 So, I've got a question to Luke because I'm really in your hands. Can I start earlier or do I have to wait until 4.45 sharp? Okay, so let me open this panel on exiting non-standard monetary policy measures. That's a fascinating topic. It's a very timely discussion, so I will be listening very carefully and taking notes. One point I wanted to make before we jump in the discussion is about the gender balance on this panel. This is an all-male panel. We've tried hard with Luke and with the colleagues. But for some reasons I'm not going to explain here, it just didn't work out. But it's something that we certainly have to improve in the future because it's certainly not our policy to have an all-male for that matter all female panels here at the ECB. So I just wanted to make the point. Second point, which is housekeeping. We have one hour and a half, which is plenty of time. As suggested by Luke, we can end a little bit earlier if needed, so don't feel any pressure either way. And I'm still going to keep you to the timeline because we want to have enough time for the discussion with the audience. So with that, let me start and thank very much the speakers for being here. I'm not going to do the introductions because they are all very well known. And we're going to have a variety of angles into that discussion. Jeremy, as I understand, will discuss issues related to the liability side of the balance sheet of the central banks. Hewn will discuss some of the more market-related issues related to the impact of large-scale asset purchases in the market and investor behavior. And Olivier will wrap it up and come back, I guess, to the core monetary policy discussion, which is how should we exit? Also, drawing on what Hewn and Jeremy would have said. So without due delay, Jeremy is off the floor. Okay. Well, thanks very much. Thanks, Benoît. Thanks to Luke for having us here. So as Benoît said, I thought I would try to talk about the liability side and I'm going to focus particularly on the Fed and focus on that. And you might think that that's a little bit curious because obviously most of the focus in the QE era has been on the asset side of central bank balance sheets. And in particular, you know, how using the asset side can act as a substitute for monetary accommodation when you're stuck at the zero lower bound. And of course, in that world, it matters what assets you buy. And so the Fed and other central banks, to the extent that they've been able to, have tried to buy longer duration assets, for example, because you get more push on the term premium that way, the Fed decided to buy mortgages because they wanted to get just a little bit more bang for their buck altogether. And there was some thought that, you know, working on mortgage spreads was going to be helpful. So that was kind of the mindset was really on that. And so as the Fed is starting to run now QE essentially in reverse, it's very natural to ask, well, what happens when you reverse all those assets and you start, you know, liquidating them and you know, who's going to talk to sort of those sorts of issues. What I want to do is point out what I think is essentially an important, I don't know if it's a coincidence, it's quite the right word. But it's, yeah, it's kind of a coincidence, which is that the QE era, you know, it was a monetary policy effort. But it happened to coincide with fairly dramatic changes in financial markets and in regulation in particular. Okay. And the point I want to make is that some of these regulations, and I'm thinking particularly of things like the liquidity coverage ratio, margin requirements, central clearing have, were essentially creating a much increased demand for various safe assets. Okay. And then you might have wondered, well, geez, wouldn't that have put a strain on the supply of safe assets? And why didn't we see sort of scarcity? And why didn't we see the system creaking in various ways? Well, it might be. And this is just my conjecture that coincidentally, the Fed, by massively increasing reserves, was creating at least a certain kind of safe asset that was actually very helpful. So things went smoothly basically because of this almost inadvertent supply effect that was just the sort of balance sheet mirror image of QE. So my point is now, as we start to unwind, one ball that you want to keep your eye on is what's happening there, and will we see some of the strains that were somehow papered over finally starting to emerge? So just, you know, most of you guys know these numbers, but just to kind of put it in context, Fed's balance sheet is just about four and a half trillion. On the asset side, very simply, think of that as being about two and a half trillion of treasuries, 1.8 trillion of mortgage-backed securities, and some dogs and cats. Less attention, as I said, to the liability side. That's about 1.6 trillion of currency, 2.3 trillion of reserves, and again, some other stuff like treasury deposits and a smaller repo program. Okay, but again, 2.3 trillion of reserves, and to a first approximation, that number is going to change by a lot. So interestingly, and I think actually very intelligently, while the Fed has been quite precise about the near-term trajectory, they have not called out the end point yet. I think that's smart for reasons I'll sort of get to. But, you know, the kind of numbers that people talk about are, you know, it could shrink by 1.5 to 2 trillion, the balance sheet. So, and that would be essentially those 2.3 trillion of reserves coming down by that very substantial amount. Okay, so why would that matter? Why would that matter apart from what's going on in the asset side? To be clear, I think it only matters, if you want to convince yourself that it might matter, you have to start out by believing that there's not only a well-defined, you know, there's a lot of work, a lot of research on safe asset demand. You have to go a little further. You have to believe that there's a sort of segmented safe asset demand so that there is a particular demand for short-term safe assets. Because if you think about what's going on, the aggregate supply of safe assets available to the non-fed public is not going to go down because as the reserves are going down, the, you know, treasuries in the hands of the public are going up. Okay, so total sort of government-provided safe assets are not going down, but of course the short maturity stuff, the reserves are going to go down. So, you know, premise number one or sort of friction number one that you have to buy into is that there's a special demand for this stuff. Empirically, that seems to be well-founded. We've done some earlier work and Arvind has done some earlier work. And if you just look at, you know, variation in things like the supply of T-bills to GDP, when there are fewer T-bills in the system, the price, if you will, goes up. That is to say the slope at the very front end of the yield curve gets steeper. So first observation I want you to keep in mind, the yield curve unconditionally, pre-crisis, so think of sort of 1980 to 2009, is astonishingly steep at the front end. So the average difference between a one-week T-bill and a six-month T-bill is somewhere between 40 and 60 basis points. That may not sound like a lot. That's a lot relative to the risk. Okay, so that's the sort of unconditional number and it moves around in response to variation in the supply of T-bills. So when there are more T-bills, when there are more T-bills, it flattens out. When there are fewer, it steepens, okay? And you can do that with sort of stupid kind of unconditional regressions. You can be a little clever and you can instrument, for example, with the tax calendar. So there's seasonal variation in T-bills outstanding because the government needs to borrow more before they get the tax receipts and then when they get the tax receipts, they borrow less. And that traces out also this variation in the slope of the yield curve. So there's some evidence to suggest that there's a demand not only for safe stuff but for short-term safe stuff. Think about money market funds, for example. What do they want? They don't want to hold long-term government bonds. They want to hold short-term stuff. If you believe that, if you believe that that's sort of a somewhat segmented market, then the magnitude of the shrinkage that we're going to get is quantitatively a big deal. So as I said, reserves are going to go down by one and a half to two trillion if they go all the way. To benchmark that, the quantity of what are other sort of short-term safe things, T-bills of maturities less than 30 days, it varies. But think of like 400 billion as being the number, okay? So if you think of the aggregate supply of short-term government safe stuff, you're going to get a huge impact or potentially huge impact in the shrinkage. All right. So let me be a little bit more concrete now because I'm sort of very loosely talking about T-bills and reserves like it's kind of all the same thing. That's not quite right. I mean, now that there's interest on reserves, it's sort of kind of the same thing. But of course, reserves are kind of like restricted T-bills because only banks can earn the interest. So let me be a little bit more specific. Who's holding those 2.3 trillion of reserves right now? Obviously, it's all banks because they're the only ones who can get the interest. What kind of banks it seems like the best data I can get, the first several hundred billion are probably in foreign banks who are doing what's called the IOR arbitrage. They're basically borrowing in the funds market and turning around and investing it with the Fed and earning a spread that right now is about nine or 10 basis points. They're just doing it because they're sort of advantaged in doing that arbitrage. And presumably, they're pretty elastic. So as the supply of reserves goes down, rates will only have to move a little bit for that to be un-tracted for them to do. So my best guess is the first several hundred, first let's call it 500 billion will roll off without any real big deal. So the first year or so of the balance sheet shrink, I think, basically these foreign banks just do a little bit less, no big deal. At that point, I think it at least gets interesting. You know, I don't know what's going to happen, but here's where the regulatory induced demand, I think, is going to be potentially relevant. So the other big category of if you just look at who holds reserves are not foreign banks who are just doing arbitrage, but they're the big U.S. banks, think of J.P. Morgan in particular, who do this to comply with this new regulation, and the liquidity coverage ratio. Okay, so again, just to give you some numbers, it looks like the overall, so, you know, the liquidity coverage ratio tells you you have to hold what are called high-quality liquid assets. Reserves count, Treasury securities count, some other stuff kind of sort of counts. Okay, the overall demand on the part of the big G-SIBs, on the part of the big banks, estimated to be about 2.7 trillion. Of that, about one trillion of what counts as HQLA seems to be held by the big U.S. banks in the form of reserves. So they are using reserves quite heavily to make up this demand. Okay, so they play a substantial role. Now, here's the tricky question. This is the elasticity of substitution that I do not know what it is. Okay, so in other words, one theory of the world would be, well, reserves count as high-quality liquid assets, so do Treasuries. When the relative supplies of these two switch, the JP mortars say, that's fine, I don't really need to hold reserves, I'm happy to hold 10-year Treasuries. Okay, in which case, the whole thing is like a non-event. Or, there could be a reason why they're holding it all as reserves, that they actually have a reasonably well-defined preference for holding it as reserves rather than as 10-year Treasuries. They don't want the interest rate risk. They want the intraday liquidity, essentially. I mean, it's very related actually to some of the stuff that Monica was talking about, that reserves provide a function, especially in this sort of HQLA world, that is not perfectly substituted by bonds. So, if that is the case, they may be a little bit more reluctant to give it up, and then you've got something of an effect. Okay, and again, a little bit more broadly, even if the banks say they only have a preference, not over Monica's sort of liquidity thing, but just over maturity. So, suppose a bank says, fine, I lose my reserves, I'm totally happy to hold like 30-day T-bills. Okay, but then those 30-day T-bills have to come from somebody. They tend to take it away from the money market funds, and then all you need to believe is that there's a segmented market between sort of short-term bills and long-term stuff, and again, you get something. So, again, I don't pretend to know, but I think just based on sort of previous evidence, I wouldn't be surprised if you got something of a price effect. Okay, so what's the big deal? So, you get a price effect. You know, again, what would it manifest as it would be something like the yield curve, which is currently quite flat over the first, you know, six months or so, even if it just reverts back to its sort of normal slope, or it might get even a little bit steeper than that. So, again, think of there being, you know, 60 basis points of slope between one week and six months. So what, the yield curve's always kind of looked a little bit like that. Now, one simple point, I think the less interesting thing, is it makes the regulation more expensive for the banks to comply with, because now they're going to be holding all the short-term stuff that has a very low yield, so they may feel, well, that's kind of not nice and unfair, and they'll, you know, they hadn't complained as much about the liquidity coverage ratio. They may complain about it more, you know, that's one thing. To my mind, the much more important thing is think about what this does. If the yield curve does slope up like this, you know, 60 basis points over the first six months, think about what this does to the incentives for private sector firms to do maturity transformation. So I think, you know, and this is work I have with Rob and Greenwood and Sam Hansen, we think sort of an underappreciated aspect of the pre-crisis period, you know. We know that the story is too much maturity transformation by the private sector, right? All these guys, Lehman and Morgan Stanley, all funding themselves short-term, we've passed a lot of rules telling them, don't do that anymore, it's bad. It has externalities and you shouldn't do it. But, you know, you've got to think about the economic incentive. That's a very high-sharp ratio. Okay? Again, 60 basis points, but it's 60 basis points against not a whole lot of duration risk. So if you're a private sector firm and you're choosing between borrowing at the six-month point and at the one-week point, you save a lot of money. And so one reason to not want to let that reemerge is I think then it makes it much harder. In other words, we've passed a bunch of regulations since the crisis and guys have apparently not been funding as aggressively at the short end since the crisis. You might get a little complacent and say, oh, our regulation is working. I think one reason it may be working is it hasn't had to fight in the teeth of an incentive that was as strong as it was at some time prior to the crisis. So, you know, I think that's something that should be monitored and it's not to say that you shouldn't be doing the best you can with regulation, but if you make it too hard for regulation, if you make the economic incentive sort of overwhelmingly strong, I think it's tougher. So, you know, what should the, what can you do? I mean, I think the Fed has wisely not committed to an endpoint and I think a sensible thing is, you know, let's watch, let's see, this could be wrong. Maybe it's not a big deal, but if you start seeing signs of either the yield curve getting steep or various spreads that are suggestive like T-bills to OIS, you know, widening, you know, you might want to either be willing to slow down the rate of shrinkage or adapt in some other way. With that, you know, a bit of communication, I think the Fed might want to start socializing the idea that what matters when you're thinking about the size of the balance sheet is not literally how many dollars of assets they have, but in a sense, the total amount of duration that they're, you know, if you think about it from any, you know, reason you would want to have a small balance sheet. It's because you want the Fed to have a smaller fiscal footprint. There's two ways to get a smaller fiscal footprint. One is to just pro-rata shrink everything. The other is to take some of those eight or nine-year bonds that you've got and start holding three-year bonds on average. So they could try to communicate the idea that we will be shrinking our fiscal footprint, but we may be doing some of it by just holding shorter maturity stuff that would allow them all else equal to keep more kind of short-term claims out there. Of course, if, you know, if the Fed doesn't do it, the Treasury can always step up, okay? The Treasury, you know, so this is just a matter, you know, the Treasury can decide, well, we'll issue less than the way of long-term bonds and more in the way of short-term bills. That would work as well. I mean, again, the Treasury historically has been reluctant to do, you know, they're sort of stopped out at about four or 500 billion. The idea that they're gonna add a trillion to the very short end is probably not likely, but I think this is a case where, in spite of all the sort of taboos against it, it makes sense for the Fed and the Treasury to talk to one another. Last thing, I've been sort of talking about this all in the context of the balance sheet shrinking in isolation, but I think this interacts in an interesting way with rates going up in the following, and I think it takes on a little bit more urgency in a world where rates are going up, because what do we know about banks? When rates go up, what do they do? They basically don't raise the interest rate that they pay on their sort of core deposits, you know, savings and transactions deposits. As a result, you know, some people just stay there and get stuck with a low interest rate, but the more sophisticated people like everybody in this room, basically takes their money out of those accounts, puts them in a money market fund, and then it gets recycled back to the banks in the form of more wholesale funding. In other words, the bank's funding mix goes from being more retail deposits and less wholesale funding to relatively more wholesale funding, which just makes all this stuff more urgent, because in other words, the wholesale funding tends to be runnier, flightier, and so at the point when they're doing more wholesale funding, I think it's precisely when you would like to have them doing it six month wholesale funding rather than one week wholesale funding. So again, I think all these issues will come a little bit more to the fore then. So again, my best guess as a forecast, nothing too exciting will happen for the next year or so, but then after that, I think it's at least an open question and it would be good to kind of keep a little bit of optionality available. Thank you very much, Jeremy. Let's move to Hyun and from the short end to the long end of the curve, if I understand it well. And then we'll have a round of discussion when we can connect the dots, not yet. I prepared some slides. So thank you, Benoit. You've given me the task of explaining why long rates seem to be so resilient in the face of monetary policy normalization and I think this is something that's familiar to a lot of you and that certainly is making the rounds in market commentary. In the long rates in all the major jurisdictions, they're pretty much where they were last year, this time last year. Certainly got a lot lower in the middle of the summer last year after the Brexit referendum in the UK when long rates in some jurisdictions went negative. But we are one year since this time last year and yet long rates haven't really budged that much. And one line of argument is, well, markets are far sighted. We do personify markets and endow them with foresight and so we tend to take market prices as signals and so if prices are not reflecting monetary normalization, well, do the markets know something that we don't? In fact, if you run a standard term structure model on current set of prices, the risk premiums, the term premiums actually come out to be negative in many cases. And so there is a bit of a puzzle but I think another way of thinking about this is to take a slightly more skeptical stance because the market is not a single individual with foresight. It is a place where lots of different actors interact and market prices are the outcome of that interaction rather than the beliefs of any one person. And so if we look into some of the details of how the market participants behave, perhaps we can gain some insights into what is driving some of the market action. So what I thought I'd do is to give you a glimpse of one small example of this by looking at the very long part of the yield curve, so the ultra long bonds, by which I mean sorry, in bonds with remaining maturity greater than 20 years. And we have some ongoing work that looks at the micro database at the Bundesbank for the German residents. And we've taken the ISEN level, the security by security level data, but of course we don't have access to the detailed data but we go there and aggregate, we get the aggregate output. And what I'm gonna show you is the result of some summary statistics of around 4,200 ISENs. It's held by German residents and it's gonna be aggregated at the sector level but this is only going to be looking at the seven Euro area sovereigns, so it's if you like the core Euro area countries, Austria, Belgium, Germany, Finland, France, Luxembourg and the Netherlands. And the focus is gonna be very much at the ultra long end but let me just show you some summary statistics just to set the stage. So here's a snapshot of the distribution of the insurance sector in Germany and it's the histogram of the holdings of the sovereign bonds of those seven Euro area countries arranged according to the maturities and into these maturity buckets. You see that the insurers hold relatively long maturity bonds between 10 and 30 years. And this is a snapshot as of the end of March 2013 and if we turn the clock forward and look at the situation as of March 2016, you see this rightward shift. And I should emphasize that these are notional values rather than market value, so these do reflect the underlying shift in the holdings rather than the changes in the valuations. Now, the other question might be, well, what's happened to the overall outstanding amounts of these bonds? Is this just a reflection that there are just many more longer term bonds outstanding? And the answer is no. In fact, if anything, at the ultra long end, so the last two categories, you see that the total amounts outstanding has, if anything, fallen and yet there is this shift towards the holding of these very long maturity bonds. Now, we can arrange this holding of the ultra long bonds as a time series and here is a chart that shows how the holdings, again the notional amounts have evolved since 2005. And the message here is that there's been, as you would expect, there's been a very big shift upwards in the holding of these ultra long bonds. So to give you a sense, compared to 2008, compared to just before the crisis, the holding by German insurers of these ultra long bonds of this subset of Euro-Area sovereigns has more than quadrupled. So there's been a huge increase in the holdings. And that's coincided with the decline in the long-term interest rate. Here is just a plot of the 20-year swap rate. And if we just put that into context, if we plot the other sectors in Germany, what you see is that there has been a shift. The yellow is the investment fund sector. A big chunk of that will also be held by the insurers, but we don't actually have the detail breakdown. The blue is the interesting part. The blue is the banking sector. And for the banking sector, that's really shrunk. So there's been a shift away from the banks towards the insurers and also to the investment funds. If we just look at the proportions, you can see that much clearer. The blue is the banking sector, where that's really shrunk. It goes from around 10 billion euros notional to less than five. Whereas as I said earlier, we have this huge increase in the holdings by the insurance sector. Now for those of you who are familiar with long-term investors, I think you can guess what was going on. I think one story here is that this shift in the holding of long-dated bonds reflects risk management motives, where if you have a balance sheet where the liabilities are of a longer duration than the assets, and there is an imperative to match the duration for interest risk management purposes, then you may see dynamics which give rise to behavior which may seem at first sight somewhat perverse. And the story here is that life insurance firms in Germany, as well as in many other European countries, serve a very important purpose in providing long-term savings products. And the liabilities have very bond-like attributes because they have obligations to policy holders which are often in nominal terms and which have bond-like flows. And those nominal liabilities tend to be of a much longer maturity than the assets they hold in order to back those liabilities. And here's a snapshot from the stress test report from EIOPA, the European Insurance and Occupational Pensions Authority as of 2013. And the way to read this is on the horizontal axis, we're measuring the duration of the assets. And on the vertical axis, the duration of the liabilities. And the dots represent the combination of the asset duration and liability duration of a number of European insurance sectors arranged by country. And you see that Germany is this observation up here where the asset duration is around 10 years and the liability duration is around 20 years, meaning that the liabilities are of a much longer maturity than are the assets that are held to meet them. The fortified realign represents the set of points where the assets and liability duration are equal. And you can see that most of the observations lie above the fortified realign, meaning that for most countries, what's true of Germany is also true, but perhaps to a lesser extent. But the key here is that as well as this duration mismatch, there is what's called negative convexity, which means that as interest rates fall, of course, the duration of both sides are extending, but liability's duration, because it's longer, is actually increasing faster. So there's a convexity here. If you imagine a pair of scales, imagine that you have the asset side of the scales and the liabilities side of the scales. And the balance means that you have the interest rate risk roughly balanced. But as interest rates fall, because of the longer liability duration, the liability side of the scales is tipping down. And so the response is that you have to add more long dated bonds to the asset side in order to even out. So what's just happened? As interest rates fall, you're buying more long dated bonds. So as prices rise, you're actually buying more. So it's a kind of a perverse demand response, but it's a perfectly understandable response because it is arising from the imperative to manage risk. And it may be reinforced by regulation, but certainly prudent risk management is gonna be enough to induce that kind of behavior. But the question is, what will this do to market dynamics if a sufficiently large segment of the market is behaving in this way? Could there be the possibility of a feedback loop where as interest rates fall, there's a greater demand for long dated bonds? And if that demand is big enough and the residual supply is not going to be large enough to offset it, you may see a price dynamic which pushes long dated yields even further. And just to give you a sense of what these demand curves look like, I've just here plotted the holdings of the insurance sector against the 20 year swap rate. And you do see this on the left hand panel, these are just in levels. You do see this very striking downward sloping relationship where the lower the interest rates get, the more the insurance sectors is holding. So on the surface, it seems that there is a downward sloping in this space, demand curve. In price space, of course, it will be the reverse. It will be an upward sloping demand curve. Of course, this is in levels, you might say, well, that doesn't tell us very much. What about in one quarter changes? And the right hand panel gives you a sense of what's happening in terms of the one quarter changes. And it's, of course, much noisier, but there is this negative relationship which seems to be there. And what we're trying to do is to dig deeper. Because we have this icing level data, we can actually do some pretty cool time series analysis. And it does seem that there is this element of a feedback loop that you see the impact going in both directions. When the insurance sector buys more, then it does seem to impact long rates in the near term. And as long rates fall, that seems to induce a greater buying. Now, the question is, are we likely to see something like this, but exactly in playing out in reverse when we see a normalization of monetary policy? And I think that is a really important question for us to ponder. But I think before going to that question, I think there are some, I think much less controversial questions, much less controversial lessons we can learn from just even these very simple charts, which is that first of all, we may take heed of nominal rates much more for their own sake than purely as an input into real rates. Because typically as economists, we care about real rates and when we talk about our star, it's very much about the real rate. But to the extent that asset holdings, to the extent that portfolios depend on the asset liability mix, and that's certainly the case, that seems to be the case for the insurance sector, nominal rates may have an importance in its own right for the equilibrium yield in the bond markets. And the second question, and this I think does bring us closer to the big questions, is we know from textbook comparative statics analysis that if you have an upward-sloping demand curve and an upward-sloping supply curve, that even a very small shift in the supply curve could have a big impact in the equilibrium outcome. And so as we look ahead to monetary policy normalization and we eventually think about reducing the balance sheet, that is going to put more supply into the market and to the extent that there are some players in the market who are behaving according to some myopic rule, for good reason, but nevertheless myopic rules, which are different from the far-sighted behavior that we normally attribute to them, we may see some sensitivity there that we may find exposed to have been surprising. And it also raises some methodological questions about how we can think about long-dating deals because we normally take the asset pricing models as giving us some decomposition between expected long rates and the risk premium residual. And that we find very illuminating as a first pass, but in some ways it is really postponing the big questions because you're not really addressing the question of why the risk premiums have been shifting around. And I think this is a way to perhaps recast some of the questions. If it is not a single rational individual that is the marginal investor here, but rather the interaction of many different individuals, perhaps we should look at some of the micro evidence a bit more. And I think there is also some interesting set of questions which have to do with how this type of amplification mechanism relate to the other ones that we're much more familiar with. For example, with the mortgage-backed security hedging demands that have been known to cause fluctuations in the bond markets in the US. So let me just conclude there. Thank you very much, Heun. What are you? I'm going to go over that. What are we? Do you want to? Okay, we can do that. So, relative to Jeremy or Ian, most definitely an ignorant outsider, but that's good because it allows me to ask naive questions without looking like a fool. So that's what I'm going to do. So I thought I would focus on the issue which is what is the, what should be the ultimate size of central bank balance sheets. And then only speak briefly about speed as opposed to final resting point. So again, most of my time is going to be spent on the first and then I'll say a few things about the speed. Now Jeremy said, you know, the Fed is smart not to tell exactly where it's going to go. And I could see the point which is if you don't exactly know how things are going to work out, you don't want to commit in advance. But, I mean, at some general level that seems a bit strange to basically start going in some direction without knowing where you're going to end. I mean, let me take two examples. The first one is if we concluded that actually the size of a central bank balance sheet should be larger rather than smaller because we've learned things which make us think that it would be good. Then reducing it doesn't seem like a very good starting point. More realistically, if we think the size is kind of okay or maybe a bit large, but we think of the composition on the asset side is not what we want because it's a result of accidents along the way. Then again, you probably would not want to start just reducing the balance sheet. You'd want to know how you're going to go from the existing balance sheet site size to the kind of assets that you want to hold. It seems to me it's important, even if it has to be contingent in the end, it is important to think about where the process will end. And then on the speed of adjustment, I'll make some remarks. I will start with something which I think is obvious to us, but is really important, which is then we need to think when we're talking about QE and balance sheets, we have to think about the consolidated balance sheet of a central bank and the treasury because fundamentally balance sheet operations of the central banks are government debt management. And in the end what the public holds is determined by the decisions of the central bank and the treasury together. So I think we have to think when we ask the question is do we want a different composition of government debt in the hands of the public, including the possibility that actually the government holds some private assets which is what some central banks do. But we really have to start from the other reasons to think that the yield curve or the term structure or whatever you want about the various premium is what we want or do we want changes either permanently or temporarily. Then the issue, if you think this way, then the natural issue is, well, who should do what? And I think then it has to do with the division of labor, what the central bank knows better than treasury where the central bank can act faster than treasury. So for anything which is relatively permanent then it seems to me that in the end treasury is where the action should be. So I had not thought about the issues before I was asked to, not seriously thought about the issues before I was asked to come here. And I went through the literature of I must admit that I missed the Jackson Hall Conference of last year which apparently was on that topic but I was in France on the beach and I missed it. But so some of what I'm going to say was that if I had known of the existence, this is a paper by John which is if I had known of the existence of the topic of the Jackson Hall Conference I would have looked but not knowing, I had no way of knowing that it was there. Anyway, I concluded that there were five arguments of which I think one is a good one and four are not good ones. Not surprisingly if a good one is by Jeremy. No, I think Jeremy has made a paper that underlies his presentation is an outstanding paper. I think it's extremely convincing on the case for providing very short maturity liabilities. I'm not going to repeat the argument because Jeremy has done it very well which seems to be a demand at the very short end of the year curve and we're really talking short end. I mean basically a few months at most which is not satisfied in the sense that we see a fairly large discount at that end and the private sector sometimes tries to fill it and we're not sure that that's a good idea. So I think the argument is there and I think it's fairly convincing. Now the point again is that there is no particular reason why this should not be done by Treasury. I think that Jeremy is offering as a solution could be equally well offered as the issuance of infinite maturity floating right that. I think it would have the same characteristics. It would avoid having auctions all the time which is one of the issues when you have short maturity and the Treasury could buy and sell it at par, buy back if it needs to decrease it. I have a sense that maybe the central bank is more aware of needs of the financial sector and can adjust faster but the example you gave about seasonality of tax receipts I think it's something that Treasury knows very well. I mean central bank can know it just as well but no better. So the question is should we basically keep a large balance sheet? Not so much because of the asset side but the liability side which is what Jeremy has argued. Am I convinced? Again Jeremy talked about it. I mean how much do we care about the 40 to 60 basis points at the very short end of the yield curve and he has given some arguments for why we might care a bit but in the great scheme of things this is for an ignorant macroeconomist like me. It looks like small potatoes. It might be important. The other thing to say is that Jeremy gave this figure about the average wedge at the short end of 40 to 60 basis points until the crisis. If you look today the wedge is not gone it's about half of what it was and this is after the balance sheet of the central bank namely the Fed in this case increasing from 1 to 4.5 trillion. So if I extrapolate which might not be the thing to do then that suggests that the balance sheet of about 8 trillion would be about the right number. Jeremy is fairly careful actually about not giving a number for what size balance sheet there should be. Clearly the larger it is even if you reduce the duration of the assets you're still going to get increasing maturity mismatch which risk but in his paper he says that 4.5 trillion could be fine and suppose it were then what are we doing trying to reduce it? I mean we have to think about it. So I think on this distribution is extremely large. We really don't know but that's clearly something that we think about. Now this was again, I think that's a liability side argument I find very convincing. Let me go through the others. So the second which I heard at the Fed is and suppose that there is a sudden stop something that we've learned can happen to some financial institutions even in advanced economies and so the central bank wants to provide large amounts of liquidity to a financial institution or to a set of financial institutions and we've learned that that can happen. If it doesn't want to increase the money supply too much it basically has to sell an equivalent amount of assets. Now the conclusion of this is that you should always have on your balance sheet enough assets to sell in case you actually need to provide liquidity and this gives you a lower bound on how much what the size of your balance sheet should be. It's the same thing with effects intervention. Now is this convincing? I find it not convincing. I see that Benoit noted the same way which is that that's not an issue. I mean basically if you want to pump back liquidity you can just issue central bank bonds when you need it and maybe you want to have them all the time so that the market exists but it doesn't seem to be a big deal. You inject liquidity at one end you take liquidity at the other end it goes where it should go and you'll find so you don't need in steady state to have a large balance sheet of I think it's useful to have the ability to issue the central bank bonds but that's a very separate issue. It's more an option that you have rather than something on your balance sheet. The third argument I've heard and this is in the context of Japan which is suppose we conclude that it's only the long rate which affects aggregate demand and I think we've become wiser on this. I mean all kinds of rates affect all kinds of things on the aggregate demand side but we have a sense that many decisions should depend on something like the long rate and then you know so the central bank decides to basically set the policy rate at the short end and the relevant spread on the say 10 year of a three month maturity bonds and intervenes in the market to do so. It needs to do both I mean it cannot if you just go for fixing the 10 year rate without the policy rate being determined you get an indeterminacy which get you into trouble but you can do both you just set the policy rate of the corridor on the policy rate and then you intervene to create the relevant spread and you achieve exactly the 10 year rate that you want. Now if you're going to do this this is a market with extremely elastic demand and supply so you'll have to have a large amount to sell and a large amount to buy depending on which way of in which way you are you want to move the spread and that requires potentially a very large balance sheet if you want to sell it. Now do I find this convincing? No I don't find this convincing again and again there's a whole question that Ricardo Reis is working on in the case of Japan which is I mean can you actually do it? And I think the answer is no you know there's a limit to how much you can affect the spread. I mean the analogy is you know can you both fix the policy rate and the exchange rate? And I think most of us would say no you can't you probably can affect both but the notion that you can fix both we think in the kind of markets that you know these assets are traded it's just not possible. You can probably decrease the spread to zero or maybe even a bit negative as you suggested but surely not much more than that and it would need an enormous amount of intervention and it seems to me that although we have learned that the yield curve doesn't quite move like we thought in the old days the expectations hypothesis is clearly not completely right but I think that by moving the policy rate you can more or less achieve a 10 year rate that you want I mean you'll have to take into account what the spread is but I think you can do a very decent job. So again I just don't find this convincing in any way. This is argument four and five actually because I've put them on the same slide in order to save space and time. So the more general issue is are we happy with the term premium structure of the spreads on different assets whether it's by term or by type of assets. And I've heard two arguments. The first one is the term premium actually is too high and if you think about how you can justify it based on risk you conclude that it's basically higher than your measures of risk that you have would justify and indeed we see many financial players playing carry trade which suggests that there's money to be made that way they don't do enough to reduce the spread but so I think if you can convince yourself that there are distortions somewhere in the term structure then you probably want to use debt management or QE and I'll come back to which one in order to change that. The other argument which strikes me as very relevant is again the probability of hitting the zero lower bound again which is that if you lower the spread on long bounds then other things equal this means you can achieve a higher policy rate than the same 10 year rate and therefore you have a bit more margin to play with a policy rate when times get bad so you can decrease the probability of a ZLB. Now we've talked about many ways of decreasing the probability of hitting the ZLB from a high inflation target to nominal income targeting to other schemes. This one strikes me as a rather costly way of doing it. If you think that there is no distortion to start then creating a permanent distortion by reducing the term premium in order to basically have a bit more flexibility when and if my 4% inflation target surely has distortions but I would trade those for the inflation target distortions. Anyway it seems to me that's an interesting debate to have but in the end I see absolutely no reason where central banks should be in this business because here you're talking about permanent changes in the term premium and it seems to me that that's exactly what that management is about. So evasive case for issuing more long bounds or taken collectively in the hands of the public I see absolutely no reason why this should not be done by treasury and I see no reason why the central bank should basically be taking on some of the operations and some of the risk. So I think arguments four and five where the two lines that there's a distortion you want to correct or that you want to decrease the probability of a ZLB. So conclusions on scope and again that's not because Jeremy's on my left. It seems to me that's the only one which in the end convinces me that I might be. But again, I mean the size of what you need to do in order to reduce something from 60 basis points to five at the very short end of a curve. Not sure that the central bank should really be in that business. Why, you know the political economy of mismatched central bank balance sheets is a big issue. On the other arguments, it's a footnote but I mean I've talked about coordination of that management between the central bank and the treasury. Now it's clearly easier to argue on the other side of the Atlantic. I mean here, you know there's a whole set of issues which is well you have 19 treasuries which are not coordinating between themselves anyway and the whole but they coordinate between themselves and with the central bank on the debt in the hands of the public I think is a hope. And so I think that's the issue then of division of labor and whether the ECB should do more. So I think that's specific to the ECB. On speed, you know I think you on has said extremely interesting things. Clearly the markets don't seem terribly worried. Let me kind of again go back to the basic economics which is you can think of increasing the policy rate or increasing the spread. These are basically the two margins and the question is which one should you use in which order. My sense is that we have a much better understanding of the effects of moving the policy rate because we've done this for decades than of increasing the spread. We have some understanding of what happens in the second case but it's not as good. So it seems to me that given that we're still on this side of the ocean anyway, not on the other side, in a situation in which there's a large output gap, there's very high unemployment and some of it is clearly higher than the natural rate, then I would use the instrument which I know works. I think that just follows from first principles. Would I do some of the reduction in some parts of the balance sheet? Yeah, I would basically look at where QE has created dangers. I mean, do other insurance companies in trouble and so on and so on. And then if there was a need to do something to help them or by going the opposite direction of what you were talking about, then I think there would be an argument for doing this. But for the moment, I would undo the order in which we went in. I would basically first use the policy rate and then they slowly increase the spread by decreasing the balance sheet if the balance sheet has to be decreased which I tend to believe. Let me stop here. Thank you very much. Before I open the floor to the audience, maybe I could give a chance to each of you to comment on what the others have said. I have a question, so I will abuse of my position but just once, so that's a commitment. Just to kickstart the discussion because Olivier discussed at length the division of labor between the central bank and the treasury. But as we learned from Jeremy, initially that's a three-player game. You have the central bank, you have the treasury and you have the private sector, commercial banking, deposit-taking banks, let's call them that way. And Olivier has given very good arguments why the treasury maybe has a comparative advantage in providing the kind of short-term assets that Jeremy asked for. But we haven't discussed that much as a choice between having it done by the public sector or by the private sector. And I feel, after your presentation, Jeremy, I feel a little bit uncomfortable. Also in the light and with the looking back at all the regulatory work that has been done since, say, the Pittsburgh Summit, so that was eight years ago. And I have a little bit of the impression that we're going in circles. So the starting point was, oh, well, these banks haven't managed our risk properly, they've been too big to fail, they've been bailed out, that's bad, which I agree. And so let's have them stand on their own feet and build the right buffers, the capital buffers, liquidity buffers. And that's most of what we've been doing for the last eight years. That's why we have 1,000 colleagues in another tower in Frankfurt and these kind of things. But now, eight years later, you're telling us, well, finally maybe the central bank would be more better equipped to do it. So does it imply that everything we've done has failed in terms of the regulatory efforts on the liquidity side? Or what's the key argument to have this being provided by this public good being provided by the public sector rather than the private sector? So that's a great question. So actually, we have a proposition somewhere in a paper which is exactly your proposition, which your intuition is exactly right, which is if you had a regulatory instrument that was a frictionless regulatory, so imagine you could impose a Pagoovian tax on banks for doing excessive liquidity transformation. If you could impose that tax without slippage, that's the first best instrument and you're done and you wouldn't do any of this sort of distortionary stuff with central bank balance sheets or treasure, okay. In a second best world, where the regulation is imperfect, so it creates some deadweight cost or there's some arbitrage, that gives you some reason to not wanna not do the regulation, you still do the regulation, but basically you can't get all the way there and so then you would do a little bit of both. So the premise implicitly, it's not that the regulation was a bad idea or was a mistake, but it's a recognition that it's gonna be not perfectly effective, especially if the economic incentive is too strong. I think that's the story. No, I withdraw from that discussion and if you want to add anything. I thought it may be just worthwhile bringing the international dimension into this discussion. I think Jeremy mentioned the role of foreign banks in the US money market. I think that's been an interesting observation. I think it also speaks to the question that Benwas said to me, which is why our long rates are so resilient because I think there may be some international dimension to that question as well, in that the long dated bonds across jurisdictions may be substitutes provided that you can hedge the currency risk and so the link between the FX market and the fixed income market really looms into view. And if there are major central banks and other jurisdictions which are engaging as a purchase programs, it may be that even if you try to tiptoe towards the exit, it may only have a limited impact if the demand from other investors who can hold your bonds on a hedge basis still keep rates quite low. It's worth thinking back to what happened in 1994 because 1994 is a good case study in that it sets a very good benchmark. If you remember, 1994 was the year of the bond market crash in the US when long rates really rose very, very sharply. And that coincided with an upturn in the US economy. In Europe, it was in a slightly different cyclical position. It was actually, if anything, needing further stimulus rather than raining in. But what we saw was that long rates really moved in unison and long rates actually rose in Europe as they did in the US. And what we may be witnessing is that the spillover now is much more evenly balanced and that the fixed income investors from the Euro area and from Japan are having quite a big impact in the dollar-denominated market. And if we combine this with the evidence of what's happening with the cross-currency basis, the deviation from covered interest parity, I think the story seems to fit in that we do see the dollar borrowing costs implicit in the FX swap to be much larger than the equivalent market dollar interest rate. And that's reflecting, I think, this underlying demand for dollars that come from the long-term investors who want to invest in dollar-denominated fixed income assets on a hedge basis. So I think if we open it up and take a global picture, I think many of the points that we've been debating, I think, perhaps make, we can shed more light on those questions. Okay, let's open the floor. Four questions or comments?