 So, good afternoon and welcome back to the afternoon sessions of this conference, the monetary policy conference. There will be also a keynote speech later in the afternoon. But I'm very happy to be chairing this one because there are two papers that are very, first very relevant, very timely, full of fresh insights into something that keeps Central Bank busy these days. So, they have one common theme is liquidity, how much banks need in terms of liquidity, how much a Central Bank should supply, how much is too much, how little is too little is a very relevant question for at least two reasons, or two processes that Central Bank are now busy on. One is the reduction of balance sheets, so how far Central Bank should go in that process and at what pace. So, relevant questions asked there and also good answers that can inform that process. But the other one is also the long-term vision on an operating system for a steady-state operating system. So, the set of procedures by which Central Banks implement policy. And there again you can do it with more or less liquidity. So again, very, very nicely done papers that give you a structured way to think through these issues. Now, I'll give you, I'll give the floor first to Viral Acharya, NYU, Stern's School of Business. On a paper that you wrote with Raghurajan and other quarters. And I understand it's a new version of the paper of the Jackson Oil paper. So, as I see very much already quoted and referenced. So, you have 35 minutes, then I'll give the floor to Lorenzo Burlon with your discussion for another 15 minutes and then 10 minutes of discussion. Then we'll come to the second paper, but I'll talk about that later. David, welcome, David, by the way. And Annette Vising Jorgensen brought this paper, very, very nice one. David is Senior Associate Director in the Division of Monetary Affairs of the Fed Board. And your discussion is here, it's a real quayet of the ECB. So, you see we put the ECB people as discussants so that we can vet your paper in death and extract the most out of them. So, Viral, please. Thank you. Thank you so much, Massimo. And thank you to organizers for having me here. Let me just check that this is working. Okay. So, this joint work with Rahul Chauhan, Raghuram Rajan and Sasha Stefan. And as Massimo mentioned, this is an extension of the work we did at Jackson Hole, which was presented last August. I think there was a fair bit of skepticism, maybe for the right reasons, that there was a sense that banks are fine and that we were sort of barking up the wrong tree. We should have focused on non-banks. So, maybe we have to thank Silicon Valley Bank and First Republic, etc., for making the paper a little more relevant. Okay. So, what we're trying to understand is why is it that even though we have injected such huge quantities of reserves into the financial system, it seems that we are still talking about injecting more. You know, we seem to be on and off getting hit by financial conditions that seem surprisingly fragile. Maybe the bank failures were the most extreme of the lot. But if you think about what led to the report rate spike in September 19 in the U.S., it wasn't a very big deal. There was some money that flew as advance tax flows into the government account. Some Japanese banks were closed, so their reserves weren't available in the system. But there was still more than a trillion dollars of reserves in the system at that point. And yet, you know, there was like a shortage of liquidity in the interbank market this time in the repo markets. Then, of course, we had the pandemic when there was a huge dash for cash on bank credit lines. And, you know, shadow banking markets froze and that stress partly got passed on to banks. We had turmoil in the UK gilts where pension funds were struggling for liquidity, even though their positions were fundamentally hedged. And then we had the bank failures maybe to top it off all. Okay, so we are starting off with the premise. It's just meant to be a rhetorical question that maybe this financial fragility is linked to the fact that central banks have expanded their balance sheets a lot. And before they complete their contraction or waning of the balance sheet, shocks come and hit. And, you know, what is it in this combination of the expansion and then the plan contraction and then interim shocks arriving that leads to financial fragility. We are trying to make sense of all this. And our starting point is our theory paper, Ragu and I did a theory paper in 2021, where we were basically saying that, you know, when central banks embarked on QE, they approached it very much from an asset pricing perspective, which is maybe there are some balance sheets or assets whose prices are dislocated. Why don't we go and have a meaningful impact on these asset prices? Maybe that will change the portfolios of banks and financial intermediaries. Maybe create some good search for yield and we'll get the economy up and about again, having hit the zero lower bound on the interest rates or effective lower bound. Our main insight in the theory paper was that maybe this is a very limited way of looking at central bank balance sheet expansion because it just focuses on the asset side of the financial system, but doesn't really say anything about how QE and then quantitative tightening are altering the liability side of the financial system. Because if you want to think about financial stability, ultimately we have to think about the demandable claims on the financial system. So the key insight that I'll first try and establish, which was an assumption in our theory paper, is that reserves on the balance sheets of commercial banks who typically hold the reserves are financed with demandable and typically uninsured demandable deposits. And that is actually our main point even though maybe at central banks there's a broad acceptance of this point that quantitative easing is not just an expansion of the central bank balance sheet, it's also an expansion of commercial bank balance sheets as I'll try and explain. So if you think about a traditional banking system in which all assets are held by banks, so banks on their capital structure, let's say, have some securities and they have their checking account at the central bank, I'm just calling it the Fed here, and there are some reserves sitting in there and they have some liability structure of deposits and capital. So when the central bank does QE, it's purchasing a security in a world in which all treasuries which are not held by the Fed are held by banks, this must happen as an asset swap with banks. And so the mechanical operation of QE is to transfer a security from the balance sheet of the banking system to the balance sheet of the central bank. The central bank credits the checking account of the bank with a reserve and at least the mechanical operation of QE doesn't need to expand the balance sheet of the banking system. Now it could still be the case that having shortened the duration of its assets, banks go and then shorten the duration of their liabilities, but at least that's not happening mechanically as a result of quantitative easing. It turns out that in practice in modern day financial system when treasuries are held not just by banks, but they are held by non-banks as well such as insurance companies, pension funds, family offices, high net worth individuals, mutual funds, hedge funds, et cetera, et cetera, that QE doesn't exactly work like this. It instead looks like an asset swap with the public or the non-banks, okay, which is that. So now there's some non-banks in the system. They hold some treasury securities and some cash, which I'm going to call as deposits in the banking system. And it's banks that I'm assuming who are the ultimate holders of reserves in their checking account with the central bank, okay. So now when the central bank does QE in the modern financial system, it's the non-banks who also want to participate, maybe because there is a price impact and they want to benefit from it or maybe the yield curve becomes too flat and they don't want to hold on to the securities. We are not going to model these motives, but they tender their securities via the banking system as the prime brokers to the Fed. The Fed of course credits the bank with the reserves, but because the ultimate tendering of securities is from non-banks, non-banks now have an extra deposit into the banking system, okay. And because these are relatively large transactions, because these are transactions with institutions, typically in the first round these deposits are going to be uninsured demandable deposits, okay. Now these non-banks may turn around and search for yield. They might say, why should I sit on so much of deposits? Why don't I convert my deposit into a long-term bond that a corporation is issuing? But you know, you have to now think through the second round, what is the corporation going to do with the bond issuance? They will distribute some salaries or make investment payments. Ultimately, some uninsured deposits are going to come back into the banking system through that operation one way or the other, okay. Now the key question then is, if you did expand the commercial bank balance sheets in this way and with uninsured demandable deposits, when you do the contraction, is it going to be benign? Or somehow, because you've expanded the stock of uninsured demandable deposits, maybe reserves are not going to move around very smoothly in the system. Are you taking on some financial fragility risks, okay. And in case the financial fragility manifests before the central bank has completed its operations, and I've always been puzzled by this forward guidance of exit policies because, you know, shocks are not going to respect the timetable of the central bank. At least that was my experience when I was at the Reserve Bank of India that I can have whatever timetable I want for my open market operations. The fiscal policy, dollar interest rates, exchange rate shocks to India, they didn't actually respect my open market operations timetable. They arrived at their own pace, at their own whims and fancies, and then we had to deal with all these shocks. And so the problem then is that if these financial fragility shocks occur before the four or five year of waning cycle that the central bank has decided for itself, it's not going to do the waning in the first place because once the fragility shock occurs, there'll be a ratcheting up in demand for liquidity. The central bank will inject more, and what was one trillion will become four trillion, and then it'll become eight trillion, and they're just sort of expanding the liquidity dependence of the system overall. Okay, so the sorts of questions that we try and answer empirically are how does QE work? Is it really working with non-banks as I showed you? Demandable deposits are of course getting created through mechanical operation of QE, but once banks are flushed with reserves, typically earning low interest rate in times of QE, do they then try to seek to sell these reserves and create more demandable claims such as credit lines in order to earn some extra fees on these reserves that they have? So, uninsured demandable deposits, credit lines which are also demandable from corporations or from individuals if they're credit cards. What happens to these claims when the central bank undertakes quantitative tightening? Do these claims shrink at the same pace as the central bank unwinds? So is QT literally an unwinding of the liability side of the banking system as well in terms of shrinking the demandable claims back to their original levels? Or is it that you shrink the reserves but the demandable claims still remain in the system so that then you could have some shocks? So that would be a time series problem. There could be a cross-section problem which is that maybe once you inject the reserves and demandable deposits that created, depending upon different banks' incentives, some banks now use their reserves to originate more assets and end up with more less reserves and more liabilities, but some other banks end up with more reserves and less liabilities, then the interbank market has to function very well when shocks arise, which may not function if there are hoarding and other kinds of incentives. So there could be financial stability implications if the time series of claims and the cross-section of claims doesn't necessarily work out the way QT could happen in a smooth manner in the market. So in interest of time, let me just quickly mention the results and then start showing you a few things which is that we're going to find in time series there is some hysteresis, which is that QE happens with non-banks, but when quantitative tightening is undertaken, it doesn't fully reverse the operations. So what I would ideally like to do is that if central bank purchase securities from an insurance company so that that company's deposit with the bank gets extinguished when quantitative tightening is done, I want to do exactly the reverse of whoever tendered the securities to me, because then I can be sure that they took out their deposit from the banking system somewhere and used it to purchase the securities back from the Fed. But these transactions are open market transactions. They are not designed as asset swaps with the same counterparty. And so there's some hysteresis, which is that the demandable deposits remain with the banking system and that's because quantitative tightening looks like the first operation which was an asset swap with banks. So what we find is that in QE, it is non-banks that tender to the Fed, but in quantitative tightening, it is banks that purchase the securities from the Fed and the demandable deposits that got created in the system in some quantitative tightening episodes remain in the system. And then there is a search for yield in the cross-section which is that relatively low capital banks in the period we are looking at smaller banks, they did actually seek out illiquidity so they were willing to give up their reserves and expand the stock of demandable liabilities even more and that kind of created fragility in the cross-section of the banks. But our main concern is that we have thought about balance sheet expansion a bit too glibly which is, okay, the situation is that ELB or zero lower bound, I have to expand my balance sheet. Of course I can just seamlessly unwind it in four or five years whenever I'm done. I think there's no evidence that any central bank is successful in doing this without arrival of interim shocks and then having to inject even more liquidity in the system which is the phenomenon we are calling as liquidity dependence and we just seem to be ratcheting up the liquidity provision in the system. Okay, so let's look at some data. So this is basically various episodes for the United States, the QE-123. This is the passive phase of the Fed when it's just not reinvesting the reserves that it gets from the government on the securities it has purchased back into the system. Then there's active QT in which Fed is selling securities in the market from after the repo market spike. I'm going to call it pandemic QE even though the pandemic part of the QE starts two quarters later. And then there is the quantitative tightening too. And of course both the QT episodes are also associated with rate hikes in QT too with a much greater pace of rate hikes. Okay, so these are reserves with the commercial banking system. This is not the total size of the Fed system which would include government account balances, cash, etc. So these are just the reserve balances of the commercial banks. So every time QE is done, reserves expand. The actual QE stops before each of these lines. So there's some reduction relative to GDP in the reserves. And then in the passive and the active QT phase, you can see that the reserves relative to GDP are coming down. Then a massive injection of reserves occurs at the time of the pandemic, both due to fiscal and QE part of the Fed. Then QE continues in the background even though fiscal sort of stops. And then you have the reserves shrinking again during QT too. So first, what did the deposits of the banking system look like? Over this period during the QE and the post QE phase, the deposits of the banking system kept growing from being at about 50% of GDP to about 60%, about 10% GDP increase. And then you can see that in QT 1, which was the first significant quantitative tightening, the deposits are very, very stable and practically flat. So even though the QE part looks like an asset swap with non-banks, this part looks very much like an asset swap with banks because bank balance sheet size is not changing during QT much. Here you can see very strikingly the expansion of demandable deposits as I'll show you later. It's not just deposits, it's actually demandable deposits with the pandemic arrival. And then unlike QT 1, QT 2, because perhaps of the size of the rate hikes, the overnight RRP mechanism and then the bank runs, the deposits are actually leaving parts of the banking system and finding their way into non-bank parts of the banking system. Interestingly, banks are also selling credit lines during QE. Once again, they are very, very stable. So even though reserves are coming down, the two demandable claims which are in the green line are at almost the same percentage of GDP. And so you can see that in some sense QT is automatically worsening the liquidity position of the banking system because demandable claims as percent of GDP are remaining flat, but the reserves with the banking system are coming down. And then if you break these up, at the top I have the two demandable deposits. So the thick black line is uninsured demandable, the dashed line is insured demandable. Time deposits are practically irrelevant over this period because of the relatively low rates. And you can see here that both in the initial phase, both insured and uninsured are rising. At the time of the pandemic and the fiscal stimulus, again both insured and uninsured are rising. However, as QE continues during the pandemic, it's only the uninsured deposit component that keeps rising and then eventually comes down. Okay. So the question is, can one put some empirical structure on all of these descriptive simple graphs that are out there? So first I'm just going to show you what happens in the aggregate that just taking the figures and building some estimates around the elasticities of these quantities to reserves and then following the work of David and Annette Wising-Jorgensen, I want to show you a little bit on what that's also doing to price of liquidity in the system. Okay. Then to build a little bit of causality, I have to go to the cross-section of banks to understand behavior, but then the problem is in the cross-section, reserves are not exogenous to a bank. They are exogenous to the banking system, but what reserves each bank holds is, of course, its private, endogenous decision. So we have to deal with the endogeneity of reserves a little bit. Okay. And then finally, if there is time, I'll talk a little bit about what does this mean for the buildup of fragility that might have led to March 20 and March 23 COVID and then SVB kind of episodes. Okay. So first aggregate evidence, very simple regressions of just quantities of deposits or credit lines, either as arithmetic changes or in preferred specification, log changes to deal with stationarity, just trying to explain them in terms of the changes in the aggregate reserves of the banking system with some lags to control for seasonality, but not very crucial. Okay. So if you do this, what you see is that deposits have an elasticity of about 0.14 to 0.18 on the reserves in the banking system. Okay. So when reserves expand, deposits, and especially demand deposits expand. So demand deposits, elasticity is slightly greater than that of overall deposits. And you can see why because time deposits actually have a reverse relationship. Time deposits shrink when the aggregate quantity of reserves expands. One reason could be that reserves are very short-lived asset and so banks actually then don't want to have a short, have a long-term liability when a big chunk of the assets have been converted into short-term reserves. And you can see that banks expand their credit lines as well at a much lower elasticity, but that elasticity is partly building up over time, as you can see in the lag term. Now because the logs are a bit hard to interpret because it's an elasticity, so you have to convert percentage changes into absolute quantities. Even though the arithmetic change regressions violate some stationarity principles, I'm just going to run them just to understand the magnitudes. And what you see here is that deposits change basically one-for-one with an expansion of reserves in the banking system. In fact, demandable deposits expand more than one-for-one, which is not what you would expect through a mechanical effect of QE. So there's some second-round effects going on and you can see here that that's because time deposits are actually shrinking and so some of that is showing up in the growth of demand deposits and credit lines are expanding by about 15% for a dollar injection of these reserves. Now you can break up the demand deposits into uninsured and insured deposits. So these are the overall uninsured and insured and then the demandable deposits are then broken up into uninsured and insured. Once again, one-to-four is in lock changes, five-to-eight is in arithmetic changes. And what you see is that bulk of the effect is actually coming through uninsured deposits. So reserves are associated with an expansion of uninsured deposits and uninsured demandable deposits have much greater elasticity than insured demandable deposits. And once again, if you go to the overall quantities, you see a similar pattern in data. Okay, so that's the impact of reserves on the quantity of demandable claims. Now the question is one way of thinking about the liquidity stabilizing role of reserves is that if I inject a lot of reserves in the system, it should reduce the rate at which banks are willing to exchange reserves with each other because I'm creating a surplus reserve system, banks should really be willing to transfer reserves to each other at very, very low cost because no one really needs such a huge quantity of reserves. Now you can immediately see that the fact that you inject the system with reserves, you think you are creating surplus liquidity is already assuming that reserves are not altering the liability structure of the banking system. Okay, because if you are injecting one dollar of reserve and you are creating one dollar of liability, it's not at all clear that you've actually created surplus liquidity in the system because you have a claim of the same magnitude that's runnable on those reserves. Okay, so what David and Annette did is this interesting regression in which they looked at the Fed funds rate relative to the interest on reserves as the price of liquidity in the system. A stabilizing role of reserves would imply that this coefficient alpha should be negative and then they also looked at the role of deposits over here in our view of the world, in our theory work with Raghu as well as the empirical work. Deposits, especially demandable deposits are like an encumbrance on the reserves, which is they are also a claim on reserves. So just because I expanded reserves, if deposits expand, the liquidity stabilizing role of reserves should be compromised because you're just not creating free liquidity in the system. And then we add a little bit of credit line related measures as well. Once again, because these are large aggregate quantities, if you prefer, you can run these specifications in log changes to deal with stationarity a little bit better. Okay, in any case, what do you find is that this is the main result of David. I don't want to steal his thunder, but that this relationship is very flat. But once you adjust for deposits and in our case also credit lines, then you recover this negative relationship, which is that in order to understand the impact of reserves on effective Fed funds rate, you have to recognize whether deposits are getting created in the system. And you can verify this in a table, but in interest of time, I'll skip that. Now, what I want to do next though is to see how the slope of this line, which is on reserves, and then there's also when you run the regressions, there's also a slope on demandable deposit. So let me just spend time on one spec here, maybe say table 8, column 8, in which reserves have a negative slope, and that's controlling for uninsured demandable deposits that have a positive slope of 0.19. So this says that if I create $1 of reserves, but that's associated with the creation of $1 of uninsured demandable deposits, it's not at all clear that I'm having a very stabilizing impact on the Fed funds rate at all because the two effects are going to offset each other. Now, you can estimate these coefficients over time in different periods in quantitative easing times, in quantitative tightening times, and so on. And why are we doing that? Because we're trying to question this basic view that central banks have had that if we create a surplus reserve system by QE, they're actually reducing the price of liquidity in the system. Maybe in normal times when the demandable claims are not coming due, there are no fragility shocks, maybe you get this effect, but maybe when shocks arise, you could get the opposite effect of the supply of reserves. So the traditional intuition is that supply of reserves will push down the price of liquidity. Our intuition is that no, because there'll be new liquidity claims that have come up, uninsured demandable deposits, credit lines, and this endogenous demand for liquidity when certain shocks arise like QTE or fiscal shocks, et cetera, can now push the price of liquidity in the other direction. So it's not at all clear how the net effect may play out. So what we did is we just estimated these coefficients on reserves and uninsured demandable deposits. The second side is once again effective Fed funds rate minus the interest on reserves and we just plotted it basically quarter by quarter using lag data. These are rolling coefficients and what you find here is that, especially during active QT periods, the price of reserves after controlling for uninsured demandable deposits becomes more negative. So an extra dollar of reserves in this time is going to have a very big stabilizing impact on the Fed funds rate. Now some people look at this and say the recipe is to inject more reserves into the system. But what they're not realizing is that simultaneously the coefficient on demandable deposits is going up as well. So if you have more demandable deposits in the system that's a bigger claim to liquidity, that's going to have an opposite impact on Fed funds rate and cause the rate to rise. So if I see these coefficients and interpret this as scarcity of reserves, if I do QE and inject more reserves and if that creates more uninsured demandable deposits, I will not get much of an impact because I'm actually completely unwinding the effect through deposit creation in the system. In any case, what's interesting here is that the coefficient on reserves becomes more negative during QT. The coefficient on demand deposits becomes more positive during QT, essentially saying that liquidity stress is a feature of this quantitative tightening episode. So then we go to our panel tests. As I said, what do we have to do in the panel? We can't treat reserves as exogenous for a bank, so we need to do some instrumentation. The simplest instrument you can think of is that the central bank controls the aggregate supply of reserves. No individual bank does. So if because of my position as a primary dealer or my relationship with non-banks, whenever central bank injects reserves, if I can get an estimate of my reserves beta, which is that if central bank injects a dollar of reserve, what fraction of that reserve comes to me as JP Morgan? What fraction of the reserve comes to me as Silicon Valley Bank? That's my reserve beta. Then I will know how much is an exogenous shock to my reserves when central bank injects liquidity, because given my positioning as a prime broker, maybe I get a certain fraction of the reserves whenever the central bank injects those reserves. So we do two shocks. We look at overall shock to the commercial banking reserves. We also look at overall change in the balance sheet of the Fed as a whole. The first instrument suffices by itself. Then we look at what has been the past four-quarters share of a bank in these aggregate reserves that are getting injected. So think of this as the share of each bank in the past, in the recent quarters, and then this is a shock to the aggregate reserves. That's going to be our reserves instrument. We verify that it sort of statistically works in the first stage. Note that by construction, these reserves instruments add up to one across the entire banking system, because someone or the other has to get the reserves when the central bank injects them. Okay. So now we can do the impact of this exogenous component of reserves on the uninsured demand deposits in the cross-section. And what do we find? We get that during quantitative easing, there is a coefficient of about 11% to 12%. So that's similar to the elasticities we had seen in the log differences earlier. And this effect is driven actually by below-median equity capitalization banks. So banks which are seeing greater growth in their demandable deposits with central banks' injection of reserves tend to be the low-capitalized banks. And we interpret that generally as a search for yield by these banks. In contrast, if you look at quantitative tightening periods, which is post QE and then 14 to 19, we pretty much don't have any impact. It's statistically insignificant. And if anything, it's negative sign, but I'm just interpreting this coefficient as zero. And so it says that when reserves expand, uninsured demandable deposits expand during QE, but there's no similar contraction during QT, at least not during the first QT. Now, two quick things. First, is this active? Is this just deposit creation? Or are banks then also actively trying to reduce the duration of their liabilities, given that QE has shortened the duration, created a huge short duration asset on their portfolio? So what we find is that especially the banks that have market power and can alter the maturity of their deposits, they seem to shrink the term structure or term spreads on their deposits so that they can get more demandable deposits into their balance sheet. So they are matching low duration reserves with low duration liabilities. And they're also selling more credit lines during quantitative easing, especially to non-investment-grade firms. I'm not going through the magnitudes of all of these effects. So then what happens is now we have to understand the fragility consequences. I may not be able to show you all the COVID and the SVB parts, but I just want to show you the ratcheting up that gets built up along the way. So why is it, like, where do I see that during, as you transition from QE to QT, these risks are actually materializing? So why is it that banks are not shrinking their liquidity? Who does not shrink liquidity? And what are the consequences? Where does it show up in the liquidity risk in the balance sheets of banks? Okay, so first things first. If you just look at the growth of uninsured, demandable deposits relative to assets in the U.S. banking system, I've divided banks by above 250 billion assets in the previous quarter, 50 to 250 billion, and below 50 billion. These are the cutoffs by which the strongest liquidity coverage ratio is applied, the moderate version is applied, and no liquidity coverage ratio is applied. And you can see here that, broadly, there is an overall ratcheting up of uninsured, demandable deposits. It's the strongest for the smallest banks in a proportional sense. But of course, in terms of quantities, it's the large banks that's going to dominate this. So that was relative to assets. Now I can look at these uninsured, demandable deposits relative to reserves. And what you see is that during QE, both uninsured, demandable deposits and credit lines come down. But during quantitative tightening, this fragility ratios start rising up again. Again, during QE, they come down, and during QT, they start rising again. Okay? Okay. So now, the question is, why is it that banks are not shrinking their deposits? So one possibility, as I discussed, is that maybe because during QT, they are doing the asset swap. So they are saying, okay, I have these deposits. I'll give you reserves to the central bank and let me get the securities back because that way I can always tender them back to the Fed in a lender of large resort and get the reserves back if I want. Okay? So now, even if you recognize that eligible assets that a bank owns are a store of liquidity, so now I can look at uninsured, demandable deposits and look at them as reserves plus eligible assets. Okay? And now what you see is that relative to eligible assets, the largest banks have actually been reducing their liquidity risk, but the smallest banks, which are below 50 billion, they've rammed up their liquidity risk from under one to almost the same as the other categories of banks. Okay? And so one of our conclusions is that it's perhaps not surprising that these relatively small banks are the ones that are bleeding quite badly because they are the ones whose liquidity risk has ratcheted up quite significantly. Last point, and then I'll stop. I'm just showing you the buildup of fragility. We have seen that SVB and COVID shocks didn't play out very well. So now you can expand liquidity risk to not just include uninsured, demandable deposits, but also include credit lines. So these are demandable claims and scale them by reserves plus eligible assets. I'm looking at QE to the passive phase after QE and then the active QT. And what you see is that in the cross section, the distribution of liquidity risk is becoming very, very skewed. Okay? So you are shifting from relatively low liquidity risk in the system to many banks actually having liquidity ratios which are very, very high at this end. Okay? So in interest of time, I'll let you read the paper on the stress parts. But I think our main policy implication, if I can just take half a minute more, is that QE is not just a central bank expansion. It is an expansion of the commercial banking system. I think that's our main message. And it's an expansion with uninsured, demandable deposits. And therefore, when you start doing quantitative tightening, accidents are waiting to happen because you are actually taking away a big part of liquidity that you've injected and which is now has a lot of demandable claims on the other side. So should you engage in QT very carefully? Should you revisit the scope of QE, worrying about the financial stability consequences and the possibility that it may be almost impossible for the central bank to exit? Maybe it's a Hotel California kind of mechanism. Okay? Let me stop there. Thank you. Thank you. Okay. So, well, while the slides are going up, thanks for the organizer for having me to discuss this great paper. I mean, the usual disclaimer applies, obviously. So, let's see if we see them. Yeah. Very good. So, I mean, this, like many other papers of Viral, is like a stone throwing upon them. I mean, we are here one year after the momentous appearance in the Jackson Hole where they're still floating on the ripples created by this paper. And the claim of this paper is quite clear. I mean, that expansions and shrinkage of central bank balance sheets has financial stability ramifications for small, weak, and tendentially unsupervised banks. And you can see it in this chart that luckily, Viral has decided to show us the very last in his slide where you see that if you take a measure of liquidity risk and you divide these measures by different buckets of banks from the largest in blue to the smallest in green, you can see that while the large and medium-sized banks that are also tendentially the strongest ones have this liquidity risk going down, even if there is a waxing and waning of the central bank balance sheet, the smaller banks are instead accumulating more and more liquidity risk. Now, I mean, to bring about this conclusion, Viral and co-authors bring to the table a compelling body of evidence. They span aggregate descriptive statistics to time series regressions, panel bank-level data to event studies. They look at assets, liabilities, quantities, rates, whatever, whatever, whatever. They also refine, obviously, the causal interpretation of the coefficients using an identification strategy, relying on this role of banks as liquidity hubs. And they also illustrate very clearly the mechanism behind these results. That is that banks that are weaker, for instance, that are lowly capitalized, tend to, you know, reach for yield. They try to, they have all the profitability incentives stacked up to build up this additional liquidity risk. Now, I mean, the implications of this paper, and this is my personal reading of the paper, is indeed that the risks coming from the expansion of a central bank balance sheet are there if regulatory and supervisory frameworks are incomplete. And in completing the sense that they don't cover especially the smaller weaker banks that are outside the scope of the provincial oversight. And if anything, there is a bottom-line message of the paper is indeed that the strong regulatory and supervisory framework is instrumental for the transmission of monetary policy, for the smooth transmission of monetary policy. Now, I will have, so, I mentioned, not that this is a great paper, and I don't think that on the technical point of view we have much that it has gone through a lot of scrutiny over the past year. So, if anything, I would like to bring a bit the Euro-era perspective and a bit, you know, what we can, which messages we can draw from this paper for the Euro-era. So, first, I will just look a bit of the ongoing adjustment in the Euro-era to the new environment, lower liquidity environment is emerging, which is gradual, and understandably so, but it's also broad-based. It's broad-based in the sense that we are not only witnessing a decrease or a stabilisation of the credit lines, but especially we are seeing very strong and potent adjustment on the unbalanced sheet long book of banks. Secondly, I'm going to look a bit on the nonlinear effects that maybe are rising in this new environment of lower liquidity. And third, maybe I will spend one minute at the end of the discussion to remind ourselves of which role actually this open, you know, and all these unused credit claims have played in one of the most stressful episodes we've seen in the past years, not in the pandemic. So, let me just start from just basic facts about the evolution of unused credit lines in the Euro-era. What you see here is the overall amount of credit lines, unused credit lines in the Euro-era that, as you can see, have gone up from two trillion in 2015 all the way to three trillion, all the way to the moment in which we started hiking our titanium cycle. And you see that they basically stabilised from that moment onward. Now, if we actually unpack this aggregate amount of undrawn credit lines, you can see that actually for the non-financial private sector, the yellow and blue bars, we are actually seeing a gradual, but still a decrease in this undrawn credit lines exactly coincident with the start of our tightening cycle, especially for the household sector. And if anything, if there is one counterpart of these undrawn credit lines that is actually expanding, is the credit claims from financial corporations, which might be a speculative interpretation of this, might be that these financial corporations are actually preparing themselves, they are building up a war chest to face the incoming lower liquidity environment that they are anticipating, but just justification. Now, moving from off-balance sheet exposures to on-balance sheet exposures, and here I would have you focus on the rightmost panel of this chart, you can see how massive the decrease in the long growth has been since we started hiking. It's massive in absolute terms, and it's massive compared to any other tightening episode we've seen, and it's even large if we take into account the size and pace of the hikes that we have performed over the last years, which will be represented by the dashed line. It's really an unprecedented slowdown in credit volumes in the Eurator. So the adjustment of the balance sheet since we started hiking and since we started reducing our balance sheet has been very large. Now, it's very difficult to elicit what might be behind this slowdown of credit lines of this long growth, but one thing that we can get from the soft information of banks is that if we divide banks into two groups, one with high liquidity, one with low liquidity, that's the blue and the yellow lines, they faced a very similar evolution of loan demand, but they differed in a very pronounced way in terms of a way in which they tighten their credit standards. And actually, this tightening of credit standards much more for the banks with low excess liquidity was associated with their perception about their liquidity position. And perhaps even most interestingly, the most pronounced factors behind this tightening of credit standards was associated with banks' risk perceptions. It was the low liquidity banks that have changed their risk attitudes in a most pronounced way since we started hiking and moderating our balance sheet, suggesting that there is this rationale of banks actually operating in an environment where they anticipate the challenges that Viral's paper is pointing to obviously helped and incentivized by the framework in which they operate in the euro area. And actually, if we move to model-based evidence, we can confirm this extra sensitivity of unbalanced sheet exposures of the loan book to liquidity conditions. You can see here, especially on the left-hand side, how this sensitivity of loan volumes to liquidity conditions is especially concentrated for the withdrawal of liquidity that is not necessarily borrowed by banks just to satisfy the immediate liquidity needs, but instead is a sensitivity associated with the liquidity that is coming from non-borrowed reserves, the ones that, for instance, accrue to banks via QE and that are withdrawn via QT mechanisms. Now, I've talked about volumes in off-balance sheet, on-balance sheet, and even the qualitative composition of the on-balance sheet exposures, for instance, represented by the duration of the loan book, has been evolving consistently with the quantity of excess liquidity floating around. What you see here is the excess liquidity of an asset in the aggregate in the euro area against the average loan duration of outstanding loans for households and firms in the euro area, and you can see a very clear upward trend whereby duration was going up as excess liquidity was going up, and the moment we started hiking and we started moderating the size of the balance sheet, this has just proceeded in the opposite direction along the same trajectory. Now, moving on to potential non-linear effects that might emerge from this environment, here I give you a couple of perspectives to speak a bit with the exercises that Viral has shown in his paper. On the left-hand side, you see, for instance, that there are thresholders effects whereby you do not have the sensitivity of unused credit lines to decreases or increases in excess liquidity only in a situation where the level of excess liquidity, so for banks that have high excess liquidity, is very high. This sensitivity emerges only once the banks have low excess liquidity and therefore they become actually very mindful of the quantity of excess liquidity that is back in their unused credit claims. And on the right-hand side, if anything, it's a way to connect the off-balance sheet developments with the on-balance sheet. Here what I look at is one of the most severe forms of credit restrictions that you can think of is rejections of loan applications. And if we mix up soft information of banks that tells us about whether they have increased the loan rejection rates, we can see that actually the probability of these banks reporting an increase in their rejection rates increases as a response to decreasing in excess liquidity only for the banks that have high undrawn credit lines. Again, speaking to the fact that banks are internalizing the framework in which they are operating, they have regulatory and supervisory pressures that induce them and also the market scrutiny that induce them to actually moderate very badly, very strongly, very pronounced way their lending conditions in response to changes in the liquidity environment. Now, with this, I would just like to spend one minute on reminding ourselves on what was the role of these unused credit lines. Because sometimes we are led to look at this dash forecast, these drawdowns of credit lines due to the pandemic as just this ominous sign of financial fragility of the system. But what these two charts try to illustrate to some extent is that it's true that the long growth that we experienced at the very height of the pandemic was associated with the amount of undrawn credit lines in the banking system as also Viral illustrates in his paper. But at the same time, these credit lines, the fact that we had those open credit lines in place was exactly the conduit through which support to the real economy actually arrived during the pandemic, where the amount of long growth that we have experienced over that period was exactly related to the decrease in the productive capacity of the economy brought about by the pandemic and by the restriction that we have experienced in that period. And interestingly, if we actually look at where the support, the liquidity support that we have injected in that period flowed, it flowed exactly to those sectors that were more affected by the drop in productive capacity. So to say that it's indeed, there is a financial stability aspect to it, but there are also other aspects that we should never forget when we look at these things. So all in all, this is a great paper. It's very consequential. It's a must read for whoever is interested into the bank-based transmission of monetary policy. So if anything, if I have to think of three open questions, and with this I would conclude that are left a bit open at the end of reading the paper. So the first one is what is the right counterfactual when we assess policy changes? I mean, just to, we have seen, for instance, the role that open credit lines have played during the pandemic. So one thought experiment that we should conduct is, for instance, what would have been the size of the public support measures that we would have had to adopt if we didn't have those unused credit line in place when the pandemic hit? This is one characteristic. And it speaks with the discussion that we had this morning on this tendency, and it's a very justified tendency on looking at the contrast between small, short-term gains against future large costs, in terms of risks to the financial stability. We should think of the fact that those short-term gains at times might be extremely sizable. Then the second question is actually what, it's not clear at the end what is the appropriate size and pace of the bank's balance sheet adjustment that we should expect based on the results on the paper. Is it about, I mean, we've seen how the unbalanced sheet adjustment was very sizable. Both in the euro era and in the US, if you think of the sluice results. So the idea is to understand whether, after having hiked as much as it was hiked in the US in the euro era over the last year, in an environment in which if we take at face value the results of the paper should have been more fragile than would have been otherwise, can we still think that the economy has not been tested by the hike that we have seen so far. And then maybe the last point is it would be very interesting and perhaps it's just analytically the part that I was missing out at the end to see the later stages obviously of the transmission of monetary policy, especially the real effects, that would be great to see in the paper. And with this, I would conclude. Viral, do you want to react to these points? And then I'll open the floor. Yeah, no, thank you for a great discussion. It's very interesting to think about these patterns in the euro zone as well. My overall sense is that the credit lines are not as big in the euro zone as they are with the American banks. But I broadly agree with the one point you made which I think has been my interpretation also that all the Fed and treasury policies of the pandemic, they are thought of as support of the shadow banking system. My view is that ultimately, they were actually support of the banking system because all the shadow banks were running on bank credit lines. And banks had about just through large companies, I think about 320 billion dollars of withdrawals in less than four weeks. That was larger than the entire year of bank credit line drawdowns. So I think the counterfactual of these drawdowns on bank credit lines would have really exceeded quite a bit. And I think if you look at bank stock prices during the pandemic until the vaccines came, bank stock prices didn't recover very well. One, there were losses on the existing loans. But also when a credit line gets drawn down and becomes sort of like a term loan, it's not technically a term loan because it can be repaid back as a revolver. But it has a greater capital encumbrance once it is drawn down than when it is not drawn. And so there are some papers which are showing that banks that had more credit line drawdowns didn't lend as much to small businesses in the aftermath of all this. So that would get to the real side of this thing, which is that at least in case of credit lines, when the fragility manifested, there was an immediate impact on small businesses because banks acted until vaccines came as though they were capital constrained. Then their stock prices also started doing better and then things got relaxed. But I agree. I haven't seen a good paper which is looking at the current stress of the actual runs on banks. I hear mixed things. Some people are saying, no, this is just like an interest rate pass through. Even though runs occurred, we have backstopped it and nothing terrible is happening beyond what high interest rates would have done. But I think we have to wait a few quarters to really tease out that data. Thank you. Thank you. I don't see comments or questions coming through the chat line, but I'm sure in the floor there will be some. From the Bank of Italy. So your story is about the symmetry between QE and QT in the ratio between deposit and reserves. But something else also can change this ratio and was something that happened after the COVID. And this was the debt to GDP ratio. So if the amount of new government's bonds is purchased by banks, initially you just have a change in the asset side of banks, but then when the government expands the money, these also increase deposits. And in particular, it increases the deposit to reserve ratio. So I'm wondering how you disentangle these two effects, the one that you attribute to QT and the one that could come from this side. Yeah, maybe if I can quickly respond because I don't remember questions very well. So, yeah, it's a great question. I couldn't show you this. So the data we have is only quarterly because we are using bank call reports. But what you can do is during the, say from Q4 of 19 until Q1 of 22, which is what we are calling as pandemic QE, you can see the sensitivity of uninsured demandable deposits to reserves and insured demandable deposits to reserves. And there are three quarters when fiscal stimulus took place, which is, I think, Q2, Q4 of 20 and then Q1 of 21. And what you see is that the positive relation between insured demandable deposits and reserves comes entirely from the fiscal stimulus quarters. So fiscal stimulus seems to be a predominant driver of insured deposits. And so while it is expanding the stock of deposits, it's expanding more the stock of deposits which is relatively more stable. In contrast, QE, so the uninsured demandable deposits to change in reserves is robust, whether you look at fiscal stimulus quarters or whether you look at non-fiscal stimulus quarters. In non-fiscal stimulus quarters, it's only the central bank that's doing the reserves expansion. And so I think my sense is that there was a bit of stimulus that went to companies, and that would have created uninsured demandable deposits. But in terms of the slope, it's not a very strong relationship. It's really QE which created the large stock of insured demandable deposits. We have the charts in the paper. Anna Samarina, Dutch Central Bank. Thank you for interesting presentation. In you offer the evidence that you provide, especially what you mentioned about the differences between large banks and small ones, there might be some concerns of the fragmentation raising up in the financial system. So in that sense, I was wondering whether you see the role for redistribution of reserves across the banking system and whether you observe this already in the data. As I imagine, as the QT goes on, there might be more need for redistribution from large banks to small ones with not that much liquidity. Related to that, there is this ongoing debate about the revival of the interbank market as the amount of central bank reserves shrink. So how do you see the potential and feasibility of reviving the market to support this redistribution? So I think what we are seeing is that large banks don't seem to be that keen actually to lend to the smaller banks. We saw that in the recent episode. Even during the repo market stress, we were seeing JP Morgan, Jamie Dimon openly said that I have to keep my reserves for resolution planning and LCR management. Now, they did relax the rules a little bit after that as to whether they are doing every night or every fortnight sort of averaging. But I think we seem to have effectively gotten rid of the uninsured interbank market, it's also uninsured but I meant unsecured interbank market. So I think in a way that would only increase the liquidity dependence on the central bank in my view which is that if the reserves are disproportionately ending up with the large banks and either because of regulation or market power they are not willing to provide support to the smaller banks when stress arises, then in a way it's even more the demands on the central bank to inject liquidity in midst of QT will be even stronger. And so I think it's not great for a central bank if your one arm is doing quantitative tightening. So it's selling securities in the market every day while your other arm is doing lender of last resort which is sort of injecting reserves against securities. And I think that's the sort of scenario I think that means at the end of the day it's got to have some impact on the monetary policy. Just a couple more questions, yeah. What is bringing is this in your paper, this fragility on the liability of banks. And we know by banking theory and by practice that this fragility on the liability might imply a change in the lending and the risk taking in the asset side. Have you considered analyzing this part? Do you think that this part is equally important on the fragility on the liability? Yeah, so I think this was also Lorenzo's point. I think we haven't yet traced out the real impact of this. And one reason why we have been a bit reluctant is because with quarterly data it's a little hard to tease it, pin it down very precisely. So it turns out that if you work inside the Fed you have access to the daily reserves and daily liabilities because for liquidity coverage ratio calculations they are now actually getting this overnight data. So I'm hoping to work with some co-authors at the Fed to see if we can, so one we can actually tease out some of these relationships much more tightly because when a QE happens you can see exactly that night whose deposits are changing. They also know which type of deposits are these. Are these corporate transaction deposits? Are these non-bank financial firms' deposits? They have it at that level of granularity. So you would know exactly what kind of deposits are getting created. And then I think once you are able to reconstruct the banking system's balance sheet on a daily basis after a QE shock occurs or a QT shock occurs then you can start tracing it out over the next three weeks or four weeks. How did that then impact the lending decisions of the banking system and so on. So I think it's probably best done at that level of time granularity rather than using quarterly data because the shocks won't give you too much to trace out. No, that we don't know because they are not keeping because for that you need to know when a bank acts as a counterparty of the Fed whether they're acting on their own behalf or whether they're acting on behalf of a customer. They should ideally collect the data but I don't think they're collecting the data. Okay, I have a question here from Enrico Perotti. Smaller banks may have or believe they have a more stable deposit base. Perhaps this is why they responded less to the worst liquidity position in QT. Alternatively, do you see a differential inclination to take risk? Yeah, so everything that we showed was all mostly focused on uninsured demandable deposits. So those are not stable for small banks. If anything, if a small bank becomes vulnerable they don't have access to capital markets, to raise capital, et cetera. So I think small banks may be more stable in terms of some of the other deposits but here the ratcheting up that I showed is all on the back of uninsured demandable deposits. There's one question. Very brief, please. Sure. So this is fascinating as imagery that you have shown, Viral. Do you see kind of it being driven by some kind of fundamental underlying force or was it driven by some kind of particular aspects of these episodes when the QT and QT kind of happened and related to that kind of... Are there some kind of policy measures that one could consider to address this issue that you are kind of... Yeah, so I think... So I think the credit lines make it perhaps the most easy to understand the asymmetry because in a way... So suppose you have a stock of reserves and banks sold credit lines. Typical maturity is say about three years on the revolver. Then just because reserves left the system they cannot just pull all these lines back. So the demandable claim remains even though the reserves in the system are shrinking. They can reduce the origination of new revolvers but now that's going to take two or three years for the stock to gradually start adjusting. But then as I was saying, shocks will come and hit in between. And so I think there is a fundamental problem that the scale of your operation is going to take so long to unwind and it's coincident with the creation of demandable claims. So if a shock comes along the way most likely you'll be back in the situation of injecting liquidity. I think we have to stop. Okay, yeah. We need to... Okay, we'll talk over coffee break. Thank you. So we move to the second paper, David, please. Thank you. Is it working? Yes. Well, thank you very much for inviting me to present this paper which is joint work with Annette. We started this paper. I think it's going to compliment what Viralia said because we started this paper from a different perspective which is now the financial stability side. We thought pretty much about the monetary policy side. The question that I'm going to just put here are really the two monetary policy mainly to the operational implementation of monetary policy depending on how do we think about the size or the balance sheet. So somehow they complement what Viralia said but also it was emphasizing. The title is quite also explicit about what we're trying to do. We're trying to understand the demand for reserve. I want to emphasize that we're not looking at the supply side which is probably what Viralia was emphasizing most of his talk. So here the perspective is to have like a some sort of way of framing of thinking about the demand for reserve coming from the banking sector and the extent to which once we understand that we can use that to think about how do we do interest rate control and how do we think about quantitative easing and quantitative tightening in particular in this paper. So I'm going to try to see if this works. Yeah. So everybody probably knows what is in this slide. So you pretty financial crisis but it's going to set the stage. We have a conventional way of thinking about monetary policy at least in the U.S. where the supplier reserve was pretty small. The reserve didn't pay any interest and the Fed basically could adjust with open market operation to change the short-term interest rate in particularly effective funds rate with relatively a small change in the supply of reserves. The differential crisis and the zero or bound create room for unconventional policy mainly forward guidance and QE but the point of this paper is to emphasize this part of the QE so the supply of reserve expanded quite massively that was pretty clear when you look at the size of the central balance sheet across the world measure interest of the GDP and importantly the Fed started to pay interest on reserve. So this is a new framework in which we need to start thinking about and this is what the paper was pretty much focusing on. What's the role of the demand for reserve in thinking about the extent to which the Fed can control the interest rate when you have an ample reserve demand and how do we think about the use of reserve demand to guide or to offer any guidance on the interest rate control and interest rate volatility and the extent to which we understand the reserve demand how that's going to matter for thinking about interest rate control. So in thinking about these questions we start thinking about do we have a framework can we elaborate a simple framework to think about these issues and the first thing that we did was to think about reserve demand coming from banks, optimization problem and the extent to which we can use a simple model to try to understand that. Then when we start thinking about demand we need to put the supply in place so we thought about can we think about equilibrium in the reserve market so putting supply and demand together and the extent to which this is important in thinking about what I said before which is interest rate control and in particular for the case of the U.S. this overnight reverse repo and take up that we've seen after COVID. So with that in mind that's going to be the first part of my presentation. Then we took a look at the data and we tried to estimate the demand for reserve and once we have the demand for reserve we can do a bunch of things and in particular if I have time I will talk a little bit about this controllability aspect and the extent to which when you start moving away from the flat part of the demand you need to think about how that matters for control of the interest rate as you start shrinking the size of the balance sheet. That's kind of what we try to address. So let's start with the demand for reserve. So to the right of demand for reserve from bank optimization I'm not going to go into the detail the paper has a lot of technical detail but you can think about the balance sheet of the bank on the left side you have what I just was showing before reserve securities and loans on the liability side we have deposit, fed funds, potentially repo, unequity. So the key point that we borrowed from the work of unequity with Arlene is that banks would always be managing the demand for reserve to manage liquid claim that they have so basically deposit. So mainly liquid deposit. So in the context of narrow banking in the past reserve and deposit they were pretty much the same. In assisting would you have fractional reserve banking reserve and construction or a fraction of these deposit in the context of ample reserves you can think about reserve being a function of many things these proxy for liquidity they spread between the federal funds rate and the IOR deposits and that's precisely this idea of how do we think about demand for reserve in this context of a large balance sheet what we're trying to capture with our paper. To do that we're going to emphasize this idea of convenience yield as I mentioned before so as I noticed in the previous slide the Fed now is paying starting in 2008 interest on reserve so IOR will be very important for thinking about demand for reserve. The second aspect of the model will be to emphasize that reserve has a liquid benefits for the banks basically the banks may not need to sell a liquid asset if the deposits drop. That's one of the key insights that this simple theory is going to be in front of us. We're going to have a function that captures this convenience value which is pretty much the expected transaction cost savings from excess reserve that is going to yield to what we call a convenience yield which is the properties of this function particularly the first order condition is going to be a function of the derivative of this function this convenience value function and this is going to be a fundamental part of the convenience yield which is pretty much the marginal value of additional reserve that is going to be decreasing in reserve and increasing in deposits. So with this simple model we're going to have a downward slope in demand for reserve coming from banks if the banks have a declining marginal value of holding any additional reserve for managing a given amount of deposits. So controlling for the bank deposit as a demand shifter is going to be key in our empirical analysis. All the aspects of the model will include that the banks are going to face some balance sheet costs per dollar asset this is a parameter phi there because of you know different frictions or different regulations you have some in the slides and if you think about you know posting collateral in the repo borrowing you might want to think about this you know foregone security lending revenues with this capturing by this function W. So all this is going to get into the simple bank optimization so this is a slide that show the profits of the bank that is you know what you earn from holding reserve securities loans minus what you pay for the passes and for private repo plus this convenience value that is going to be the function B as a function of reserve and deposit minus this cost of you know a balance sheet cost and posting in the private repo so we can define efficiency condition for borrowing in the federal funds rate borrowing the repo market against you know investing in reserve or borrowing via deposit investing in reserve in all these cases you have a condition like one two and three you're going to just emphasize one which is the one that we'll be using is the market interest rate will be the highest interest rate the bank will be willing to pay to borrow to invest in reserve and it's going to be equal to the net benefit of holding the reserve which is in our case the LER plus this convenience aspect which I will specify in a second I captured this idea that I mentioned before adjusted by the parameter five we captured the cost the balance sheet cost of the bank so that simple model give us this nicely looking downwards sloping demand for reserves as I said before is just coming out of this first condition that you have on the top right of the slide so the demand for reserve depends on the interest on reserve this liquidity benefit of reserve and the bank's balance sheet cost five parameter this demand for reserve that is downwards sloping will be shifting up and down depending on this bank balance sheet cost and the IOUR and the shape the elasticity of this demand for different values of reserve and for different value of deposit will be a function of this convenience value convenience yields function that we will need to estimate one interesting result of this simple model is that if the convenience yield from reserve goes to zero then the reserve demand has an as into that approach to this floor which is the IOUR adjusted by the cost the balance sheet cost of the bank and another important insight of this simple framework is that we can define the demand for reserve relative to any source of funding for holding reserve so there are many markets including the repo market so we can define this cost liquidity this idea of convenience yield relative to many in the people we emphasize relative to the Fed off on trade market but we can use all the rates if you think about the reserve supply to think about the demand and supply and how the equilibrium will be determined in the reserve market so this is the Fed reserve balance sheet up to the end of June 2023 which is the H4 4 so the Fed had on the asset side treasuries and NBS securities for almost $7 trillion a little bit of loans to banks and other assets but pretty much as you can see on the asset side are treasuries and NBS those are if you can think about the government securities there and on the liability side we have reserve overnight reverse repo for almost another $5 trillion we have and then we have a shift or autonomous factor like the treasury general account currency and other factors so if you think about the Federal Reserve from the perspective of the supply of reserve the way we like to organize these is by thinking that the amount of reserve is securities adjusted by the autonomous factor this is what we call net securities plus so long to bank as you can see are pretty much negligible and this reserve will be the net securities adjusted by this non-bank investment facility which are reserve lent to the central bank by non-banks this is something that you know but I also emphasize so a key aspect a key identification of our demand is to think about can we trace out some shifter that move the supply reserve so we can actually think about pin down the slope of the demand so that's what pretty much what we would be emphasizing total assets corrected by the autonomous factor is going to be a fantastic instrument in thinking about the joint reserve and IRP estimation but before getting there I'm going to put together the demand the nice looking downward slope in demand with the supply the supply have different form so it's not the usual vertical line to the net securities because at the lending facility rate and at the non-bank facility rate the Fed will be supplying elastically any amount of reserve at those rates so depending on the case in which the lending facility for banks is open but there is no investment facility for non-banks there is no dollar bound but if you introduce an investment facility for non-banks then you get the non-bank facility rate which actually is bounded the equilibrium in the money market radar that you can see there so of course you can think about combining the demand with the supply as I said before you can see at least three ways of thinking about the equilibrium this is the standard in which you can see the downward slope in demand is crossing on the inelastic part of the supply so the demand and the supply cross above the non-bank facility rate and this is the case in which as you can see in the title there is no take up at the investment facility for non-banks so you can see that the the locus of the equilibrium will change so the spread of the market rate will be moving depending on a factor that are moving the demand for reserve IOUR, deposits the banking cost and potentially the amount of net securities a non-bank facility rate if the reserve demand evaluated at the market rate equal to the non-bank facility rate and that is below the amount of what I call their net security so in the case of you remember what I put there net reserve in this equilibrium is equal to reserve is equal to net securities if this is not the case then we can have two situations one in which there will be a positive take up at the investment facility by non-bank so you actually see the demand shifting downward and crossing at the non-bank facility rate and the other equilibrium will be one in which there will be a take up at the lending facility for banks and you can see that this you're crossing and the equilibrium is found at the lending facility rates so in this case in which you cross at the non-bank facility rate the difference between reserve and net securities is what is called the non-bank facility this is what is right now happening which is we have a bunch of overnight reverse report that are sitting on our balance sheet for some of you that are familiar with the past history of money banking in the U.S. this is what in the past was called non-borrow reserve as opposed to this sort of equilibrium that is showing borrow reserve which is the case in which reserves are above the net securities okay so key takeaways from this I'm 17 minutes over that's fine so key takeaways the central bank can in principle control the short market interest rate it does that with administrative rates in our case the rate of a lending facility which is the discount window and the rate of this investment facility for non-banks the overnight reverse report facility and on top of that they have a choice for the net securities too on the balance sheet so the role of the central bank lending facility for banks and investment facility for non-bank is very important and one of the key lesson that is important to take into account is that the private sector will use these facilities and by using these facilities will change the equilibrium supply of reserve endodinously to keep the market clearing interest rate in a particular desired range so this is a very simple policy framework and most of the time has been successful to control interest rate the effective federal fund rate has been clearing target in the target range except during Sunday since September 2019 we can talk probably after my presentation about what will happen in there because the time is extremely successful so this part of the presentation is trying to estimate this demand for reserve using data monthly data starting in 2009 until 2022 so to do that we specify a functional for these convenience yield aspect we borrow from Lucas in Econometric in 2000 thinking about basic model of money demand and we specify something that it was working great in the data is kind of a semi-lock so this convenience yield is a function of the lock of reserve and the lock of deposit what happen when we look at the demand with this functional for the semi-lock show up that the spread between the market rate and the IOR that Viral was emphasizing this convenience yield so liquidity component is a function of the lock of reserve and deposit plus because we're going to look at the data we're adding some external shifter UT there so you can see that from estimating the parameter B and C and A we can back up the alpha the beta and the gamma which will be a deep parameter specifying the semi-lock function that we have in mind for thinking about how the demand for reserve is a function of the scale deposit and the spread between the market rate and the IOR as you can see in the slide very simple pretty straightforward so let's look at the data I'm going to show in this slide something that we already emphasized from a slightly different perspective you see that the policy went up this is the deposit growth in the United States starting in 85 so you see that the deposits went up a lot during the 2009-22 which is our sample period even relative to GDP this is no matter what you look at this either the small or large time deposits H6 or H84 you see this run up in the demand for deposit a little less so for a small and large deposit let's take this into account well let's look at the data and this is for many of you this is going to be something very familiar so if you try to estimate this demand for liquidity you see all kind of shifted around so there's no sense of stability when you look at the spread and reserve no matter whether you scale a reserve by GDP no matter what sample period you look at so you see this kind of you know shift up and down that you know people call changes in velocity so the key contribution of the paper on thinking about the demand for reserve is that controlling for the amount on bank deposit is very important to get a stable relationship of the one that I presented just a minute ago so when you do that you'll see what happened so let's take this equation to the data and Viral was pointing before that of course reserves are non-exogenous in this regression and deposits are not exotic so we try to identify these parameters in making an effort of what a potentially good instrument for reserve and deposit I'm going to get to that in turn here so in thinking about reserve just by looking at the basic insight from Jim Hamilton J.P. 1995 is that you might want to find exogenous supply shifter coming from autonomous factor that's going to trace out helping identify this V prime R so in our case it's kind of really nice to just using the identity coming from the central boundary that the amount of reserve and over an IRP is what we call net securities in the previous chart which is the difference between securities and autonomous factors and securities is controlled in an exogenous way by the central one so this is our candidate for being a good instrument for our reserves and of course deposits are not exogenous and we'll try to instrument deposit just looking at something that we thought it could be also interesting in thinking about this equation which is bringing the household financial assets so thinking about the real side of the economy are part of an asset for household in this economy and it turns out that this is also very natural and interesting instrument as I will show you in a minute plus the level of IOER controlling by this IOER or not really doesn't change the result much as I will show so briefly what you can see here is the estimate of the equation that I just had on the previous slide. The panel is the second stage so we run the spread between the market rate and the IOER on reserve and deposits. You can see in the sheeted area extremely significant and with the right sign so it depends positively on deposit negatively on reserve which is a downward slope in nicely looking demand curve and the first stage is putting there to see that also we're actually close to a demand. You never guarantee that this is a demand but at least the exclusion restrictions of this simple analysis is working nicely so you see that the deposit doesn't enter into the first stage and reserve over an IRP is a really nice nicely correlated with reserve which is a great instrument. So that's pretty encouraging. So what we do now is to translate these reviews for estimating to the parameter that we have in the semi-lock specification of our B prime function. This is done in this slide so we translate this parameter into a semi-elasticity of the demand reserve with respect to the interest rate spread and what we got there is that 10 basis point change in the spread leads to an increase in reserve holdings by 50%. So it's a pretty flat demand. I'll show you exactly what I mean by pretty flat but not totally flat. So it's a pretty elastic demand for reserve what we find for the U.S. in this sample period. On top of that we also can identify the elasticity of this demand for reserve relative to deposit and this elasticity is slightly about 1.5%. So 1% increase in deposit increase reserve demand by more than a 1%. So this is kind of the other side of the coin of what Viral was emphasizing in his presentation. So what do we mean by flat demand so you see here how the model works. So this is in sample the feed of the of the model. The blue and the red line tells you the combination over time of the affected federal fund rate relative to the LUR and the feed value. So it's a pretty nice feed. Nothing that makes us think that this equation has been affected by QT and QE and different QE. So the overall feed is kind of nice. As you can see on this scatterplot on the right there's no sense of large shocks and most of the dots are well close to the feeder line except the one that you see there on the top left which is the one in 2019 September which is what happened in the particular repo market. We can talk about that. But overall the feed is really nice and you can see that this is flat but not completely flat. That's what I mean by very elastic. So why did the deposit grow? There's an increased deposit demand. So we look at this from the perspective, as I said before, can we just instrument the deposit in the banking sector and we'll look at the the real side of the economy in particular by looking at the flow funds. We look at the financial asset of the household. You can see that yet financial assets trending up as you can see on the left panel since 1985 dramatically. So when we put the increase in deposit relative to this run up in the financial assets of the household, you see that deposit relative to financial asset has been pretty stable. So the way we think about that is well you can think about portfolio choice here from the size of the household which is giving you a stationary fraction of the financial wealth allocated in this deposit. So we use this instrument as potentially useful to see whether the previous specification get affected by it. So this is what happened with the previous estimate. So the panel A again is just the second stage now instrumenting both variable. The results are extremely similar, it's slightly different of course, but the elasticity is at both part. And the panel B are the first stages. As again you see that we have reserve instrumented by autonomous factor adjusted securities adjusted by autonomous factor which is reserve an IRRP extremely significant financial asset exclusion restriction doesn't enter there in the same way that deposit didn't enter but are very highly correlated with deposit as you can see on the second column of the panel B. So there's potentially useful instrument doesn't change significantly what we have in the previous specification and that make us kind of happy in the sense that maybe this is really capturing something that proxy at demand for reserve coming from the banking sector. This is the feet of the model seems very stable again regardless of whether we have different Kiwis and different QTs there's a small size and barely autocorrelated velocity shots in this specification. So in the next five minutes I'm going to try to use the simple framework to tell something about what are the implications for industry control and in thinking about QT thinking about QT from the perspective of the controllability of the short term industry by the Central Bank. So to that end I introduce a concept that was developed theoretically in a paper by Bianchi and B.G. in Econometric in 2022 which is this idea of the ISOFED funds curve which is very straightforward given what I just show you already. So an ISOFED funds curve is useful because it give you a combination of the I OER and reserve plus overnight RP that imply a predictive effective federal funds rate equal to a particular choice. You give me a value of the effective federal funds rate and I can trace out a combination of I OER and this reserve and overnight that can achieve a particular value for this market rate. So how do we do that? You see that in the bullet in red so you give me a value for the federal funds rate. I know the interest on reserve and given my specification I know the parameters A, B and C and I can evaluate these computing what you have in the panel on the left. What you can see in the panel of the left those are the ISOFED funds curve give me a combination of liabilities of the central bank and I OER that deliver an equilibrium rate a 2% or 4% federal funds rate. You can see that this is tracing out billions of dollars trillions of dollars that are reflecting the elasticity that previously estimated so to achieve you know a certain federal funds rate you see that the steepness of this ISOFED is pretty large when the balance sheet size is becoming smaller and smaller but it's relatively flat for relatively large size of the balance sheet that's important for what's coming next and what's coming next is how much then reserve an open IRP the liability size of the central bank can be adjusted or reduce it in a way that by moving away from the flat part of the demand interest rate become not very volatile so what we do here is to use the specification that we estimate to run this counterfactual and to do that one important insight is that these calculations depends on the level of deposit in the banking sector for each level of deposit in the banking sector which is our scale variable you can identify this trade off that you see there so what we do there is just imagine that we set the deposit the value of the end of 2022 which are around $18 trillion for that we say that reserve an open IRP as I show you before around $5 trillion close to 20% a lot less now so the calculation that you see there is that to achieve rate that is tighter than the one that we saw in September 2019 which is this spike in the federal funds rate that we observed at that point you need to actually move the size of the balance sheet to in our case a little less than $2 trillion if you want to keep the same spread with the one that we saw in September 2019 it becomes a little bit more than $2.8 trillion around 11% of GDP or if you want to set that spread to zero maybe enough to avoid any kind of daily spike in that rate then the size of the balance sheet needs to be relatively larger around $3.5 trillion of course these estimates will change as Viral was emphasizing depending on the deposit in the banking sector for different value of the deposit this number will change and of course this will be evolving assuming that we keep even constant the estimate A, B and C in the specification that you have there of course the funding repo facility may help to get these numbers right so this slide which is almost my final one summarized how much reserve and over an IRP which are these liability can be reduced assuming that what I show you is a ballpark reasonable estimate we think that this estimate if anything a little bit conservative why? we haven't taken into account just give me one minute Massimo please so we introduced the standing repo facility in July 2021 that makes dealers and depository institution able to borrow as you all know funds from the Federal Reserve using repo borrowing and that may help in reducing any additional volatility of any spike that you can see for a given level of reserve and over an IRP and then volatility on the autonomous factor is important taking into account so it might be prudent also to take that in consideration given our estimates in a way that may induce undesirable volatility QT may lower deposit so but it also can increase deposit may not necessarily reduce deposit QT may reduce deposit in the banking sector if the house will buy more bonds even directly or indirectly through bonds funds or money market funds who actually will be investing in the Treasury repo something of that sort here but also QT may not necessarily reduce deposit if the overnight RP take up falls with reduction in the amount of total security adjusted by the autonomous factor and the money market fund just replaced overnight RP with private repo or hedge funds or other leverage investor may hold more bonds in the repo markets I'm going to stop here and thank you all very much Thank you Thank you for this great presentation of interesting paper for the time that my slide is put on let me thank the organizers for having me to present this to discuss this paper and of course the usual disclaimer apply This being said to summarize the paper I think this paper makes a great work in terms of taking all the boxes of what to expect from nice papers good model, very tractable, very nice very elegant estimation and finally policy exercise that they speak to hot topic that is currently monetary tightening so in terms of models the author design a reserve demand on supply model that speak to the control of the short-term policy rates and that embeds all the key elements you want starting with supply we start with the level of net security that is the vertical policy on the top chart and that is the initial reserve that the central bank is supplying to the entire banking system then it is also willing to lend any amount of reserve to the banking sector at the lending facility rate providing the banking sector has enough collateral that to the left the right side of the horizontal red line and it is also if it implements an overnight reverse repurchase facility central bank is willing to absorb any quantity of reserves that the banks are willing to deposit at the central bank by moving cash from the banking system to the central bank it decreases reserves regarding demand the key is demand function is the following that banks are willing to pay for an additional unit of reserves what they get, what the monetary yield that they get from depositing these reserves at the central bank which is the IOR and the convenience that it gets from this unit of reserve that depends on the initial amount of reserves it has and it decreases with the initial reserves because the more reserves you have the less you care about any additional unit of reserves and that's key in the model underestimation it increases with deposits when you have more deposits on your liquidity side you want to self-insure against potential outflows on your liquidity side by having liquid reserves on your asset side and final final item this fee parameter that captures the balance sheet cost of running a large balance sheet that typically can be a proxy for the leverage ratio requirements that were introduced after a great financial crisis then making some assumption on the shape of this VR prime function those sorts arrive at the main equation that is colored here and it is very nice in two ways first it links the main policy target rate of the central bank that is in green that is the effective fund rate for the reserve that are in red which is first the IOR so it's a remuneration rate on reserves and second the amount of reserves that it push on the system pushing the system and it creates an interaction between those two instruments when you want to target a particular fat fund rate meaning that the more reserves you have in the system and because this big efficient is negative the more reserves you have the lower the spreads between the effective fat fund rate and the IOR meaning for instance that if you implement a QT contact training you reuse the reserves which increases the spread that you need to keep between the fat fund rate and the IOR to keep the fat fund rate at the target level and the second nice feature with this equation is of course that you can estimate it actually and the authors do it in the second step and they find the bottom chart that you have on this slide that shows a very tight fit of this spread suggesting that first the model is adequate in capturing the main elements, the main drivers of this demand equation and second that the coefficient are stable which is something that I will discuss in a minute and the authors insist that it's key to obtain this fit that you control for demand into the equation or into the equation which suggests that the key driver of reserve demand thinks really want to self-insure from the present outflows by holding reserves. Next the authors use this estimated reserve demand equation to run two policy exercises to guide policy monetary policy tightening. First they wonder what is the IOR that is feasible for a target effective fund rate and a target path of QT. As I told you the lower the reserves the higher the spread between the effective fund rate and the IOR which means that as you reduce reserves you must also reduce the IOR to keep the effective fund rate stable and this is what you have on the left top chart. When you reduce reserves and you move on the left toward of the X axis you have also to reduce the IOR that is on the Y axis. Second policy exercise what is the possible QT total tightening that you could implement without making the effective fund rate too volatile which means for the author without having a too large spread between the effective fund rate and the IOR and they find they make several scenarios and what I think is their preferred one is to finally reach 3.5 trillion dollars of remaining net securities that would leave the spread at zero which you consider will good enough not to have too many spikes in the effective fund rate. Finally, during all that they also find that deposits affect reserves as I said but not the opposite wants to control for demand factors which means that more reserves do not create more deposits supply only maybe more deposits demand but not supply. Overall I think it's a very great paper it makes a few main contributions first it provides a very elegant model including all the key elements of a rate steering operational framework both supply and demand that allow to speak about liquidity as a weather but also can speak into other questions that I will discuss in a minute second it is highly policy relevant of course in the current context and turns to nice policy exercise as I mentioned you can either take one of the two policy instrument as given and it gives you a mapping for the other policy tool and finally it contributes to debate on the link between reserves and deposits more deposits increase reserve demand from banks but more reserves supply do not increase deposit supply in the authors view. So I have a few points to discuss the first one would be if you want to bring this model to the Eurora case it would be of course very interesting to do but I would be cautious to keep in mind a few factors the authors and David say at the end of his presentation that estimates on the possibility of quantitative taking might be conservative so that you could go a bit further but in the Eurora context there are some elements that would warrant more cautiousness first we do not have a non-bank facility which means that there is not this flat left-hand side line of the supply side which you could reduce without affecting the effective financial target policy rate so we should be cautious on that and second we already have some form of reserves in the Eurora which are the Teltros the long-term loans from central bank to banking system that are already being repaid so we are already removing some reserves from the system and as you can see on the right-hand side chart when the middle of it we repaid the banks already repaid about three quarters of Teltros but there is still one remaining quarter to repay until the end of 2024 so there is already some form of reserve addiction to nice theoretical framework to be empirically adjusted for each jurisdiction that you want to apply it on second as I said this paper is already nice and the authors focused on liquidity provision and liquidity demand but I think it can also inform on other aspects on central bank balance sheet and regulation in particular the role of central bank debt security portfolio and long-term financial operation at the end on the long-term framework not only the QT but what's the final level that you may want depending on what's the shape of the rate steering framework that you would be interested in. Second, it can also speak to the convenience of reserves what's the impact of this convenience on the control of short-term rates in particular now that we talk about post-tentional CBDCs that would reabsorb reserves now because they would be moved out of the banking system into the CBDC and how this would impact potentially money market rates so that's something that is discussed for instance by a paper by Abad Nunez and Thomas. And third the impact of auto moves factor shocks so you quickly touched on that at the end of the presentation but there could be other type of shocks maybe it has increased over time maybe it varies also depending on the jurisdictions for instance in Europe depending on governments it's not always the same type of shocks that you serve that the banking system can experiment so this would also be useful to use your model to inform how those shocks may affect the money market rates and the steering of the policy target rate. Finally also the balancing cost which is these three parameters in your model and that should be affected by the regulation but for instance the leverage ratio how this could affect this cost. On this I find very interesting that you find a very tight fit of the reserve demand equation which suggests stable coefficients you have the very stable fit over 13 years which suggests that the parameters did not change much or they offset each other which would be a bit surprising. Why is the environment changed a lot? No. The convenience here should have changed because of liquidity requirements on higher relative auto moves factors so you have expected that banks value more in the additional units of reserves at the given level of initial reserves and initial deposits and second the balance sheet cost I would have expected it also to increase given the new leverage ratio requirement so I don't know if you have already thought about it how to rationalize the result and what it means for the regulation. Next how to explain the absence of impact of reserves on deposit supplies that's something that is an important result of your model of your paper it's not maybe the core paper it's an important result and I wonder how we can think about it because when as Viral mentioned in his own presentation when the central bank pushes more liquidity into the banking system by into the system by buying the securities it creates liquidity it creates deposits now so how is it reabsorbed by the central bank is it that non-banks immediately deposit it at the central bank through the overnight reverse repo facility or is it that then it is somehow transferred to households and firms that turn then this deposit into less liquid venture assets. Second I would expect deposits to create more credit no when banks have more reserve to create more lending that creates more deposits so when banks have more reserves they maybe feel more comfortable into lending more that mechanically creates more deposit for those that borrow so is it that those reserves do not create credit or is it that this credit exists but then it is not transferred into finally into more deposits maybe again because of the same factors that these deposits are really are changed converted into other forms of financial assets and to deal with this question is there a role for different source of reserves borrowed versus non-borrowed for instance as also Lorenzo mentioned in his discussion of Veyron's paper. It is something that is discussed in the paper at the ECB by Alta Villarro Stegno on Schumacher. And third point that may touch to this issue you say in your paper that you may have an important driver of household financial assets in the past great financial period but that puts in place on deposit demand not deposit supply so that putting more reserve to a system would increase deposit so more deposit but it would not work through supply and I wonder if you have thought about what implications this would have that it works through demand and not supply and if there is a way for you to test it empirically because this would be a nice addition finally as I said it is already a very nice paper that already includes many nice features but I think one possible refinement could be about the link between the spread between the effective FedFend rate and the IOR and the volatility of the effective FedFend rate because that's what the key constraint of your second policy exercise now you want to avoid this effective FedFend rate to be too volatile and the way to do it is to say okay to do it you have to keep this spread low or to constrain it so this is true that there is this link for instance in the IOR it is very clear I wonder if you could add it maybe to the paper empirically to show that it holds and second maybe include it in the model to find a way to include it to really close the model or maybe a way to do it would be to consider the implication for cross sectional heterogeneity in deposit volumes not all banks have the same ratio of deposit to total assets which may lead to different heterogeneous reserve demand elasticities in your model this VR prime reserve deposit function might be different across banks and this may lead to different volatility to different demands that may lead to volatility in particular volatility because would exit this flat part of the demand line as you said and move to maybe the more vertical part and then this heterogeneity would be even more important but it's already as I said a very nice paper and very useful for policy makers in the context of entry titanium so thank you very much David yeah so just briefly so thank you thank you very much those are really really nice nice comments that are actually part of our ongoing work so we're trying to look at this not just from a time series perspective but using some cross sectional data too that's taking a little longer than we thought we also start started thinking about applying these to other countries and of course to the at least to think about the ECB as well but we're not totally familiar with what's the best way to do that there are a couple of things that I want to just put some thoughts in here so one is the extent to which in our specification these changes in regulation get not reflected because the fit is so nice this is something that surprises us too and we start thinking about well can we use these kind of also changes in regulation in the supplemental liquidity ratio capital requirement so on so forth as potentially way of capturing some of the parameters of the model the boundary cause and so on and use these potentially interesting instruments too so we're in the process of doing that as well it doesn't seem that matter much but it's a good suggestion the other before going into the supply and demand on the deposit which I think is also part of what we're thinking because we're pretty much motivating as I was telling yesterday I mean the work that the Rago Vidal doing is also pretty fundamental to understand all these correlations in the data but before getting there one aspect of our estimation that is important and we're trying to understand in the context of published in the paper is that the properties of the demand are the convenience yield demand is interesting because is technically deviating a little bit from being homogeneous of the degree one so the elasticity of deposits and reserve is not just one it's just slightly above one so we we're now using this you know cross-sectional variation to see if there is something that has to do with the way this is aggregated that make us in the aggregate finding this larger than one effect that is connected to some of the aspect that are related with what we see in the deposit demand or supply driven and the extent to which the supply aspect of these are related with actions coming from QE and QT in sequence I think that we need to do a little bit more work on that I think that we this is a fundamental question for us part of the discussion before I don't think that we have a good way of thinking about the real side of the economy and the extent to which the real side of the economy is also being behind what we see with the aggregate deposit and reserve in particular the example that we've seen during and after COVID there is some element of insurance and some element associated with the demand of these kind of insurance by the private sector that is behind some of these liquid assets and deposits convenience value attached to them by the private sector that probably are behind some of the correlation that we're picking right now with aggregate data so combining that real side of the economy with this aspect with the effort that we're trying to make in thinking about the operational framework and the demand component is definitely something that we all of us will be pretty much involved over the next few years in thinking hard so those are really hard questions, good ones and I don't know the answers this is where we aren't right now so again thanks very much Any questions? Yeah, Philippine to the puzzle of this more reserve supply would not increase so much the deposit supply have you looked at the open economy aspect for the US with the Chinese demand for treasury bonds having US treasuries really declined and so maybe you have some leakage in that way you want me to respond one by one maybe we can collect a few further questions maybe you go ahead we have a look at that we have a this is a good question but we have a look at it in detail Carlo thanks this is a great paper just a clarification if you can probably spend like 30 seconds on explaining as what are the determinants of large balance sheet with respect to the rate controllability so what is that makes really so the size of the balance sheet large and so large like you have in your estimates so what is it? well we do not claim that you need to have a specific size for controllability so what we claim is that given the slope of the demand you need to tell us what exactly is your idea of controllability what is the spread that you would like to target and given our estimate on the spread that you would like to target that give you a potentially size that is going to make you reach that in the medium run but it's not necessarily tying at all the balance sheet of a certain size with the controllability I mean of course the estimates are calling for if anything kind of the opposite because the demand that we are estimating is pretty flat it's pretty elastic so the kink between the really flat part when you have a super ample reserve and the slightly less it's just pretty big so in terms of controllability those are kind of a good news there is an intriguing question coming through Webex which is close to what I wanted to ask you intriguing the latest the latest New York primary dealer survey shows median dealer expectations are for system reserves to fall to 2.7 trillion dollars in about a year's time this is well below the upper estimate of necessary reserves quoted by both recent San Luis Fed estimates and by your own estimates of a safe upper limit does this suggest primary dealers intuit a lower limit to reserves than your own studies if so why might this be the case I don't know the answer crazy people deserve primary dealers I mean I don't know what the primary dealers are thinking about the joint determination of the federal funds rate the rest of the instrument and the size of the balance sheet so I don't know the answer to that it's also important to think about what's the macro outlook that these primary dealers responding are taking into consideration in thinking about a certain amount of reserves in the system so I don't know there's nothing normative in what we say just try to be positive just to give you an example of how to think about this as a way of framing controllability and the size of the balance sheet from the monetary policy perspective there's some interesting aspect attached to financial stability as well that we don't mention ok so with that we close thank you very much it was a great session I learned a lot but we need to now we have a coffee break and we need to come back at 4.30 sharp because there is a pretty famous person giving a keynote speech from Chicago so I think we need to be timely thanks a lot