business School, I'm a little intrigued by the new empirical results So are you sure adding treasury reduces funding spreads and increases liquidity to me adding treasury drains cash from the system Which is the more liquid one? So it should go the other way around I think about the treasury has to be financed and they increase the TGA actually drains bank reserves are the things being said Let's collect two or three. No, I saw another hand and left on that side. Yeah Hi, me close very from Bong de France. So actually my question will be more for and the guard you mentioned the idea of the LCR in this paper that maybe That justifies liquidity regulation what I was wondering since you seem to have put some thought around this topic is if you had in the model a land of last resort who commits to Liquidity whenever there is a run My first intuition that that will be maybe superior to be to be LCR That's actually a general feature. I haven't seen any model which tells you that a liquidity regulation is Actually greater than the lender of last resort. This people seem to be kind of allergic to The LOR for political reasons. So I mean you have some views on the literature If you have any thought on this it will be interesting. Thank you There's a third question there in the front. There's a micro needed there in the Second row third row on the right. Yeah, thank you. Hi. You mean mark will be a business school I'm curious about your treatment of liquid assets of both being reserves and treasuries because it is true that reserves are held On bank balance sheets leading to the effects that you described But you know treasuries are not primarily held by commercial banks, right? And so if you have like a non-bank was holding a lot of treasuries and perhaps the non-banks is issuing Liabilities that then our substitutes this deposits, right? Would that create an additional channel where actually the issuance of more Treasuries could actually crowd out your bank deposits. I Don't see any questions online, but Christian I think then the back has a last question. Maybe then it's time to close. Thanks Christian Kubitsa ACB I just want to add to something that you mean said is that you Treat reserves and treasuries in the same way you focus a lot than an empirical part on treasury rates I was wondering whether or what are your thoughts on using your model to quantify or understand the effects of QE or QT So what is really the role of reserves in your model? Can can we think about the central bank increasing or decreasing the M that you have in the model and what effects will that have? Thanks. You have a very nice question because it would also allow us to make the transition to the next paper very nicely So you have an amazing amount of time to answer but I don't get courage. I'll encourage a long monologue though Thank you. So if it's okay, I'm gonna start from the from the end So, yeah, we wanted to keep things simple in this model And that means we have one liquid assets and one in liquid acid and it felt Okay to sort of bunch together reserves and treasuries in the liquid category But sort of if you take that seriously, you know QE as long as QE is literally issue reserves and Buy T bills then it wouldn't have an effect. I don't think that's really realistic I mean, I think possibly reserves are more Liquid than some other than some of the treasuries at least the long term maybe But yeah, we just didn't do that. It wasn't sort of we thought that the additional complexity of the model wouldn't have been paid off by by further inside but but it's a good point That is probably how we want to go about if we really want to think about QE Okay, so what Okay, let me go to the TGA question So it is true I mean, we actually have the balances of the TGA in the as lags It is true that so the way it works when when the government sells treasuries it Sort of in a way banks pay it in reserves and these reserves end up in the TGA balance Of the of the treasury that means that there's fewer reserved for sort of financial intermediaries to hold That is from day one. Now these shocks are persistent. I mean, these are innovations, right? But All these variables are persistent So if we run a local projection what happens is that the money in the TGA after sort of an innovation to Treasuries get spent pretty quickly not all of it like a small fraction gets spent on day one But we see that almost all of it is spent within a week so in this sense more treasuries increase the amount of liquid assets available to To intermediaries within sort of it takes a while Okay, so I think you may want okay, so I it's a good question. I think it depends on What exactly you think the market failures are so it in a background model? You know if you think the most prominent market failure is just failure of coordination Then the lender of last resort is the optimal response, right? It's that's what diamond and Divick already pointed out right like you might want to suspend convertibility But that's not very palatable So lender of last resort sounds great or deposit insurance sounds great All these things sounds great because you don't actually have to use them You just have to build a big house that says FDIC and then nobody runs anymore because they think the FDIC is there Or the lender of last resort, right? But in some of the broader models for example with fire sales like I said you get the same kind of condition But you also get an externality and now Having a lender of last resort just there would actually generate moral hazard It would it would make sure that people exploit the externality or cause more externalities than before and so that's maybe more problematic. I think So it really depends on where you think the frictions are I think can I follow up on this? So I mean the model is about sort of ex ante liquidity in a way and you're saying well, what about our exposed liquidity? So we think we're thinking about an extension one interesting thing that I can say is that sort of it's true But like if you know you say I'm gonna buy those illiquid assets at one So no liquidation costs exposed you you you make the friction go away But sort of even though this is never taken up Ex ante this has a this has a fiscal cost in the model because it you know the liquidity premium would go to zero Right. I mean there's no reason then ex ante to hold any any liquidity So that's one interesting thing. So like ex post Lender of last resort actually has a cost in this model even though it's never picked up So I mean I thought that was interesting. Would you mind reminding me of the question? I'm really bad at this Were you able to hear it? Yeah? Yeah, I was able Yeah, I think fundamentally This is a question about sort of what are your banks sort of who what institutions get into that? I think this is a bit of a It's a difficult feature for all these macro finance papers to think about What should really go in there, I mean we are thinking that you know the key characteristic of what should go in that category is Institutions that do maturity transformation and therefore sort of through this mechanism we get that liquidity crowds in investment I think what you're saying is that yeah, maybe not you know That's not a primary concern for every for every institution and therefore for some institution may go the other way around Yes, probably yes It's I would have to think how to add this ingredient and maybe then it's fair to do a horse race between two. Yeah, thanks Alright, I'm afraid you have to give your price back because you went into the red now, but okay, but not very much. So that's okay I hope that Viral is slowly coming online here. Maybe we can put him on screen because we're moving to the next paper Ah, here we go. Okay. Good to see you Viral. You look good on the screen So the the title of the paper is much too long. So it's actually impossible to read out But I believe what he but Viral and Ragu want to convince us is that if there's more liquidity injection That may not me lead to more liquidity in fact or they may some be some countervading effect So actually a paper that has led to quite some discussions in the central banking world already So I'm really looking forward to seeing it. It presented Viral. So you have 25 minutes I don't think we can Can you see the screen for that? Yes, we see you and the slides together beautifully So no adjustments needed. That's all good It's there go ahead Okay, perfect Okay, thank you so much everyone for allowing me to participate remotely Unfortunately are or fortunately our empirical paper got on to the NBR monetary policy program and I Had to sort of do double duty. So I'm up early, but I'm wide awake and I hope I can say something interesting So Ragu and I started looking at this issue perhaps because both of us were In the midst of liquidity management while in India and we started wondering about Sorry, I think after there's a question to switch the full mode, right? Okay Does that address it Stefan? Okay, okay, very good. So yeah, we started asking this question of How does the central bank balance sheet alter the balance sheet of the banking system? Because ultimately if you want to know how the transmission is going to happen or happen the right way We do need to go from the central bank balance sheet to intermediary balance sheets Perhaps in the grandest version of the paper one would also think about non banks But as of now we are treating non banks in a very reduced form manner This is mostly about the banking system and Thanks to Silicon Valley Bank and Credit Suisse. It seems we'll still keep caring about banks for some more time. So Okay, so the real question that we started with is why is it that Even in a world of what seems to be large quantum of central bank reserves the ultimate form of liquidity as mentioned Do just explain Why is it that we are still getting these episodes of liquidity scarcity? It's it perhaps seems unusual that from a world in which we used to have 20 to 50 billion dollars and manage liquidity We now have several trillions of dollars and yet somehow it's not enough You know every now and then we are getting punctuated by Problems in money markets. We had the repo market spike in September 19. We had a dash for cash in March 20 We had the pension funds collateral or liquidity crisis in Around the fiscal problems in UK in October 22 and then perhaps topping it off all were the uninsured depositor runs Based on bank solvency concerns in March of 23 in the US Now there is a proximate cause for each of these and you can write a separate model and a separate explanation for every single one of this And each one I think has some Shades of its own. I think we are trying to see if there is something common. That's happening to the system as a whole At the level of the macro balance sheets of the banking system and the distribution of reserves within the banks, okay, so The theory we have may not necessarily be a perfect explanation for each of these episodes But we think it as something deeper to say about why the balance sheets of the banking system have been evolving the way they are Okay, and so we're going to start off with the assumption unlike the most current setting in the US in which money market funds do have access to Parking reserves at the central bank, but we're going to assume in this model that It's the commercial banks that are holding the reserves. So at least for most part of History to date this would be a reasonable assumption So the question is how does the Fed expand its balance sheet or a central bank expand its balance sheet and it can happen in two ways this mechanism Either the Fed buys directly from the commercial banks So it could be a world in which essentially it's a bank dominated financial system Or more in the modern advanced economy setting The central bank is also buying or primarily buying in some cases from non banks, okay and it turns out this matters for The size and the way things may play out in the commercial banking world, okay So the first mechanism is a is an asset swap with banks. So in the pre QE world on the left-hand side You know the Fed is running some balance sheet Banks are running some balance sheet and And there are non banks as well, I think I'm starting with the second mechanism here first, sorry So so banks have some deposits. They have some treasury securities and they are holding some reserves Okay, now there are also non banks for simplicity their balance sheets are modeled in a In a waste kind of non leverage fashion So they have some wealth on the right-hand side But they have some bank deposits and they are also holding some treasury securities So, you know, this could be insurance companies pension funds mutual funds Cy net worth individuals family offices, whatever it be so If if the Fed expands its balance sheet here It goes to the market and purchases treasury securities by a billion dollars in the in this mechanism Where it's the non banks that are tendering their assets to the Fed effectively via the banks So what happens is the Fed buys the securities at the auction The the bank that's participating in the auction tenders The securities to the Fed and gets the reserves in exchange But actually ultimately there is the securities were tender to the bank by the non banks So the non banks are the ones whose treasury securities are coming down And their financial deposits are actually going up by a billion dollar at the end of the QE It turns out If you have access to Fed data, which currently I do because I'm on a sabbatical at the New York Fed You can actually track this even at at the daily level for some banks And at weekly level for a larger set of banks and see whether it's the financial deposits That are actually going up and QE is done and it turns out Most of the assets swap of QE happens in this manner Which is that the non banks are tendering their assets via banks to the Fed as a result financial deposits go up And importantly the bank balance sheets are expanding. Okay, so that's the point I want you to focus on that in this version of QE when central bank balance sheet expands bank balance sheets expand by dollar mechanically and backing them are the reserves in the aggregate banking system as a whole There's another way in which assets swap can happen Which is simpler which is that banks simply tender over their treasury securities or agency securities to the Fed and In this case at least mechanically there needs to be no alteration in the size of the commercial bank balance sheet Okay, there might be some second-order stimulus effects Maybe maybe this transformation of the asset side creates some economic economic activity But at least make the mechanical operation of the QE need not expand the bank balance sheets So the question is what happens in data So in a parallel paper that I was mentioning I'm presenting later today here What we show is that when reserves expand in the system It does seem to be the case that the deposits in the system go up So the reserves here are in the blue line relative to GDP the vertical lines are the various phases of central bank waxing and waning of the balance sheet In the first three rounds of QE as the reserves go up the deposits are going up as well Of course, it takes a little bit more to establish that there is some causation link here And importantly when the reserves are coming down in the passive phase After QE of the balance sheet and then active quantitative tightening, which is QT1 You see that the deposits stay sort of relatively flat Then you have the pandemic QE when the reserves in the system are expanding massively and you can see this huge surge in the deposits If you drill down further, you can show that it's not all coming from the fiscal stimulus There is actually a piece linked to the QE as well And then of course from March in in 22 onwards the deposits actually started shrinking ultimately ending in runs Furthermore if you divide the deposits into insured and uninsured deposits, just focus on the demandable deposits at the top here The thick line is the uninsured demandable deposits and the dashed line is the insured demandable deposits In the initial phase both are growing in the first rounds of QE and they're both stabilized But especially at the time of pandemic QE the continuing increase of deposits with QE is actually that in the uninsured Deposits and then ultimately they correct in a big way So At some level in the central banking world This is now almost accepted as given which is that when the central banks expand the commercial bank balance sheets expand to When we started out the work perhaps there was a little bit of uncertainty about this assertion But at least now this seems to be proved in a variety of central banking data all over the world But importantly it looks like there's some hysteresis that once banks expand their balance sheets They are not keen to reverse them fully At least this didn't happen in the 17 to 19 Qt period So we are trying to build a micro model that explains this like What happens in bank preferences on their balance sheets and why might there be some fragility? associated with this creation of a stock of uninsured demandable deposits in the system and our explanation is going to be that Once the reserves are injected even if in the banking system as a whole they are Hacking deposits one for one the distribution of reserves in the banking system afterwards may not necessarily line up with where the demand Or where the demandability of the deposits is and for this not to be a problem It has to be the case that the interbank market works beautifully and you know under simple assumptions you can break that and therefore This creation of a stock of Demandable claims in the system may be a problem Because you're dealing with a larger stock of demandable liabilities. Okay, and so that's our main conclusion that that we should Think about stability of fragility consequences of QE because it's not just an expansion of central Bank balance sheet. It's actually an expansion of the commercial bank balance sheets and importantly with uninsured demandable claims Okay, so the basic model to to tie it up nicely. It takes a little bit of work We have firms banks investors Because banks are lending to some entities in the economy that pins down their overall intermediation size They are raising some deposits and equities so there have to be some investors And importantly, there has to be some heterogeneity in the banking system Even if just exposed through some realized shocks because you want to think about an interbank market that will shuffle around these reserves To where the demandable claims are running So we organize the model at the as these bank firm pairs which are regionally or sectorally matched Firm and bank owners in the model are risk neutral. They are expected profit maximizers. So nothing dramatic happens through them Demanding liquidity themselves Firms will have some liquidity demand, which is that they make some initial investments Invest so this is like a homestorm to your old sort of a setup. They make some investment at date zero There's some initial wealth, but they will still get some term loans on the margin the term loans are going to go into their own cash deposits at banks and When they get hit by some shocks at an interim date one They may also be drawing down on these cash deposits from their Relationship banks. So the firm deposits are D zero with the superscript F at the top Now the most important balance sheet is that of the banks. So at date zero, they have some long-term assets Which are these term loans to banks. We're going to assume that they are illiquid at date one, but that There is enough for them to pay off the claims at date two And they may be holding some liquid reserves as zero This is going to be determined the size of this is going to be determined by the central bank only the central bank can Control the aggregate quantity of reserves in this model And therefore with the banks in this model because there is no Reverse repo facility that can have reserves moving over to non-banks or money market funds in particular But now we'll assume that at date one there is something that happens which takes out a part of these reserves The most natural interpretation is that a quantitative tightening is going to happen at date one And that's going to be for a proportion of the reserves tau in the model You can model it as a constant quantity. It doesn't alter the results qualitatively Okay, so the question then is if this is zero is going to expand the bank balance sheet the way I described it How will the banks finance the reserves on their expansion? So they can raise uninsured deposits and there's an unlimited stock of this in the model The simplest way of thinking about it is that it could just be an asset swap And the other way is that they can raise some equity capital Equity in this model is just a claim that doesn't run at date one So if you want this can even be our franchise of core deposits The important part is that to raise more of it. There is some convex cost of Issuing them whether it's a due diligence cost or whether it's your franchise cost of maintaining branches It doesn't matter so much in the model. Okay So then with some probability There's a stressed state of the world in which some of these firm bank pairs Up to a proportion theta by law of large numbers get affected Firms are now going to have to rescue their investments, but more importantly the depositors who are very stressed And we assume this sort of like a Stein 2012 tile extreme risk aversion that these depositors are also going to run on the bank Okay, so what's the exact mechanic that's at work? Your firms are getting affected. They are drawing down on the bank So it's like the tech sector drawing down deposits on Silicon Valley Bank and Seeing that the depositors who are extremely risk averse say now There is a small chance of an insolvency at day two Even though there's no expected insolvency at day two But because these depositors are extremely risk averse they'll precipitate a run at date one itself Okay Now the run is then a claim on the reserves of the bank But a fraction tau of the reserves have now gone out So the firm bank pairs which are distressed are now going to get depositor runs and a scarcity of liquidity So where do where can they get the reserves to settle their deposits? They have to either go in the interbank market or they have to raise more equity What's raising more equity in the model? It's a conversion of someone else's deposit claims into an equity claim Okay So that is going to be also a transfer of reserves from surplus banks to the needy banks So either surplus banks can lend to the needy banks in the interbank market All the needy banks can issue equity which in the aggregate is the conversion of someone's deposits claims Into an equity claim and as long as those depositors are with surplus banks There'll be some transfer of reserves that takes place with equity issuance. They're going to assume It's one for one. It may be even less than that if it's your own depositors who are converting claims into equity But we're going to assume it's all coming from the surplus banks out there Okay, now importantly, we're going to assume that You know, there's some market rate R1 in the interbank market There's some quantity of equity that's going to be raised even there'll be a convex cost of that So there'll be a marginal cost of equity capital issuance and In equilibrium these two rates have to match which is that the equity issuance rate on the margin is the same as the Interbank rate at which the reserves will get lend So we're going to assume that for to start with just exogenously that there's a fraction fee of banks That's willing to lend in the interbank market and a fraction one minus fee think of this as JP Morgan It just wants to maintain a fortress balance sheet and not have its reserve balance is going down during the day And they are going to forego lending, but they're going to now get flight to safety deposits Okay, so in ultimately we endogenize this fee and that's going to be the trade-off between these two decisions Do I lend in the interbank market and allow my reserve balances to go down and In the process give up the flight to safety deposits Or do I just remain like a JP Morgan on the side getting all the flight to safety deposits and giving up my high Interbank market returns Okay, so that's the additional wrinkle in the model which is that the stress banks Getting money in the interbank market from what we are calling as tainted banks a fraction fee of them But the flight to safety is then removing those reserves back to the safe bank And the safe banks are now growing and there's some advantage to them from doing that because they've become Disproportionately large even though Exxon they all banks are identical in the model So market to some market power gain on reserves or deposits at the end of the day when this happens Okay, so I won't go through the exact optimization problems firms are Like in homestorm Tirol doing some management of liquidity shocks Banks are doing some capital structure Optimization the key equation is actually the budget condition of the bank Which is that on the right-hand side there are deposits and equity in their capital structure on the left-hand side They are intermediating loans. We have some Intermediation cost to pin down the size of the balance sheet But there's this talk of reserves that the central bank is supplying that they need to finance Okay The key so so that's at date zero and the way this model works is that everyone anticipates What's going to happen in the stress state of the world at date one is the world going to be in a In a in a state of interbank friction. So the rate R1 is going to be above normal So there's going to be some interbank liquidity premium some equity issuances taking place Which are then going to pin down the interbank premium or is that is there going to be so much surplus liquidity? That actually the interbank rate will be pinned down to no premium and no equity issuances will be taking place in the model Okay, so that key condition that ends down the equilibrium rate R1 is the following Which is that the market for spot loans at date one has to clear at a going interbank rate R1 But the stressed the stressed banks and the ones that are arbitraging in the interbank market They have a need or a benefit to issuing equity So they will incur equity issuance costs and then there's a cross market arbitrage that pins this rates down Okay, so what is that arbitrage condition? This is the core of the model in some sense You have some equity issuance is taking place. That's the additional liquidity being raised via date one capital issuances So that's by the theta banks which are hit and the one minus theta times fee banks which are in the interbank market There's some net demandable claims running on the stress banks in the system. How large is that? That's theta times the new demand for investment from the firms that are hit by this home storm to roll shocks and The risk of us depositor runs that's taking place Okay, so the tech sector and the uninsured depositors are running at the same time And then there's some liquidity of reserves in the system but a fraction one minus tau that was not with the Stress banks and with the interbank active banks is now disappeared from the system. Okay, so the reserves of the JP Morgan are not in the system and and a fraction of the QT reserves that were taken out are also not in the system anymore Okay Now why is this equation important because it actually captures the two key forces in the model? Okay, which is that if you just jack up reserves somehow the s0 gets increased as an exogenous shock Without altering the size of the bank balance sheet at date zero Then of course if you increase s0 it has a negative effect on the equilibrium interbank rate R1 Which is being determined on the left-hand side. Okay, so the sign in front of s0 is negative That's the standard intuition that when you do QE you are injecting reserves The world hasn't changed it's just somehow that the central bank changed the stock of reserves and so the interbank rates have come down Okay, but what the point of the model is that no no not quite because how are you altering the stock of reserves at date? Zero the banking system changed itself and in fact in the model There's a one-for-one increase in the stock of deposits with reserves because equity issuances are costly Whereas wholesale deposits are just going at the going rate of zero in the model at date zero So deposits at date zero increase one-for-one and therefore now there's a trade-off because these deposits are a demandable claim With some probability and on a part of the system at date one So the key intuition is is the injection of reserves at date zero How much surplus liquidity does it create and does it always create surplus liquidity or not? Okay, and we have this main condition of the model that if the size of the shock is large Which is a big proportion of the system theta gets affected if Qt is going to be large Which is that a big part of the reserves got taken out before this shock arose by a fraction tau And if not many banks fee are active in the interbank market So if theta is large fee is small and power is large Then you can get that you get stress and R1, which is this interbank rate at date one Can actually be abnormally high. There'll be liquidity scarcity and therefore a capital scarcity In the system in the model Okay, the way the model is written the insolvency happens at date two with some probability But that's enough to precipitate a run at date one because of the extreme risk aversion of deposit Okay, just a couple of quick points as to what we do then in the model There are there are few extensions that are important. We do endogenize this interbank market Active share there's a simple way to do it Which is to just assume that whichever bank becomes bloated as a result of the fragility Typically, it's been JP Morgan in the last 15 years that it earns some convenience yield on this reserve It can you can think of it as market power on deposits or if you want It's a convenience yield on reserves and this means that they would rather sit as a fortress And then in the model we endogenize what fraction feeds it's goes in the interbank market and one minus fees It's as a fortress. Okay. That's one Second we endogenize the reserve shrinkage tau. Okay, which is that we we endogenize or at least sketch How can it be that when the central bank engages in QT? It's shrinking reserves, but banks don't want to shrink their balance sheets and what we show now We just realized this. Yeah, I have to come to an end We just realized this last week that the same force that makes commercial bank balance sheets expand during Q E, which is that banks want a larger balance sheet They want an extra deposit for an extra reserve is the same force Which is going to prevent them from shrinking their balance sheet when QT happens because they actually don't want to lose a deposit for a loss of reserve and So we show in the model I haven't had chance to go through this that there's some force which ensures that the That commercial banks will expand during QE, but they're not shrink their balance sheets during QT They'll shrink their reserves, but they'll just do an asset swap with the central bank and get securities instead Okay, so I'll stop here. This is a model in which fragility happens because of the total size of The reserves as zero if reserves as zero become very large the uninsured demandable deposits become very large But not all of the reserves are available to move around in the system. Okay, so let me stop there And I look forward to hearing your comments. Thank you So the discussion is will diamond one of the upcoming young scholars from the Wharton School at UPenn Here we go great Well, thank you so much for the opportunity to discuss this really interesting paper and for having me out to the ECB It's my first time here seems like a wonderful place And I also am being a little short on listing the whole title of the paper. That's the first line I didn't include the whole thing, but I really enjoyed reading this paper and let me quickly summarize what it's trying to do Which is a it's a model of the impact of central bank reserve injections on financial stability So there's other parts of QE in terms of the yield curve, you know Treasury supply which is not modeled here, but what the paper is really trying to do is understand in a nice model with good Institutionally relevant frictions if we add reserves to the bank balance sheets How does the commercial banking sector respond to this and in what conditions might this actually have counterproductive effects on? stability now the econ 101 Benchmark here is if there's a supply for liquidity and a demand for liquidity and we supply more liquidity The demand will be satiated and there will be fewer liquidity crises now That can still happen under some conditions in this model But because we endogenize the choices of bank balance sheets over a wide range of dimensions Once the liquidity these reserves are injected into the system the banking system can become quote-unquote Dependent on that liquidity through its endogenous choice of capital structure making stability more of an issue going forward So there's two frictions in this paper which really lead to this possibility and they both are emphasized in for all's talk One is that banks supply more deposits after reserve injections, which are hard to unwind later And what this gives you is a notion of path dependence where if I do Injections through QE right now. I take out reserves through QT later That does not net out to a zero total effect because the banking system now has more liabilities Which require the liquidity on balance sheet the second thing is what happens in a banking crisis? So when bank health is uncertain and different depositors might be worrying about where they are or are not going to lose their money They likely are going to run to the sort of heavily regulated strong balance sheet large too big to fail banks And as a result you might have a mismatch of which banks have the reserves in which banks need the reserves I think we saw this in the Silicon Valley crisis where lots of reserve rich banks sitting in the center Deposits are being reallocated out of other banks into those whereas in a frictionless world the aggregate liquidity supply in the banking system Could potentially be reallocated to possibly prevent some runs and the paper models both of these really nicely And that's what leads to these interesting counterintuitive results where we have an increase of the liquidity supply Making the risk of severe liquidity crises worse Okay, so a quick overview of the model itself. There's going to be firms in banks So firms have a project initially and it's going to get a payoff at the end of time if everything goes well The more I invest the larger my payoff But what's going to make things interesting is at time one in between the beginning at zero and the end at two The project could potentially fail and the firm gets a chance to rescue the project So if the firm invests I one at time one, it's going to get an amount of output G one of that Which is increasing in the amount invested so I start with an initial investment if I'm in trouble I lose that initial investment, but I have an opportunity to sort of come back and invest a second time This generates a precautionary demand for deposits and this also leads to the banks who lend to these firms When the projects blow up are going to have a bad shock and because of that there can potentially be a run on that bank So the banks then who lend to these firms are also going to hold an exogenous supply of reserves supplied by the central bank They provide these loans with a quadratic cost Which makes them a little bit sticky to adjust and then they're going to finance this with deposits which they can adjust flexibly an equity Which also goes go going to have an adjustment cost now because of this when you inject more reserves into the system The most flexible way of adjusting is to add more deposits So $1 more reserves $1 more deposits all the lending and equity issuance costs are held fixed And that's why we have in the model that reserve injections lead to this expansion of bank balance sheets with more deposits coming in So unlike reserves deposit quantities here are endogiously chosen and as the reserve injections happen That's how we get this response of this levering up an expansion of bank balance sheets So at time one there's either going to be a good or a bad shock the good shock nothing interesting happens everything None of the projects fail and there's no runs When there's a bad shock because each bank lends to only one firm Some fraction of the banks now have this disaster where they've lost a large share of their balance sheet And they can potentially experience runs and what's going to make things interesting here is that a fraction only of the healthy bank of the Healthy banks are going to participate in the interbank market Now again kind of like the Silicon Valley example I talked about some banks are in trouble some banks are healthy and of the healthy banks some say I do not want to enter the interbank market and solve this problem because by staying out I can convince depositors I am super super safe and you might as well pull your money out and run to me instead and that leads to a Mismatch between where the liquidity exists and where the liquidity is needed So if there's a crisis even if aggregate liquidity is enough to solve all the issues Because of this issue of depositors not knowing where to put their money and this sort of signaling idea of wanting to keep your balance sheet Strong that's how we can have larger reserve injections leading to runs actually being worse ex post Okay, so my first comment on this is there's an implicit notion of dynamics in this So the model is simple, but very sort of institutionally relevant and realistic And that's how we can kind of tell informally a very subtle and complex story with nice tractable math But what the papers really emphasizing here is the notion of how deposits are building up during booms and being taken out during busts And there's other papers which has talked about related things Which is deposits tend to flow into banks at low interest rates and flow out of banks at high interest rates Because of the substitution between cash and deposits deposits are sort of a worse deal when nominal interest rates are high and One thing we see recently in this tightening cycle is rates have risen really really fast We've knocked out a whole bunch of the banking system But the substitution channel seems to be a little bit slower than usual so the path dependence in this paper I think so the paper is accepted at a journal now I'm not going to nitpick on small details But I think it's raising this broad issue for the literature that we really need to understand Dynamics and state dependence in terms of how deposit quantities move So let me just quickly show a raw picture. This is from the recent tightening cycle in the u.s Usually there's like ballpark half of the increase in the Fed's funds rates passes through to savings deposit rates Here it looks like it's effectively nothing Which I think tells you that the speed of tightening really matters and not just the level of interest rates And I think the details of this are complex, but the fact that this is true I think you can just sort of see very quickly in the data Let me just quickly get back to the sort of more micro details of my thoughts here So Verral has another paper showing empirically that he mentioned that aggregate deposit quantities seem to be responding to quantitative easing and quantitative tightening I have some related work with the eming where actually we find a slightly different conclusion Which is that the crowding out of lending to firms is also very important So there's multiple dimensions in which a bank can adjust their balance sheet So a bunch of reserves get thrown into the system and the bank has to figure out. What do I do? I could raise equity, but that's very costly. I could raise more deposits. That's what this paper finds I could lend less to firms. That's what we find is somewhat of the response or related to some of Wenschen's work I could say the synthetic dollar lending I'm doing in the CIP deviation market I could say well reserves are crowding that out instead and I might adjust my arbitrage trades instead So I find the paper plausible and interesting here But there's a laundry list of different ways bank balance sheets adjust to shocks because banks are so complicated in the modern world And I think to understand how severe the mechanism they has in mind We have to have some sense of for each dollar of reserves How much substitution is there in these various dimensions? And I think there's sort of some difficulty of coming to consensus about that in the literature And it's really important to understand this idea of sort of attention and switching costs here because it might be the case that for a Small shock we identify a relatively low elasticity and that's sort of behind some of our empirical results Whereas for a large shock like the crises they're studying here You actually might find these large responses, which would be more in line with the theory paper here So for the literature, I think that's sort of something which needs to be understood The model then extends to have both an asymmetry between QE the injection of reserves and then QT the withdrawal of reserves and The version of the draft I was looking at basically what happens is there's reserves s not Tao s not some fraction of that are taken out and then in reduced form a fraction Tao d s not of that is a reduction in deposits where there is a difference between the two So reserves go down by a dollar maybe deposits go down by 50 cents or something And there's this remaining gap in the balance sheet which needs to be filled So the bank needs to rebalance from holding reserves to other assets And what happens here is the bank is going to be holding securities instead in their model and because securities are less liquid than reserves You can't serve reallocate them so easily by lending them in the interbank market If we first inject reserves lead to this big growth in deposits Then we withdraw reserves the aggregate supply of liquidity on the asset side of the balance sheet shrinks But the liabilities with the deposits are effectively stuck there And that makes the financial crisis worse if we hadn't done QE or QT in the first place And this actually gets back to a notion of dynamics of what we really need to understand here is the dynamic process By which when you do policy over time banks first expand and then they contract And I find the idea here really interesting and thought-provoking I think empirically this is complicated hard to identify but as a sort of checklist for the literature This is something we really need to understand too It's precisely this kind of state dependence that makes things matter And what we need to know is do what extent do depositors keep their money in a bank due to adjustment costs After that went in in the first place And I think what makes this a little bit complicated is there's multiple types of deposits here We could think that there's retail deposits which are slow and sticky slow to enter slow to leave There might be wholesale deposits uninsured deposits which are fast to enter and fast to leave And what I haven't quite figured out is do we need there to be a type of deposit which is fast to enter and slow to leave or is enough for their mechanism to just have this heterogeneity of different types where the Sleepy people gradually enter as the stimulus becomes stronger and stronger And then when there's a sharp tightening the wholesale money runs out and then their mechanism holds Or do we really need it to be the case which I think is a little bit tougher to have some notion of the asymmetry of sometimes people are sleepy Sometimes they're awake within a subclass of deposits And I think the papers serve as you know high level abstract and doesn't zoom into these details But for the literature going forward. I think this is a question. We'd like to know something about The last thing I'm going to talk about is the paper's normative implications where it has a really interesting and kind of Counterintuitive at least at first to me notion of what the externality is So Jeremy has papers based on fire sales where why do we think that there's an inefficiency why we need to do financial stability? I get in trouble. I have to sell my assets that pushes down the mark-to-market price That's going to affect your bank regulations your service collateral constraints to and therefore you might want to regulate ex ante Things actually kind of go the opposite direction here where the friction here is about entry into the interbank market where what's going to go on is Imagine that there's a really high interest rate lots of money to be made by doing these interbank loans to save the failing banks That's going to say I don't care if I scare my depositors, and I don't get these inflows I'm going to pick up the free money on the table and for every bank that does that They're going to bid down the equilibrium interest rate on interbank lending The free money is not going to look quite as attractive and what's going to happen then as other banks on the side will say Well now actually the spread between a risk-free interbank rate and the rates at which I borrow is not so attractive anymore I'd rather sit out on the sideline and therefore get the flight to quality flows from Depositors pulling out of the risky banks and coming to me instead and this actually tells you the banking system in certain sense should be taking More liquidity risk not less And that I think potentially depends upon the details of the assumptions here But I want to just point out that the actual frictions here are unique and counter-intuitive and tell us some interesting stuff about What it's really about is incentivizing banks to lend in the interbank market is the inefficiency expo in a crisis So the paper does this in a fairly reduced form way And what I'd like to know about this and more detail may be an additional work by people in this room is How exactly does this stigma in a crisis work? Is it the case for example that if these interbank loans before they're kind of scary? And as more people enter sure the interest rate goes down, but the degree we scare people goes down, too We could easily flip that result potentially and then lead to sort of crowding into interbank lending And if I know everybody else is saving the bad banks, I want to help save them, too But because things are sort of reduced form here with the sort of amount of stigma of lending that's being held fixed I think this makes a sense a lot in terms of understanding what happened in the Silicon Valley Bank crisis It's aftermath where what seemed to be happening is these small banks which might not be so too big to fail had outflows of deposits Which then went to the big banks even though the big banks were not paying attractive interest rates at all I showed you there's effectively zero savings accounts rates that this banking system was paying and What could potentially be going on here is that? Regulation might have something to do with this it's reduced form in the paper But I'm inclined to say that these big banks who are too big to fail could potentially understand I don't want to save these failable banks because I need to keep my regulatory capital requirements aligned and Then there might be a sense in which actually this is not just about QEQT monetary tightening But actually some sense in which we need to think about whether ex post We should let the private sector do some bailouts if the public sector doesn't want to do them on their own the last thing I want to mention is The model I think also implies that if you do the COVID checks which lead to a huge increase in deposit supply That might have some financial stability implications too So I have a paper showing that this has something to do with you know There's too much nominal equity that leads to inflation for sort of standard monitorous reasons and caused a housing boom But this paper if I interpret that episode in their model says because the deposit increase here might actually be even bigger than some Quantitative efficiencies and events it might have some persistent negative effects on financial stability that are even worse So let me quickly conclude this paper presented a simple and tractable tractable model of the downsides of reserve supply for financial stability it models rather reduced form some interesting ways that there's this Dynamics or state dependence and how monetary transmission works understanding empirically how that is is I think really important for the literature going forward and just big picture as the complexity of central bank policy grows more regulations more different ways of Injecting things we need to understand more and more micro details of how these frictions works to do it Well, and the other possibility would be as was raised yesterday This notion of Mari condoing our policy regime simplify it only keep things that maintain joy and therefore We actually don't have to know exactly the details of how every little bit of plumbing works to miss these unintended consequences Thank you will I'm afraid I have to literally eat into your lunchtime a little bit at least So I'm the floors open for you questions. I don't have questions online at present So Matthias, hello Hi, Matthias German from the BIS I have two questions for you Viral One is you always, you know the introduction is you you say, you know you buy assets from non-banks And then essentially but later in the model you talk the bank has to finance the reserves with With with deposits so if I think about in the moment the central bank buys these non-bank assets It automatically creates reserves and deposits on the the bank's balance sheet because the non-bank can't have reserves So the question is then why does the non-bank actually continue to hold these deposits and rather buy something else? So which is maybe it's the same in equilibrium But I would be just interested in your thoughts and then you've mentioned and will talked about it as well this Asymmetric reaction, you know once you shrink QT and you said that there's a sluggish response by deposits And it's simply because the bank does an asset swap so he keeps Part of the QT if I understand you correctly But maybe you can just expand a bit more because this asymmetry in the data is very puzzling and in the model would think Initially, you know it should be symmetric There's another question there. Thank you very much. Thank you very much for this interesting presentation I want to expand a bit on the previous previous question So I was also wondering to what extent you introduce in the beginning these mechanical effect of buying assets from non-banks And thereby creating deposits versus like the more endogenous effect by expanding the balance sheet You also incentivize banks to create loans which create deposits what it is distinction is important for your model Okay, okay, so I take one more so more eating into the lunch But then we are we're done. I then is only the answer left I was wondering so that the key thing is that when some banks get into trouble some of the others your JP Morgan want to hold liquidity instead Which kind of fits SVB really well, but then SVB wasn't a systemic crisis, right? It was relatively small So I'm wondering if there's a non-linearity here where if things get really bad if there's you know a large proportion of banks are in Trouble maybe nobody can really afford to hold liquidity anymore and maybe the arbitrage you can do by By lending to some of the stressed ones becomes too attractive. So I was wondering how you It's bit outside the model, but but how you would think about that Yeah, let me add me quickly something to these things. So so maybe you can comment also Viral on what all this implies for a potential action as a Lender liquidity provider of last resort by central bank exposed So, okay, try to be concise if you may to the benefit of our lunch breakers Yes, absolutely Thank you will for a very great discussion. I'll come back to Will's comments Maybe at the end just start with the questions that were asked first So Matthias your question is a good one So what happens in the model is that the first round of transaction happens with non banks But we give banks effectively the ability to optimize the capital structure if they want So even though the deposit comes first from the non banks They can go out and raise equity and what happens in general equilibrium in these banking models when equity is raised Is that equity raising has to happen at the expense of deposits if you want it to matter It's just at some equity holders are buying other equity That that doesn't sort of change anything in the model So that's the way to think about it Which is that the banks can reoptimize their capital structure after the first round has happened But then you asked why don't non banks? Why should non banks hold the deposit so it could also happen from that side So maybe non banks then engage in maturity transformation. They buy corporate bonds The their deposit then gets converted into corporate deposit The corporate may not want to hold cash. Maybe they go and make investments They pay salaries the salaries then go into bank accounts The assumption in the model is that when reserves happen after all these round of optimizations There is still a creation of deposits now whether it's the financial deposit initially It's the corporate deposit in the round one whether it's a retailer and in short deposit in round two We don't have much to say on that because we don't have those amplification mechanisms And that's an important point which is that if QE is actually creating some amplifier Then the stock of deposits may even increase more than one-for-one with reserves And then it's even easier to get fragility note that in the model the way it's structured Deposit can only increase one-for-one in our benchmark model in an extension We have a sort of reserve requirement that gets relaxed when you inject reserves And then there's a multiplier of deposits that gets created then it's even easier to get fragility I think I think the lending point is quite important And this will also allow me to discuss the point that will mention which is that I didn't stress this but in the model There is a contraction that happens in the real lending and the way it happens is that when QE is very large It will be the case that equilibrium into the interbank rate at date one rises That's R1 so when R1 increases the rollover risk on deposit increases So term premium have to increase because the bank optimally says it Rationalizes that yeah if I take on a deposit there's a rollover R1 cost at date one with some probability So if I'm making a term loan to the corporation I have to factor that into the term premium and actually the lending therefore decreases because firms demand fewer Quantity of term loans what happens in the model is that the way it's written the size of your lending book doesn't affect the likelihood of the run And that's the reason why it doesn't play any role except that QE When it destabilizes at date one it also has a muted economic impact at date zero because it actually shrinks this Which is very much consistent with the result that will emig and Zhang Yang have in their excellent paper which which which perhaps preceded some of what we have done Couple of other points that were asked yeah SPB was not systemic But what's going to happen is that if the shock happens to a larger part of the system Then there is also a bigger flight to quality flow into JP Morgan So it depends on how you think of the convenience yield if the market power from becoming Super abnormally large is very high then it's not the case that JP Morgan will want to start lending They'll say if 30% runs are taking place I'm going to become 30% larger and so the benefit of that is also increasing because our convenience yield is modeled on each Unit of extra deposit rather than even if it's diminishing returns to scale, you know as long as it's increasing That's still going to be the case in the model I think it's a great idea to think about fiscal stimulus in this kind of a setup We haven't done it yet But I think the model has a lot of ingredients to make it work because you have firms You can model the fiscal stimulus via some tax collections of the government on the corporate output And then that could be getting injected into the system as deposit So you can actually close the model and start looking at it But wills intuition is exactly right. Which is that if there's an exogenous stock of deposits that you create And it's just going to be some lump sum taxation on the corporate sector Which for simplicity you can assume doesn't affect anything else you will actually increase fragility in this model Now fiscal stimulus may have some amplifier effects Then you have to build those in to do some kind of welfare analysis one last point I think which is interesting is that Why do we get this? So I didn't realize this until the last couple of weeks But it's actually the case that what what looks like an asymmetry between QE and QT is actually not an asymmetry So let me just take two minutes if I can just say this Philip Which is that it looks like in QE the transaction is happening with non banks So non banks are tendering and in QT banks are swapping their reserves for for securities What is symmetric about both of these is that banks don't want to contract their balance sheets Okay in QE. They are expanding their balance sheets and in QT. They don't want to shrink it So they are always choosing to have an extra unit of deposit if they can in their balance sheet Either by acquiring reserves in QE or by swapping reserves for securities in QT And it turns out that they ignore the fragility because they don't realize that they are all collectively Taking on deposits is creating fragility in the first place But they realize that if I have an extra unit of reserves I can park it around in the system with some probability and become actually Either the acquirer of flight to quality or the interbank market abnormal gains in the case So actually it turns out in the model that what looks like a symmetric Mechanical operation of QE and QT is a symmetric response of the banks not to shrink their balance sheet size last point And I'll absolutely stop with that which is I think the main policy implication of this model as I see it It may not be the best model to do capital requirements analysis because we don't have the kind of effects that's time 2012 has for example But one thing that I have been thinking about through my experience at reserve bank of India is that We implemented liquidity coverage ratio more as a band, which is that it's not a daily requirement It's to be implemented over a fortnight And surplus banks can actually deviate from 100 percent up to 80 percent if they want on or 90 percent And we vary this band based on the condition of the interbank markets So what so effectively what that does is that surplus banks are allowed to do arbitrage And then within a fortnight, they still have to come back to an average of 100 percent Okay, and so that makes the reserves a little bit more mobile It's not going to be a perfect solution because the convenience seal will still be at work But at least in practice it could be that if reserves are getting held up for regulatory reasons Then you can relax that by having a fortnightly averaging of these liquidity coverage ratios Rather than requiring them be met every single night because then you run into this good art problem That there are taxis at the station, but there always has to be one taxi and therefore no one can use it Yeah, thank you so much. Will there was really a great and transparent discussion It gives me some thoughts also to improve our exposition in the paper. Thanks so much Very good. Excellent. So, uh, that's not fair to say that's my last point And then now it's really my last point and now it's really really my last point So but anyway, so I think we saw a great they had a great session to wonderful papers and very good discussion So we should thank all the speakers with a resounding round of applause