 Fy roi'n gweithio i ddeithasio'r forum. Yn ymdegfynol fod ydych yn Cintra a Llywodraeth. Felly, today we looked at the latest thinking on inflation from aspects ranging from supply shocks to who pays for price pressures. This morning we want to delve into other aspects of the consequences of strong price pressures. Mae'r cymdeithas yw wedi'i cymdeithas yw ein bach yn ddechrau'r cyfranteidol. Rwy'n meddwl am erbyn yn dda i ddisiguol y balans. Gwin ein cwestiwn ar ddisiguol y balans sydd gwis i'r cymdeithas yw yw ydy, a y gallu o'r sicurau dioledig mae gennym ni risio. Mae'r cyflwdoedd rheswm, na yw ddwy'r ffysgol a'r polisi taelwyrraeth sydd yn ôl yn gweithio a'r llwysiau. Mae'r gweithio'n gwneud ond yn ysgolwch yn ysgolwch, neu'r gweithio i gael y cydwyddiadau cyfaint, yr eistedd gyda'r prysgol bydd yng nghylch yn ysgolwch, yn ddiddordeb yn ddiddordeb yn ysgolwch. Fodd i gyd, mae'r prysgolwch yn fawr o'r gwirionedd hawddol yn ydym ni'n cydweithio'r ffocl yn ysgolwch, ac yn ddiddordeb yn ddiddordeb yn ysgolwch. yna sy'n brafio i ddweud hynny yw'r cyflog gyrraedd, yw'r ysgol wedi'i dechrau yw'r cyflog gyrraedd, o'r cyflog gyrraedd yn gyrraedd, o'r cyflog ar y gagell dechrau yn ei ffordd y dyfodol. Y ddweud hynny yw'r cyflog gyrraedd yw'r cyflog, o'r cyflog ar gyfer y bydd y cyflog ac ydy'r cyflog. Ond o'r eich felwyr yn rhaid, rydym ni i'n dweud i ddweud y cysylltu'r rhai o'r cyfnod. Yn dearf yn cyfnod, ac mae'n ddweud i'n ddweud sy'n gweithio'r cyfnod o'r cyfnod ynghyd ym gyffredinol. Yn ddweud i'n ddweud i'n ddweud i'r cyffredinol. A'i ddweud i'n ddweud i'n ddweud i'n gweithio. Jordan Galeys, a senior researcher at the Center for Research in International Economics, a professor at the University of Pompeo Favre. And a research professor at the Barcelona School of Economics. So, perfectly, the flow of events. So you will have the floor for 25 minutes to present your paper. Afterwards Jordan will make some comments and he will have 15 minutes and afterwards we will open the floor for questions. So just in case that we have any questions from, you know, from Webex, please you can raise your virtual hand and we will try to collect all the questions that you are going to ask. So without any further delay, Annette, you have the floor. So I'd like to talk about central bank balance sheet policy when the short interest rate is above the effective lower bound. And as I'm now working for the Federal Reserve I'm going to be a little bit more diplomatic than I normally am. And just highlight in large red font that these are just my views and not those of the facts. So please don't give Chair Powell a hard time about this later on in the panel. All right, so the question is what should central banks do with their balance sheets when they're not really needed in the sense that above the effective lower bound the central bank can adjust the monetary policy stance by adjusting up and down the interest rate on reserves. So I'm going to argue that convenience yields are going to be very useful for thinking about central bank balance sheet size and asset mix. By convenience yield I just mean benefits on investments that you get as an investor above and beyond the principal and interest rate payments. They can come from liquidity in the form of safe transaction costs so they could come from safety in the sense that if you hold the security at very low default risk you don't have to expand resources on credit risk analysis. Now central bank reserves are perhaps the ultimate safe and liquid asset and so I'm going to argue that providing these reserves adds value to the economy in a similar way that the supply and cash does. So the main result of the paper is that I'm going to derive theoretically and empirically what I'll call convenience maximising reserve supplies. I'm going to distinguish two cases that depend on how the central bank supplies those reserves. In case A the central bank is able to supply reserves without holding assets that themselves have convenience yields. As a result the convenience maximising reserve supply will be larger in a sense that it will become precise as we go along. I'll contrast that with case B in which reserves are supplied by the central bank holding assets that have convenience yields and therefore the convenience maximising reserve supply will be smaller with the main intuition being that when the central bank is in case B it's supplying liquidity through reserves but it's quote unquote eating liquidity by holding treasuries or bonds for example. So I'm going to lay out the argument in the five steps. First I'm going to argue that which kind of assets the central bank is able to hold is to large extent driven by political constraints, how to contrast the Fed and the ECB. Then just for those of you who don't think about this stuff every day I want to just remind you of the sense in which the central bank doesn't really need a particular balance sheet size to hit its target policy stance above the effective lower bound and therefore it's going to be relevant to discuss what are some factors that could guide that choice of balance sheet size when it's not, when a particular size is not strictly needed. Then we'll dig into the main framework of dividing the convenience maximising reserve supply as a function of the central banks as a choice and then I'm going to estimate this out for the U.S. and the Eur area in the last step. Okay so let's think about the politics of this. So the Federal Reserve as you know has announced plans that it will hold mainly treasuries in the longer run quote thereby minimising the effect of Federal Reserve holdings on the allocation of credit across sectors of the economy. Now this choice really fundamentally goes back to the Federal Reserve Act under which the Fed can hold securities that are guaranteed by, that are issued by or guaranteed by the U.S. government. Now of course the Fed has had very successful programs for corporate bond purchases during COVID but those were constructed under emergency lending conditions and similarly the discount window is generally priced to be used mostly in crisis. So I want to give you a quote here by a Protestant good friend who expressed well the common sentiment in the U.S. that the Fed really should mainly hold treasuries. So I quote, the Fed's asset acquisition policy ought to give priority to preserving public support for the Fed's independence by insulating the central bank as much as possible from potentially damaging disputes regarding credit allocation. When the Fed purchases Treasury securities it leaves all the fiscal decisions to Congress and the Treasury. So that's the Fed by contrast. I think the ECB could likely hold only assets without convenience yields in the longer run. Remember the ECB has historically supplied reserves through collateralized lending to banks. In fact, in the Euro area it's the government bond purchases that have been more politically sensitive as you know they have been challenged in court. Isabel Schnabel actually in a very nice recent speech states well the sentiment amongst many in the Euro area that there are additional considerations relevant for the assessment of whether large bond portfolio is desirable or not. One is that the lack of a consolidated public sector balance sheet raises more fundamental concerns about monetary and fiscal interactions in a currency union with sovereign member states. These concerns may potentially undermine the credibility and independence of the central bank. So in other words, across the Atlantic what's politically sensitive for central bank to hold differs with government bonds being the politically safe choice in the US and a politically biscuit choice in the Euro area. That means that from the perspective of supplying convenient assets the ECB is at an advantage because it fits case A better where it supplies reserves without at the same time subtracting bonds say with convenience yield or it could decide to do so. The fact is that it is advantage because when it supplies reserves it is at the same time withdrawing a treasury supply from the private sector to hold. All right, so moving on to the second step in what sense is a particular balance sheet size not needed when the short rate is above the affected lower bound. I'm going to review this well known point within the framework of a recent paper mine because we're going to use that framework later on in any case. The framework models the banking sector cement reserves and there's three simple ingredients. The first one is that reserves pay interest. The second one is that reserves have liquidity benefits for banks in the sense that banks that have reserves can manage deposit in and out flows without having to sell illiquid assets such as loans or securities at high transaction costs. The third factor is that there's a balance sheet cost for banks from the banks perspective this captures capital requirements such as the supplementary leverage ratio. Now in terms of that second factor the liquidity benefits we capture that through what we call a convenience value function V which just simply measures the expected transaction cost savings to the banks from managing deposit in and out flows with reserves rather than illiquid assets. A crucial element in the analysis is going to be the derivative of this V function with respect to reserves. I'm going to call it V prime R. It's the convenience shield from reserves. It's the marginal value of an extra dollar of reserves and that marginal value is going to be decreasing in reserves as additional reserves are less and less useful but it's going to be increasing in deposits since having more reserves is more valuable if there's more deposits to be managed. Reserve demand then it's simply the bank's first order condition for borrowing at the market interest rate R. Think here the Fed funds rate or the European short term rate for borrowing at that interest rate R and investing in reserves where on the right hand side you have the net benefits of reserves or what the bank gets out of borrowing to invest in reserves and that net benefit comes from the interest rate in reserve. The convenience shield on reserves V prime R and then subtracting out the bank's balance sheet cost. If we graph this out, focus here on the left graph for the reserves market, the demand curve is in blue. I'm going to have two of them. I'll go through that in a second. The demand curve in blue is just the formula that we just derived. It's going to shift up with the interest rate on reserves naturally as reserves are more attractive if they have a high interest rate it'll shift down with the balance sheet cost and the whole shape is just driven by this convenience shield function V prime R which eventually goes to zero so it's going to completely flat now way out to the right. Okay so consider the two points A and B. At both of these points the central bank hits its target interest rate R, the target R. At point A the central bank has chosen to have a low interest rate on reserve accordingly. The reserve demand is low and in order to hit the target the central bank needs positive scarcity of reserves which it achieves by choosing a low supply. In contrast at point B the central bank has chosen a high interest rate on reserve accordingly the demand curve is high and to hit the target the central bank now needs negative scarcity on reserves and that's achieved by a high supply. So in other words there's actually many different combinations of the interest rate on reserves and the quantity of reserves that will enable a given policy stance. I summarized that in the right picture which I have labeled the iso market rate curve. Along that curve the short market rate hits the target. Now you can think of it as basically how to set the balance sheet size given interest rate on reserve or conversely how to set the interest rate on reserve given the balance sheet. Actually as a policy maker what you really want is the iso market rate for the long interest rate because the long interest rate is at better measure of the overall policy stance. It's that iso market rate curve it's in the paper it's a bit steeper than the current one because at lower balance sheet size term premium are going to be higher. But the basic point remains is that there's a whole schedule of choices for the central bank when the short rate is not constrained by the zero lower bound or the effective lower bound. So therefore let's talk through some factors that might be relevant for setting balance sheet policy above the effective lower bound. I'm going to focus on the first one, the central bank's supply of liquid and safe assets. A second commonly mentioned factor is side effects of large balance sheets. Those stem from the fact that the banks have to fund the reserves which essentially can happen in one of two ways. Either reserves have to crowd out the bank's other assets that could lead to associated welfare losses from less lending to firms and so forth or reserves have to crowd in the size of the banking sector such that its overall liabilities grow. That of course could be good if those liabilities have liquidity and safety benefits to their holders but it also could be bad from a financial stability perspective especially if those extra liabilities are uninsured deposits for example. A third commonly mentioned factor is interest rate volatility. As I graphed out before the reserve demand curve tends to be flatter for high quantity which means that if in practice the autonomous factors of the central bank are very volatile that leads to volatility in reserve supply and that in turn leads to less interest rate volatility if you are out on that right tail of the reserve demand curve. Finally central bank profits of course are an issue and in that sense a large current balance sheet may be viewed as limiting headroom for future QE and may be reasonably emphasized in a speech by a household. So I'm going to focus on the first one but let me just suggest that you guys as policy makers start from my numbers and then add and subtract based on your personal preferences regarding the importance of the other factors. All right so let's then dig in to the convenience maximizing reserve supply and I just need a couple of formulas and then we can get to the data. Okay so think about the Friedman rule for the optimal supply of money. It says that to maximize the welfare of money the central banks should keep supplying them until the last supplied unit has no value for managing payments. Translated to reserves this would suggest setting the convenience yield on reserves and that of these balance sheet costs to zero. Accordingly this is a reserve number that central banks calculate it has many different names across different central banks. That is of course very useful but what if the central banks assets also have convenience yield. So to think about that let's talk now about not just convenience yield on reserves but convenience yield on bonds and for the sake of concreteness let me illustrate this for treacheries but of course it would work the same for bonds. I'm graffing here yields against default probability. The straight line illustrate the normal upward sloping relation and this idea of convenience yield on bonds is just a statement that bonds such as treacheries that have extremely low default risk tends to plot below the line as you can see with that bottom red dot where the treachery bonds are plotted far below the line. So that means that if we think about the yield spread between treacheries and corporate bonds that don't appeal to these safety and liquidity investors then that yield spread is not only going to have the default component it's also going to have a component from the treachery convenience yield essentially the spread is going to widen out because the treachery yields are artificially low because some investors are willing to hold them due to their safety and liquidity properties. One can implement that decomposition in practice by observing that the treachery convenience yield should become really small for sufficiently large treachery supply as its demand is saturated Implementing that decomposition in work with Krishna Morty we estimate in early work a very large convenience yield on treacheries it's a bit lower when you benchmark relative to AAA bonds which might suggest that they themselves have some appeal to these safety and liquidity investors. I'm going to use this spread to measure treachery convenience yield going forward to just keep in mind that if anything my results will understate just how special the treacheries are. All right, so now we're ready to derive the main result and focusing on a central bank that's in case B where it supplies reserves through holdings of bonds with convenience yields we need the following two formulas. The first one is just the central bank's balance sheet on the left hand side it's the central bank's holdings of convenient bonds on the right hand side the central bank's liabilities which is the reserves plus the astronomers factors. The second equation we need is the private sector's convenience from holding reserves that's a red term and convenient bonds that's a blue term and notice importantly inside the blue term that the private sector's holdings of convenient bonds is the total supply B minus the central bank's holdings BCB. All right, so now the main theoretical result of the paper is the following contrast in case A and B. If a central bank is in case A and is able to supply reserves without holding assets that have convenience yields then it should just focus on convenience from reserves. That's maximized by setting the net convenience yield on reserves to zero as we have discussed. By contrasting case B if the central bank has to supply the reserves by buying bonds with convenience yields then it needs to focus on both parts of the formula the reserves and the bonds and that overall convenience is maximized by equalizing the convenience yields on reserves and bonds. I show in the paper there's a footnote that this result holds regardless of exactly how the bank can take the funds that we serve through this crowding and the crowding out mechanism that I mentioned before. Right, so here's the fun part, let's grab this out. So let me start with case A. So the graph to the left illustrates the reserve market. We have already talked through the reserve demand curve. The total convenience value from reserves is the area between the demand curve and the interest rate on reserves. You can split it up into the consumer surplus and the producer surplus as indicated. They sum up to the net convenience yield on reserves. Okay, so you can see that the supply and the picture here in case A is not optimal from a convenience maximization perspective because you could increase the sum of the consumers and producer surplus by supplying more. So illustrating that here with the green line, this illustrates case A where the central bank supplies reserves to the point that there's no reserves scarcity. By contrast, in case B, we need not only that graph in the reserves market, we need also a graph from the bond market. That's the graph to the right where in blue I'm graffing the downward sloping private sector convenience yield from treasuries. Now I talked through point A in the left graph for case A. Let's discuss whether that's optimal now in case B where the central bank has to supply reserves by buying bonds. I have, for that purpose, plotted in where would that point A be in the bond picture and it's really high. So that's because at point A, the central bank supplies a lot of reserves. It does so by buying bonds for concreteness. Let's focus on the fact and say they're buying treasuries. And because the central bank buys a lot of treasuries, there's not a lot of treasuries to be held by the private sector. Accordingly, the treasuries are really scarce. All right, so what is the central bank to do in this case? Basically you should grab those two green supply curves and just pull them out. So I was trying to come up with some snazzy term for this. You could think of it as sort of open the curtain. Just go to your hotel room, grab those two sticks that hang on the curtains and just pull them to the right. If you remember that, then you have remembered the main point of the paper. So at then the red supply curves, overall convenience value is maximized by equalizing that on reserves and treasuries. OK, I want to just make a comment about the ECB. Since I have contrasted these two stock cases, case A and case B, of course they could be intermediate cases. So for example, suppose that the ECB, for reasons outside my framework, decided to supply reserves with a mix of bank lending and government bond holdings, some of which were convenient. Then it's straightforward to show that convenience is maximized by setting the net convenience shield equal to the average convenience shield on the ECB's holdings. So for example, suppose, with an apology to the Dutch, that only the German bonds have convenience shields. The right-hand side here would be the convenience shield on bonds times the portfolio weight on bonds, which would be the product of Omega, the ECB's portfolio weight on bonds, and Alpha won the weight of bonds in the ECB's bond portfolio. All right, so now we're ready for the data. Let's start with the US. So what we need out of the data is, empirically, we need to figure out how to measure the convenience shields, and then we need to estimate those two convenience shield functions, the V prime R and the V prime T, so we can start setting stuff to zero or equal to each other. So starting with the market for reserves in the US, I'm going to use the Fed funds rate as a proxy for R, the market rate, and therefore I'm going to use the Fed funds IOR spread as my empirical measure of the net convenience shield on reserves. It's currently really low at negative seven basis points. To estimate the whole reserve demand function, I'm going to follow that earlier paper of mine with Lubasolito and assume that the net convenience shield function is log linear in its inputs that gives us the estimating equation in bed. Now I need to talk you through why I have to instrument for excess reserves, and that's because they're correlated with the error term. That also gives me an occasion to illustrate the role of the O&RP facility as well as the ceiling facility, the discount window here for the Fed, for this whole reserve demand framework. If you look at the graph here to the bottom right, we have our reserve demand curve as before, but now that we have this floor and ceiling facility, they basically cut off the tails of the reserve demand. So focusing on the role of the O&RP facility, any supply of liquidity from the Fed past the VET vertical line is going to go into the O&RP facility, thereby preventing the market interest rate from falling below the floor. Now that means also though that if there's a reserve demand shock shifting the blue line up and down, even if the Fed doesn't react and keeps the total supply of reserve plus O&RP constant, you're going to get a change in the mix of those two which is going to imply that you can't run over less in the regression. So I'm going to instrument reserves with reserve plus O&RP. It turns out not to matter much if you implement, so if you instrument for deposits, but it is crucial to control for deposits. All right, so here's the estimation then for the US. The estimated equation is at the top. And to emphasize the role of deposits contrast the two graphs at the bottom here. To the left is just the data. It's a Fed Fund's IR spread graphed against reserve plus O&RP supply. You can see it's trying to look like a downward sloping demand curve for reserve, but it's almost like something is pulling on it, pulling it up over time, and that's a growth in deposits. By contrast, in the right picture is the result of the estimation. I'm changing the x-axis there. You'll see that it's a function of both supply as well as deposits. That comes from rewriting the estimation equation as stated in blue at the top to define deposit adjusted reserve plus O&RP supply, which is supply adjusted for the need for supply. You can see you get a beautiful looking demand curve in the bottom right in that case. All right, turning to treacheries. I'm going to use the triblade treachery spread as my starting point for estimating treachery convenience shield. It's despite the fact that there's so much treachery supply currently, that spread is still large. In a second, I'm going to estimate a default component of about 31 basis points, meaning that the convenience shield on treachery is currently despite the high supply is still 35 basis points. This is when you're supposed to have an aha moment and say, wait, there was minus seven basis points before for reserves, and now here it's 35 basis points for treacheries. There's something wrong that is kind of the point. Okay, so I'm going to use a triblade treachery spread as a general, which is a long maturity spread. I'm going to use that as my general measure of the convenience shield on treacheries because there's not a whole lot of term structure, at least down to the three-year point where in the paper I have data to show that. All right, now, again, we need, not just the number, we need a whole function. So to estimate the treachery convenience shield function, I'm going to follow a different paper on mine, where the top left graph is the main picture from that paper with crystal more T. It graphs the triblade treachery spread against treachery supply, thus tracing out the V prime T, the convenience shield on treacheries. Now in the top right, if you add in the post-GFC points, you can see they look like outliers. And that basically suggests that this whole convenience stuff has become even more important after the GFC that the demand curve has shifted to the right. That's because of fat and foreign demand shocks. I'm showing you that in the bottom picture is where on the left, I'm subtracting fat holdings from supply. You see the outlier points migrating a little bit to the left. And then in the bottom right, I'm subtracting out both fat and foreign holdings. You can see then you get back to a more normal looking demand curve. Now from the perspective of our estimation, what we need is the bottom left picture. We need the one that focused on the private sectors of all holdings of treacheries. So assuming again a log linear functional form and including year dummies post-GFC to fit those outliers and an asymptote C to account for the default risk. We estimate, I estimate this in annual data. And that means we're ready for the final numbers. For the US, I'm going to show you both case A and case B, but keep in mind, case B is the empirically most relevant one. Here in case A, given today's deposits of around 17 trillion, I estimate that a beserper's own IP supply of about 3.3 trillion would set the convenience yield on reserves net of the balance sheet cost to zero. That's of course much smaller than the current value about five and a half trillion. Importantly, that number grows over time. Remember, deposits grow over time, deposits matter. So in that I'm illustrating the evolution of this line over time and the blue line is the actual for comparison, the green line to the right illustrates the growth in deposit. All right, turning to the more important case B, we are back to this open the window, open the curtains thing where we equalize the convenience yields. So here I'm graffing in that V prime R, the convenience yield on reserves. It looks small because I have put it into the same picture as V prime T, the one for treacheries, which is much higher. So this is just a simple way of saying there's a lot more demand for the convenience of treacheries than there is for the convenience of reserves because treacheries can be held by others than the banking sector. Okay, so currently, if the fat were to only hold treacheries assets we would be at the A points from a convenience maximization perspective, we should move to the B points where the convenience yields are equalized. That happens at a convenience shield of about 29 basis points for supply of about 600 billion. All right, again, that evolves over time because deposits changes, but also treachery supply changes. So I'm going to skip the math on that. Okay, finally, let me just take a second to do case A for the euro area. So there I'm going to measure the net convenience shield on reserve by the spread between the European short term rate and the deposit facility rate. I've used the only a minus eight and a half basis points in the period before the ESTR is available. There's a spike in the series around European sovereign debt crisis, likely to do bank default risk. So to not have to model that, I'm going to start the estimation here in 2013, assuming again a log linear functional form. And in this case, not having to instrument since the ECB doesn't have an O and RP facility. We get the following result. Focus on the right picture, which is the reserve demand fit for the euro area. You can see the fit is quite good. I do want to flag that the blue line, the fitted line is a little to the red of the data right around zero. So my numbers for the ECB context might be a little too high for the community maximizing supplies. One could work a little bit more on the functional form to get a slightly better fit. In the euro area, it turns out that from an R squared perspective, it actually doesn't matter much to control for deposits because they're highly correlated with liquidity. But of course, it does matter in the sense that if you ignore deposit, you would incorrectly think that the convenience maximizing supply was constant over time, which is not. Let me also just flag that the Bank of England has a great blog post showing the role of deposits for reserve demand in the UK context and how that changes things over time. Okay, so then the number for the case A for the euro area I got is a liquidity supply of about 1.4 trillion. And again, I'm flagging that this is probably a little too high because of that functional form issue. Now, if I can take this couple of seconds, Philip Hardman e-mail me saying, look, I know that for the euro area you're pushing case A, but what would happen in case B? I haven't done a full estimation of this because it's difficult to estimate the B prime T function for bonds because it may have changed with the introduction of the euro and so forth. But the basic point should hold that there is a large convenience you'll learn bonds. It really does matter if the ECB can go with case A. Using the KFV bond spreads as practice for the bond convenience shield, you get pretty large numbers. So if you remember the formula that I had before for the euro case, suppose we set the bond convenience shield to 40 basis points and let's say the ECB only supplied reserves to bond holdings and they did so in proportion to the capital key which is about 20% for Germany. And then the right-hand side is eight basis points which would imply liquidity of about 500 billion. Now that's too low because as the ECB held few and fewer bonds, the bond convenience shield would shrink so the right-hand side would be smaller than eight basis points but this is just to give you the basic point that it really does matter whether the ECB chooses to be in case A or case B. So to conclude, I have laid out a framework for thinking about balance sheet policy when the short rate is above the effective lower bound. I have pushed for a central role for convenience shields and argue that the convenience maximizing resource supply depends crucially on the central banks as a choice which in turn is determined by political constraints. I contrast the case A and case B and argued that the ECB could likely choose to return to case A whereas the Fed has announced that it will be in case B. Thank you very much, Annette. Well, I think that you have exceeded your convenient limit time by four minutes but it's acceptable. It's acceptable. It's in the range, in the range. Okay, now, Jordi, it's your turn. Okay, good morning. Let me start by thanking the ECB for inviting me to this great event and also for asking me to discuss this paper which I have enjoyed very much and I have learned much from. So in my discussion, I will do two things. I will summarize Annette's framework and then I will discuss some of what I think are the implications for ECB policy. Now, first, in terms of motivation, I think the motivation for this paper is straightforward. As part of the normalization of monetary policy, most central banks have started to shrink their balance sheets or are planning to do so soon. So this is the shift from QE to QT. And then in that context, there's a very natural question that arises which is how much QT, no, when should central banks stop? Equivalently, we could rephrase this question by asking what is the optimal supply of reserves given autonomous factors? And this is the question that Annette is seeking to answer in this paper and she adopts a particular perspective to answer that question, which is the one on optimal supply of what she calls convenient assets, that is assets that are highly liquid and extremely safe. So my assessment of the paper is that, well, this is a very nice paper. It brings together some of Annette's previous research. It's extremely timely and highly policy relevant, so it's the perfect paper, I would say, for this event. So what are the key ingredients of Annette's framework? First is the notion of the convenience value of some assets, that is some value that investors attach to those assets beyond the pecuniary payoffs that those assets have. So this may be related to extreme safety or the very high liquidity of those assets. Reserves are certainly a convenient asset from the point of view of banks and we can express the value that banks get from the reserves at the central bank with this convenience value function V of R. Now, there's also a holding cost for these reserves, the reserves which has to do with the capital requirements of holding such reserves and she denotes that by theme. So given this function, we can derive a demand for reserves, which now I'm going to use a diagram that I find somewhat more useful that Annette has used to make the basic argument. So here we have on the vertical axis the spread between the money market rate and the interest rate on reserves and on the horizontal axis the quantity of reserves. So the red line is the demand for reserves from banks. So essentially given the spread on reserves, banks will hold reserves up to the point where the marginal convenience value of those reserves, net of the holding cost, equals that spread. That is exactly compensates for the fact that a money market rate is maybe higher than the interest rate on reserves. So that's the red line. And as you can see, it asymptotes horizontally to minus fee that the holding cost per unit of reserves. So that's a way to reconcile this theoretical framework with the fact that on many occasions we observe that the money market rate is below the interest rate on reserves. Now the central bank will supply reserves that's the vertical blue line and that determines the spread. Now as Annette has mentioned, there is an important separation result that makes this analysis meaningful. And just looking at the equation above that is the equation that describes the demand for reserves, you can see it immediately. In the old days, some central banks had a zero interest rate on reserves. So IOR was zero. In that case there was a one to one mapping between the money market rate, which is the presumably the interest rate that the central bank wanted to target and the amount of reserves. So once the central bank decided on the monetary policy stance that is on R, there was just one value of reserves that would implement that R. Now with interest rate on reserves that changes because any given monetary policy stance, any R is consistent with a continuum of values, of configurations if you want, of values of the interest rate on reserves and the quantity of reserves. So that makes the reserve policy problem meaningful. What is the quantity of reserves that the central bank should hold? And that will determine the interest rate on reserves given the policy stance. So, and that's the problem that Annette attacks in her paper. And she considers two cases. The first case is the one in which reserves are adjusted through purchases or sales off of what she calls inconvenient assets. So, for instance, loans to banks as it was the case in the euro area before 2015. So in that case, the optimal policy, and again, let me use the graph to represent that optimal policy, goes as follows. The central bank should provide reserves up to the point where the marginal convenience value of reserves net of holding cost equals zero. That is, you know, from the point of the central bank supplying reserve is cost less and from the point of view of society in a sense. So it should satiate the demand for reserves from banks. Now, so it should, the reserve supply, the vertical blue line should be adjusted so that we are at the intersection of the demand for reserves and the zero horizontal line, which is equivalent to saying that the spread on reserves that is the gap between the money market rate and the interest rate on reserves should be zero. Okay, and that's something that we observe, okay? In principle. Now, so what has that spread on reserves been in practice? And all my discussion will focus on the euro area case. So look at the red line. Please ignore the blue lines at this point. Just that's the spread, that's the spread between the money market rate and the interest rate on reserves. And you can see that since 2015, it has been negative. So that means that from the point of view of the previous graph, we are to the right of what would be the optimal point, okay? Now, Annette goes beyond that. In other words, there is an oversupply that has been an oversupply of reserves. Annette goes beyond that and she estimates the demand function for reserves. And that allows her to quantify what is the reserve gap, the excess supply of reserves. And in some of the numbers that she gets, I reproduce here the optimal amount of reserves is would be 1.77 trillion as of today. And that's much less than the actual amount of reserves which is 4 trillion. So the prescription would be that the ECB should lower reserves through reduced funding to banks. Now, is the previous framework the one that is relevant for the ECB these days? I would say no because since 2015, the margin that the ECB has used in order to increase reserves and increase liquidity has been not through injections of loans to banks but through the purchases of securities. And some of those securities may have a convenience value themselves when the ECB purchase those securities, it is withdrawing them from the private market. So it's reducing the convenience value of private investors have access to or the amount of convenient assets that private investors have access to. So the ECB or any central bank should take that into account. So we can think of a convenience value for treasuries, say government bonds, whatever assets central bank buys, and of course associated with that convenience value, there is a demand for those convenient treasuries. So investors will hold those treasuries up to the point where the marginal convenience value of those treasuries exactly compensates the interest rate differential between an inconvenient asset, say a corporate bond, and the convenient, say, highly liquid government bond. So that's the gap between RL and RT. So in that context is the optimal policy. Well, the central bank has to take into account that when it injects reserves that are highly desirable, which is highly desirable, it increases the convenience value that accrues to banks, it is actually removing convenient assets from the private market. The optimal policy, and again I will use the previous diagram to represent it, is such that the optimal policy, the central bank should provide reserves up to the point where the marginal convenience value of private investors from holding treasuries equals the marginal convenient value of banks net of holding costs from holding reserves. And that is given by the point at which this intersection between reserve supply and reserve demand at which the spread on reserves equals the spread on treasuries. So it's the point that is represented in this diagram. Now, again, let's look at the data for the euro area. Now the blue lines represent different measures of the spreads on treasuries and that we see that since 2015 they are well above the spread on reserves. So we conclude from the perspective of a net framework, we conclude that the ECB should lower reserves that would bring money market rates up through sales of government bonds that would increase the yields on government bonds up to the point where we equate the two spreads or which is equivalent to equating the marginal convenience value of private investors and central banks. Now, of course, we can ask ourselves, well, why is the ECB policy on reserves so far from optimal, okay? And I think the answer is clear. In recent years, the reserve policy and the policy of buying assets through asset market program has not been driven by the desire to maximize the convenience value to society. But it has been driven by other considerations like stimulating aggregate demand and increasing inflation when that variable was below the target, okay? So now let me just quickly go over some implications for ECB policy, okay? So first is again related to what I just said, optimal versus actual policy, okay? So let me just use this picture from a paper by Philip and Frank that shows the different interest rates that the ECB sets together with the short-term money market rate, the Ionia in this case. So up to 2015, okay, the ECB was injecting liquidity through by funding banks, not through loans to banks. Now, an edge framework implies that in that environment, the optimal reserve policy should have been such that the money market rate should have been equal to the interest rate on reserves, that is the deposit facility rate. But we don't observe that. Systematically, the money market rate was above the deposit facility rate, suggesting that the ECB was undersupplying reserves. Now, this is true up to the beginning of the financial crisis in which between roughly 2008 and 2015, we see that that optimality condition is actually satisfied. But then starting in 2015, the ECB starts injecting liquidity through purchases of bonds, okay? In that case, the optimality condition for supplying reserves should be one in which the spread between the money market rate and the interest rate on reserves should be positive. Instead, it's been zero or a slightly negative as we have seen before. So in the more recent period, the ECB seems to have oversupplied reserves, okay? Now another, and as I said, the reason for this deviation in the latter period may have to do with the fact that obviously the motives for expanding the balance sheet were different from maximizing convenience value. Now, what is not so clear is why there was the deviation from optimal behavior in the early period. Now, I understand there is a debate, second point. A debate within the ECB regarding how to implement monetary policy when the interest rates are already positive and there's this discussion about the corridor or floor system. And I think an edge framework can shed some light on the desirability of the two. So this is what I understand as the floor system using an edge framework. The floor system is one in which the central bank flots, banks, the banking system with reserves, up to the point where the market interest rate does not respond to changes in the quantity of reserves. So that would correspond to this flat part of the demand for reserves. Now, we've seen that this is not optimal under either of the two scenarios that we have considered earlier. This implies an oversupply of reserves in an edge framework. The corridor system is one in which the central bank chooses a stance for monetary policy and an interest rate on reserves and then adjusts the supply of reserves so that it's consistent with those. Now, in that case, what's the main drawback? So this could be in principle consistent with the optimal condition for reserve policy if the central bank supplies liquidity through the purchases of assets. Because the spread on reserves would be positive and it could be such that it equates the spread on treasuries. Now, the drawback from this is that if the demand for reserve shifts around, that may generate a lot of volatility on market interest rates. So an alternative is what is known as a demand driven floor system, which I think is something that could describe well what the Bank of England does, which is you set the interest rate on reserves at what at your desired market rate and then you supply reserves perfectly elastically in order to meet the demand for reserves at this point. I mean, this could be adjusted. So this would be optimal in the case of when the central bank purchases inconvenient assets, but this could be adjusted so that the spread on reserves that is between the money market rate and the interest rate on reserves is positive in a way consistent with the optimality condition when the central bank purchases securities. Now, a quick comment on optimal portfolio management implied by an edge framework. Now, a distinctive feature of bond holdings by the ECBs heterogeneity, heterogeneity in risk, heterogeneity in other types of heterogeneity, liquidity and so on. So for a given total value of the ECBs portfolio, what should be the allocation across different assets? Say assets issued by different jurisdictions. So this paper has a very clean implication, okay? The ECB should equalize the convenience yields across issuers and maturities. That is for any given maturity, it should equalize risk adjusted yields. That is it should sell the bonds with the lowest risk adjusted yields. Now notice that this is very different from closing spreads, okay? Because the equalization is not an equalization of the raw yields, but of the risk adjusted yields. Of course, this may run against political and legal constraints, but these constraints can be changed and I think an edge framework provides a useful framework to think about ways in which these constraints should be and about reasons why these constraints may be changed. Then another question is how to implement QT. I mean, the starting point as we have seen in the euro area is that the ECB has a portfolio with assets that have different convenient yields, okay? And in particular, the spread on treasuries is higher than the spread on reserves. So an edge framework implies that the ECB should keep lowering reserves by selling the bonds that at any point have the highest convenience yields up to the point where the two spreads are equalized. And the implication of this is that for any given maturity, the ECB should sell the bonds with the lowest risk adjusted yields. Now whether the bonds with the lowest risk adjusted yields correspond to the bonds with the lowest yields, I don't know, I haven't made that calculation, okay? But the distinction is very important, okay? Again, there may be political and legal constraints that may prevent the ECB from doing so, but as I said, an edge framework provides a way of thinking, provides a rationale for maybe relaxing those constraints. And finally, the framework of this paper assumes that the total supply of treasuries is taken as given by the central bank, okay? That's B in an edge notation. But obviously in the real world that supply is not an exogenous variable, the fiscal authority has something to say about that supply. And the fiscal authority may internalize the impact that its supply of safe assets will have on investors' convenience and hence on the interest rate that they will pay for them or the price that they will be willing to pay for those assets, okay? So there is a recent paper by Choi, Kirpalani and Perif that looks at the optimal supply of assets from the point of view of the fiscal authority, and they show that if treasuries are partly held by foreigners, treasuries with a convenience value are partly held by foreigners, it may be optimal to under supply those assets in order to keep their yield low. I think it would be interesting, maybe in an extension of this paper, to consider the interaction between the fiscal authority and the monetary authority and to see to what extent your optimal prescriptions for central bank may completely offset this market under supply of assets by the fiscal authority. Just to conclude, because I'm out of time, I think this is a very nice paper, highly policy relevant, very timely. It focuses on just one factor, which is the optimal supply of convenient assets. Now, this dimension or this factor may be overshadowed by other considerations, but I think it should not be ignored at all. Thank you very much. Well, thank you very much, Jordi. You have equalized the time that you have exceeded of your time, so in that respect, et cetera. I don't know, before opening the floor for questions and remembering that you are connected via WebEx, you can raise your ritual hand if you want to ask a question. I would give the floor to Annette. I don't know whether you want to comment something on the comments. I took more than my time initially. Let me just emphasize that I'm not arguing that the ECB has done anything suboptimal in the sense that at different points in time, additional considerations were also relevant. The paper is above the period where the short rate is above the effective lower bound, and what should the ECB do in this context, which is of course different from when the ECB did QE to stimulate the economy. But thank you very much for an excellent discussion and let me open up the floor. Well, thank you very much. And now, I think that Pierre Olivier, you and Andrew afterwards, and there are another question. Pierre Olivier. Thank you, Pierre Olivier Goransha from the IMF. I thought this is a very stimulating paper, really fantastic, so congratulations. One question and one comment. On the paper is a sort of Friedman rule flavor. There is some utility of holding certain types of assets, and the central planner would want to sort of supply until you saturate and bring the marginal benefit to zero. But that's true, and I think that's a point that Jordi made in his final comment. That's true both for reserves, but it's true for government assets. And so if I were to sort of take your paper and say now I want to think about the optimal supply of government securities, I would also want to conclude by the same logic that you'd want to issue whatever it's German bonds or it's US treasuries up to the point where the marginal convenience yield of those assets is also equal to zero, and then the central bank could issue reserves up to the point where there is no spread between the two and then we live in a happy world where all the spreads have been eliminated. And obviously it's not that, but it's a useful long run guide. So I want to push you a little bit and ask, well, do you think that the prescriptions you have for central banks also apply when you think about government issuants of assets? The second comment is, you've been using as a measure of the convenience yield, I was struck by this, the spread between KFW and bonds. And of course in principle, one could think that KFWs and surprise or whether this is also true whether you'd use them ESM bonds or European Union bonds. There is a spread there. It's kind of surprising there is a spread there. There are super nationals, there are guaranteed, et cetera. So that sort of brings the question of what is driving really these convenience yields and if they can move in ways that are sort of not really controlled that sort of shifts the demand. And again, so do we want to sort of have a system where we would have the supply adjusting to that, those changes in the demand? I'll stop here. Thanks. Thank you very much, Peter Olivier. Andrew, I have a lot of questions and we do not have much time. Please be concise and to the point. Thanks. I think that's relatively easy because my question is a sort of slight development of Pierre Olivier in the audience last point. So central banks provide reserves for two reasons. One is monetary policy and the other is financial stability. And I can see that if monetary policy was the only reason then providing reserves against liquid assets would be optimal. But for financial stability reasons that there can be an optimal case to provide reserves against illiquid assets. And we're all trying to optimise the two things and work out as you were saying in that in a sense what the optimal post-QE stock of reserves will be. So the question for me is, I think how you develop the framework to say how do you essentially derive the optimum where it's a combination of monetary policy and financial stability reasons? Thank you very much, Andrew. I think that you have the next question. And I have three more questions. I have Ricardo, Janis and Vito. And I think that I will close, you know, because otherwise we are going to exceed the available time. Please go ahead. Fogavilan, Goethe University, Frankfurt. Thank you for the thought-provoking paper and discussion. Quick question, one, it seems the main driver of asset purchases was really crisis fighting. First, the crisis fighting of the deflate or disinflation, deflation risk following the Euro debt crisis and then the brief period of the Corona crisis. And we got to a stage of between 25% of public debt to close to 50% of public debt being held by the ECB on the ECB balance sheet, you know, depending on which country you take, 25 for Italy. So it seems there is a big, if you are in a crisis, there's really a big increase you might want to implement. And so that concern that having enough room, having reduced the balance sheet enough so that in the future one can do that again should also be one, I think, factored in. And actually it also matters outside of the zero lower bound because the TPI is a new instrument precisely with the same objective to buy these assets with positive industries. So, you know, if you, in two crisis, go to between 25% and 50% of public debt, that would argue for really reducing it maybe much more public debt holdings than what the framework you presented implied. And also it might have the opposite sequence of what Jordi said because if it's about crisis fighting, you want to, you know, in good times, lower the holdings of those sovereigns who are more likely to get into crisis, right? Okay, thank you very much, Annette, with our full questions. Yeah, let me reply first to Pierre's question. So, in the paper, I argue that one should equalize the convenience shield on everything that's the optimal thing to do. That's not the same thing as supplying more government bonds to set convenience shields all to zero given that taxation is likely distortionary, right? In order to keep the, there's a reason we go through these endless debt ceiling dramas in the US, right? So we can keep that safe and liquid. So, I think your prescription is correct only if taxation is not distortionary in which it is in practice. In terms of the KFV bond spreads, I think they probably understate the convenience shield on bonds precisely because of those government guarantees. They may pick up mostly the liquidity component whereas the safety component is not in there. So if anything, my results will be stronger accounting for that. In terms of the contrast between monetary policy and financial stability, I don't think they point in different directions in the sense that I suggest supplying reserves with, quote unquote, inconvenient assets when possible, which could be lending to banks, which is the same as you would do in a crisis, understand the central banking principles. And I think in terms of also the headroom argument, these are all valid points that, remember I had the suggestion that you say, okay, start from my numbers and adjust up and down depending on the weight you put on other factors. I haven't quantified how to do that, but conceptually, if you're very concerned about headroom, you could want to go lower. If you're more concerned about in-straight volatility, you would want to go higher than the numbers that I prescribed. Thank you very much, Annette. We have three more questions. I have Ricardo, Janice, the governor of the central bank of Greece, and Vito Constantia, and we close. So please to the point because we are sitting on that. Absolutely, two questions to the point. First, why don't we mention other factors? Those factors sometimes are very correlated with the fact that you're emphasizing. So I'd like you to expand on how they correlate. One, in my, when I ask the same question as in your paper in Jackson Hole in 2016, I focus instead on the central bank losses and profits. And the key determined there was the maturity of the treasuries that the central bank buys. If you buy is very short, you can have a much larger balance sheet. If you buy is very long and you have a much larger one, because as worked by Seth Carpenter and myself many years later had shown, this leads to a more volatility in those central bank losses. And as the last experience you confer. How do convenience yields vary across short-term treasuries and long-term treasuries? Because that will interact very strongly with this other motive for the size of the balance sheet. Second question, then a year later, challenged by Pierre Garincia at the IMF annual conference, I wrote a paper on what's the optimal size of the balance sheet if we have different bonds or different default risks in a monetary union. And there are the key consideration, very much along the lines of what Andrew just said, was financial stability and on how the size of the balance sheet depends on whether you can absorb more or lesser that default risk, which then affects bank's ability to lend and the risk they can take. There likewise you said control for deposit is very important, from a fast stability either for the liquidity reasons Andrew said, or the risk reasons in my IMF annual conference paper, it will be the loans in the total amount of loans. How do I distinguish deposits of loans given their clear correlation and given that the size and risk-in-size loans is going to determine also the optimal size of the balance sheet? Thank you very much, Ricardo. Jarnes o Stunaras. Thank you. This is actually a very useful and nice paper. I have the following question. Given the fact that the architecture of the eurozone is far from perfect, one of the objectives we have in DCB is to avoid fragmentation. So my question is, if we follow the rule that you propose, what would be the implications for fragmentation in the euro area? Or can I ask it in a different way? Do you assume a perfect capital market for your result? Thank you very much, Jarnes, and finally, with the question here, I did not deserve this. Thank you. Well, you chose to determine the size of reserves and then the balance sheet, the criteria of maximisation of the net convenience value of supply of reserves. It's a very particular objective and perhaps not the only one because it's not directly related to monetary policy because indeed you said at the beginning that above the ELB, the size of the balance sheet doesn't matter. I don't agree with this statement, the initial statement, another criteria to discuss the size of reserves and the balance sheet is and should be related also to monetary policy. Because the topical discussion that is ongoing about the size of reserves and the size of the balance sheet has to do with the choice between implementing monetary policy with a corridor system or a floor system. And the approach you take in the case that the supply of reserves is backed by purchases of sovereign bonds also with convenient yields leads to a very small balance sheet to operate the floor system. And I see there a great difficulty. And the floor system is important because both the ECB and the Fed have been operating the ECB approximately a floor system since 2008 because we had fixed rateful allotment in October 2008 and then we had the LTROs and that created the situation of excess reserves and the money market rate approach very much already since 2008, the DFR, the deposit facility rate. And then of course after 2015 it really was practically there since then. But and the floor system has worked well and has to be advantages to the operation of monetary policy. One is that it's easier and more certain to choose and determine the policy rate. If it would be through auctions and the corridor system, the ECB would have to organize now auctions of almost $2 trillion which would be messy and very difficult to achieve the desired money market rate and the volatility that Jordi mentioned could very well be the consequence of that approach. So with the floor system you have the fixation, the determination of the policy rate and you have other advantage that you have a second instrument of monetary policy because you have the determination of the policy rate and you have the size of the balance sheet being separated from that determination of the balance sheet. And if ever for possible reasons you need some QE, then you can do it and keep the desired money market rate that you want. So we test these two big advantages, the floor system and this is related to monetary policy implementation and it's not covered by your objective function of maximizing just the net convenience value of reserves. How do you think these considerations impinge on the discussion about the desired size of balance sheet and if your approach can be made compatible with the managing of the floor system? Thank you. Well thank you very much Vito, Annette and I will give the floor also to Jordi. So Annette, Vito. Let me just do it in order. So, Ricardo, I have investigated the maturity structure, the jump structure of the convenience shield using a different approach. There is not much in the US down to the three-year points where you have data. Bills have lower convenience shields. So the prescription following your thinking would suggest that the Fed might benefit from going a bit shorter in maturity than just holding their representative treasury out there. In terms of the fragmentation, I think that's a different issue than what I bring up here. My convenience shield can't go negative. Fragmentation is a fundamentally different issue and the paper really does not speak to that. So, in terms of a consensious question, I disagree that central banking is not about providing payment systems. I mean central banks, they historically printed money for good reason. Many of them directly as a mandate that they're supposed to facilitate payment systems. So I view the whole supply and safe and liquid asset as a natural extension of something that central banks also are supposed to do. And we'll need to have a longer discussion since we're already running over time about the very good question that you asked subsequently. Okay, thank you very much. Well, I think that this is a very important contribution. I will read twice your paper before deciding on QE the next time. But, well, it's a partial equilibrium analysis, so I think that we need to have other kinds of factors and to take them into consideration when we decide about the evolution of our balance. So, thank you very much. This brings to an end this panel. Thank you very much and well done.