 Let's start back. So welcome to this keynote lecture. We are in a very, very good company of Professor Ricardo Riz. I don't think Ricardo needs any introduction here. We all know he has made very, very important contributions in macroeconomics, finance, and in particular, on a topic which is of special relevance in this institution and in all central banks, which is the analysis and the measurement of inflation expectations based on the financial assets. So today Ricardo will present us his most recent piece of research precisely on this very critical topic of how to measure and how to interpret existing measures of inflation expectations based on inflation link drops. So Ricardo, please, the floor is yours. Thanks again for accepting our invitation. Very good. Thank you very much. It's a pleasure to be here, and thank you for having me. So yes, I was going to use this opportunity to present to you my most recent piece of research, of which there's a draft online, although it will be an even better one, I think, in a couple of days as part of preparing for this keynote. And that is a work that tries to explain, as the word says, the market for inflation risk. What I mean by this is trying to understand what's behind the inflation swap market. So for some of you, this will be 40 minutes of everything you wanted and didn't want to know about inflation swaps. For those who have used them, and you know, Isabelle showed several pictures, I thought, more than one using inflation swaps, it's going to be how to understand them better, how to refine them, and how to take better signals for monetary policy and not only from this market. The starting point is, and this I took even from the, this was in the FT three days ago, and so I thought, let me just put it into the slides, but that was even a version of it in Isabelle's talk, is the fact that we often measure inflation rotations using financial markets, especially as we look along our horizons, it is very hard to extract signal from surveys at horizons above three, five years. I've tried, I've written papers about it, but it's just much harder to extract signal. Financial markets, on the other hand, know the difference between 10, five, and two, or you can explain to them in a way that in survey is hard, and you can do this thing of looking at things like a five-year, five-year horizon, which becomes extremely relevant to understand when inflation rotations are anchored. And if you look in the Euro area over the last 10 years, what you see is that this number initially, when quite significantly down around 1014, for those of you who have some memory of that time, this was exactly when then Mario Draghi, then ghost Jackson Hole, shows this picture and says, we need to start doing QE because this measure was falling down. It's the first time the five-year, five-year actually becomes somewhat famous even among the financial press. And he does QE and that works a little bit and then not so much lighter in terms of moving that back up to 2%. But now over the last two, three years, this has gone up a lot. And at first we thought that's great because we had been trying for this to be around the 2% of the target. But now should we start getting worried that it keeps on rising and rising and rising to the point where it's around 2.6, 2.... I think the last reading was 2.63, if I remember, but I may already be a little out of date. So I want us to think about this measure and what is it due to it. So what I did with my co-authors is I'm gonna show you, I think we made progress in three ways or we're gonna do three things through these 40 minutes. The first is we got great data. And what's that great data? Since 2010, starting with the crisis, every time after the very financial crisis, every time you buy and sell a swap, you now have to tell a repository so that we could track because these are over-the-counter markets with counterparts we could track. So we have the whole universe of this. I first learned about this data a few years ago when a team here at DCB, I remember Sam Langfeld, I remember hearing it from him, was using the interest rate swap. These are truly billions and billions of prices because every single swap they get sold and many of them get sold, gets reported. And not that many people have used this data yet. I think partly also as we start presenting this, now I get many, many emails of people saying, where can I use it? And you can. But essentially every swap counter with its interest rate, exchange rate or inflation, we have the entire universe of the microdata thanks to the regulatory features. I'm gonna use them for the UK because I'm at the Bank of England and we have that. Or I'm nearer to the Bank of England, sorry, and we have that. But similar work to what I'm gonna show you could be done for DCB, I think. And now what I'm gonna show you just from looking at the data in a fairly, again, I don't wanna say it's simple way, this first contribution took us like nine months because you really have to clean it. We're talking about billions of data. It's to show you that this market has a feature that really blew me away, which is how incredibly segmented it is. It is very segmented across many theories. The players in the 10 year and in the two year are just very different players. Which if you're coming at it from the finance theory, which I come a little bit on, was pretty exciting because we've been talking about segmented markets, preferred habitats in the context of the interest rate market. This is a preferred habitat. There's just two habitats in this world, very striking way. So now if I ever keep preferred habitat theory, this is the market I'm gonna show it, not so much the interest rate yield. Then after showing you some facts, I'm gonna show you in the lines of again, a recent strand of work in finance that thinks about how do we estimate demand and supply functions. Let's move away from efficient financial markets where demand curves are all horizontal and therefore no arbitrage conditions price everything into where quantities matter and who's buying and selling matters. And I'm gonna show you how we're gonna propose not one, not two, but three identification strategies to identify demand and supply in this market. From a finance perspective, I care because we're trying to understand financial markets and understand demand and supplies about understanding market power in these markets with the price we reveal. From a macro perspective, it's gonna mean that I'm gonna be able to get rid of those pesky liquidity premium. What is liquidity premium precisely the fact that prices are not just about expectations or risk adjustments, they're also about non-horizontal demand and supply functions, meaning that therefore shifts in those demand and supply features having to do with how much is supplied at these contracts and distorting your measures of expected inflation, like the ones I was showing you earlier. And then finally, third, I'm gonna apply this to three ways. I'm gonna have three applications. First, what are the slopes of the supply and demand? If you're coming from a finance angle, I think that's pretty important to understand how financial markets work. Second, now more from the macro and link to Professor Schnabel's initial speech, how reliable are these measures of expected inflation? And third, if you're coming more from a behavioral angle, how much of the dispersion beliefs there is, who drives these markets? We're gonna have, there's different players with different expectations. So when we talk about expected inflation, who's expectation are we talking about? And I'm gonna be able to show you, I think some interesting things about that. So with that, let's talk about the facts on this market. I told you, I very told, well, let me just skip the data. It's all the derivative transactions. It's billions of observations. I'm gonna be focusing on daily observations from 2019 onwards until February. There's some issues that are going earlier, some of which are just lack of energy and money and research assistance to clean the stuff before, but one can go earlier than that. But facts, first fact, it's striking fact. If you look, sorry, this market, I guess I should say the basics. I should really go with the basics. There's dealer banks. You can approach them and they will sell to an inflation swap contract. An inflation swap contract is a contract that says, look, a year from now, if inflation is up high, you get paid. If inflation is lower, you get paid less. The return is indexed to whatever happens to be the realization of inflation up until then, okay? So over the counter market, and very important, there's a lot of trade in between the different counterparties. There's a clearing mechanism. However, that clearing mechanism is all within the dealer banks. 17 of them in the UK. So you don't want to do that because then all you'd have is whether Paribas trading with UBS or something. What you really want is the non-centrally clear part. That's the hard part. That's what requires this data to understand when just a company, a pension fund went and bought an inflation swap from the banks. A lot of the key here is the non-centrally cleared bit. It's exactly the one that you really want to care about. There's millions of transactions in between banks. First fact, dealer banks, if you look at their net notional. So this is all the transactions they've sold against inflation. Mine is all the one they bought, okay? And this again, within the banking sector, so not an individual bank. And that's the blue line there. It's not zero. It's 120, 140 billion throughout the spirit of 2020. Why is zero the null hypothesis? They're dealer banks. If you're coming in from a model of a dealer, they buy, they sell, they hold nothing. And yet that is not what's going on in this market. Moreover, it's not only not zero, it's well above their holdings of index bonds. If you're a regulator of these banks, you may get a little worried because they're not supposed to be holding inflation risk. They were supposed to just deal the risk, match buyers and sellers, and when they have to warehouse some risk, they're supposed to be holding inflation index bonds that exactly heads them against them. That is not the case. These banks are not behaving as dealers. They're behaving as insurers. They're selling inflation insurance. It's probably related to the fact that there are nice fees to sell insurance. Insurance is a good business. Until of course inflation comes and you have to pay it out. So much so that by our calculation over the last two years, simply on this account, these banks have lost something like 20 billion pounds in payments as inflation went up. So again, not whether the dealer was supposed to do. So think about them as insurers. Fact number two, if you look at the long horizons for 10 years or more to be stark, the net buyer are essentially pension funds. This not surprising at all. Pension funds need to hedge their liabilities. They have very long-term liabilities. They want to buy inflation protection against some of them. More interesting is that the pension funds are essentially absent from the three year or less market. Inflation swaps on inflation of three years or less. Fact number three, if you look instead at hedge funds, I'm using here hedge funds as all, well, actually hedge funds and financial investment corporations. There, they are much smaller in terms of their net positions. For the pension funds, we're talking about net positions of about 100 billion. Here we're talking about look at the scale. We're talking about five or 10. But they are very active at the short horizon and note they went from selling to buying inflation protection, just as inflation took off. Justifying why sometimes you think about these as the informed traders, if you want, okay? But they're almost not at all in the 10 year market. And we can talk at length about why that is, having to do with how you clear and make bets on the 10 year market and collateral carbons and others, but essentially completely absent from the 10 year market. Indeed, if you look at trading activity as opposed to stocks, is there literally day by day how much was purchased and sold or some of the purchase and sold? You see that in the long horizon, it's all blue. It's all pension funds with some spikes. You know, the pension funds, something interesting goes on in the 10 year market. Something interesting happened about here, for instance, with this LDI crisis in the UK backfall and the pension funds do come in and buy itself. But usually they don't. Whereas in the short horizon, again it is the hedge funds who seem to dominate in a very clear way. So it's an extremely cemented market with to complete the picture, the dealer banks being on the other side. So you have dealer banks passively if you want selling in true suspension funds in the long market. And in the short market, it's a dealer banks and the hedge funds who the hedge funds sometimes are buying, sometimes they're selling protection depending on where you are. An extremely segmented market of this. Okay? By the way, this is data for the Eurozone. My data is very good until a vote happened that meant that the UK is no longer, we no longer get the data bank being on but the ECB would have it, but that's why you have this break, meaning you start, it's not just about data collection. I misspoke there. The issue is that with Brexit, London is the key market for selling and buying these swaps. And so as a result of data bank being very good, that's just where the market is. Some of the market has now moved to Frankfurt and there's some inflation softening trade there. So before here, I think I had the universe. Now I have a sample of the universe. It's another reason why I wanted to focus on the UK. But at least for this, you have very similar in the, in the your area in terms of the dealer banks, the pension funds and the segmentation that we seem to observe as far as we can see. So that was part number one. First thing to remember is of extremely segmented market. Number two, let's think about then this market segmented market. Think about the long horizon market where we have the dealer banks applying and the pension funds demanding. Why do we have a balance here that's positive? Retreat, note, this is not a fundamental asset. There has to be a buyer and a seller. We're swapping things. Well, it could be that maybe we think pension funds expect higher inflation than dealers and that's why they will buy and the deals will be happy to sell them thinking, ah, I'm gonna make money out of you because you think efficient is higher or not. Or more plausibly, there may just be different risk aversions. The banks may be more, the pension funds are more risk aversed to inflation because they have more of their capital tied to it. They have more of their bank run risk tied to it. They're more tightly regulated in some way than the dealers in terms of the risk they can carry. And as a result, you'll have some positive amounts. What happens when there's an increase in expected inflation? The things that the monetary policy makers care about. Well, if there's an increase in expected inflation, it may be that the pension funds update more and the dealer banks update less or the other way around. There's disagreement, heterogeneity. Either way, demand and supply will both shift up and so the price will shift up. And so Isabelle in writing her speech will see that we move vertically here because expected inflation went up. It's gonna be a mix of how much was it the dealers versus the pension funds expected inflation went up. As you'd imagine, right? Again, only in a purely efficient market with less expectations would they move vertically by exactly the same amount and there's a single increase in expected inflation. But beyond that, there's also non-fundamental reasons, not have to do with expected inflation that affects the demand and the supply by the banks. For instance, the pension funds may have tighter regulations sometimes, may have to manage some withdrawal, some cancellations of life insurance contracts, they may have trading constraints in terms of what they can buy and changing those are gonna be demand shocks. And those are gonna be pretty bad news for the monetary policy makers because you're gonna see an increase in expect in the swap price, you're gonna call increase in expected inflation but it really isn't, it was just increased demand for inflation protection for whatever reason from the pension funds. And likewise, you're gonna have supply shocks because the banks also have trading constraints, operational reasons whereby they may be able to sell or buy, they'll be able to supply more or less its insurance, especially in the sense that they are insurance sellers much more than they are dealer. Things get more complicated or more interesting, actually easier, not more complicated because we have two markets. I told you about the long market. I could tell the same thing about the short market but in the short market, what do we have is that we have a completely different set of demanders, the hedge funds with the eridicine credit shocks, but I have the dealers, the banks for short, that are common in the supply operating in both markets. They're the segmented markets. What that means is that a liquidity shock to the dealer banks is the shift in the supply in both markets or is the liquidity shock to the pension funds is a shock only in one market and likewise for the hedge funds. And I'm already trying to give you very strong hints of why it will be able to identify this. If you have players in all markets, then you're hopeless because all shocks are shifting everything at all the places. But with the segmentation, you can make some progress. And to make some progress, let me be a little more formal. Imagine that we have linear demand curves by pension funds, of which there's I of them, dealer banks, this is the short market, sorry, the long market in small letters and in capitals, the short market. And those demand functions are such that they fall with the price of the swap. They increase with expected inflation. And I try to be very clear this is expected inflation together with a risk adjustment. That may be a risk premium. This is risk adjusted expected inflation. And then there's gonna be this liquidity shock so I'm gonna call Lambda. Note that long and short inflation may be obviously correlated. One three year ahead inflation, 10 year inflation are correlated. But for these shocks to even make sense, they are uncorrelated with the liquidity shocks. That's why they are orthogonal. Otherwise, what does he mean even talk about as a shock? But absolutely the banks and importantly the banks a shock to the banks affects them both in the long and the short market. Okay. Now equilibrium this market from market clearing implies that the price we observe is risk adjusted expected inflation. Again, Isabel would like to know PIE. She has to struggle with the fact that it's risk adjusted. It comes with a miscompensation. And as I told you in a world in which we think that there's no rest expectations, this is also PIE that's really a weighted average of the pension funds and the dealer banks as well as their differential reactiveness to the price and then for the slope that's the plan demands. But as bad as the fact that we have all this liquidity premium coming on. Okay. I'm gonna have shocks to frictionless price. I'm gonna keep on calling this the fundamental. I'm gonna have shocks to the liquidity. And so what I'm gonna try to do and I think I'm gonna do an okay job today is really try to get rid of liquidity premium. And then towards the end, I'll do a little bit about the risk premium. But let me be clear about that. With these equilibrium prices, I'd like to separate those. As I said, Isabel has a problem that if she only observes the prices as she did for her talk, she cannot tell us if it's been a liquidity premium. She can't identify the liquidity premium separately from where there was the frictionless price. There are shocks, both of the fundamentals as well as to the liquidity premium of the hedge funds, the pension funds and the banks. Now I can already do much better. Why can you do better? Because she had two prices and I have to price in two quantities. Since I have four shocks, I'm obviously gonna be able to do better in some ways. Okay. And to remind you, these are the price that I have in quantities of the net purchases of these across different sectors. Still, at least I'm in business. I have four shocks and four observables. I still, familiar to this macro audience, still have to identify some matrix. You could put this of course in a VAR, but let me talk about it in static terms. That links the shocks to the observables that I have. We're gonna do this always with VAR. And here are my three identification restrictions. First, I'm gonna exploit the fact that my data is very high frequency. In particular, it's high frequency over three years and three very interesting years, very painful years for the staff here in the building. It was worked very hard in the last three years. But for me, very interesting years because there's been a lot of inflation shocks. Why is that good? Because during those years, every CPI release, by the way, these are RPI because it's the UK, but we would never hear me say RPI, I think CPI. Every first release every month came with a lot of news. A lot of news. But then, approximately, we can think of every time there's an RPI release and I have 48 of them, a lot of fundamental news is released about inflation. Much more than in the 30 days they go until the next price index gets released. But then I can exploit identification by interest elasticity. I know that, and I observed it very strongly that in whatever the RPI release, there's massive movements in these markets, there's a lot of trading, prices move around a lot, and those are the days when fundamental news are being released. So all I need to know is that as long as there's more fundamental news there than in the other one, and you can test these first stages and they're extremely strongly supported, I can do heterosascist identification that says that when I observe the price of the quantities, I'm gonna be able from that various covariance matrix try to defy some of the shocks. That seemed like a good idea at four years of interesting data. I had a time series. So then we had a second idea. This is also a market that's extremely granular. What do I mean by that? There are 18 pension funds here, 18 dealer banks approximately. There's about 40 hedge funds. There's a lot more pension funds, about 300, 200, 300. But when you look at the data, what you see is that their size distribution is very much periothic-tailed. What that means is that there's gonna be, whenever a pension fund, a particular pension fund, one of the big ones, gets a new de-syncratic shock, that's gonna move the market. The law of large numbers doesn't apply in these markets because they're very concentrated. What that means is that I can use this granular IV recent strategy of saying, look, given fundamental, let me extract all common factors from the demand. Those residuals give me the pension fund, life insurance, dealer bank specific change in quantity demanded, but once I have that, while those will average in zero to zero by partisan linear regression, once I weigh them by assets, because the assets have such a strong periothal, this GIV, the weighted sum of the residuals, these mean zero residuals, once weighted by assets, are not gonna be mean zero. By construction, they are orthogonal to what I wanted. The hard thing is for them to be relevant, but as long as the asset distribution is peredo, they will be relevant, and indeed, they are extremely peredo. You have periprams of 0.13, powerlock coefficients of almost one, there's almost Zipp's law when you look at the distribution of quantities held in this market. And so I can build these instruments, and with these instruments, I have very powerful instrument in order to identify my three shocks, my three liquidity shocks, since I have four shocks, then the fundamental falls as the fourth one in the system. Okay, but I had even other ideas, because not only I've already exploited the fact that I have a time series, I've exploited the fact that I have a cross-section across institutions that have to be very peredo, not wanna exploit the fact that I have very high frequency by making two assumptions. At a high frequency, I would expect hedge funds to be very smart and or jittery to news, and pension funds to be much less smart and or jittery to news than dealer banks. Again, within the day only, within a week, within three days, they may adjust differently. But whenever there's really a fundamental news about inflation, I expect the hedge funds to react or maybe even overreact within that day, and the dealer banks to, sorry, and the hedge funds to be slow to move with the dealer banks in between, okay? Moreover, I will, from just how this market works, you will have that in the dealer bank, they'll be the guys selling the short-term market and they'll be the guys selling the long-term market. And they're actually different guys, why, because remember, it's an over-the-counter market and the person who calls you to buy the contract is a different one in the long and in the short market. It's the pension fund, this is the hedge fund. So they actually have two people almost always doing this, more than two. Well, those guys have to adjust their book to risk exposure, but they usually do it at the end of the day. What that means is that if they have a liquidity shock to in the short market, they only adjust their exposure the next day. So within the day, they're gonna have their desk separated. In other words, if there's a lot of demand for short hedges, within that day, I can say, oh, I don't have capacity to supply to you, but that will only affect whether I also lower my capacity supply in the long market tomorrow. And because I have high frequency, if I did this assumption at a monthly level, this would be stupid, but at a daily frequency, it actually makes a lot of sense. I mean, again, you talk to these dealers, what they do is they meet up at the end of the day and say, this is your capacity for tomorrow. And then they meet up every 24 hours across the segmented market. Well, with those two assumptions, I have a beautiful sign restriction, VAR across the things or our regression. Why? Because essentially, well, I don't have time to go over the shocks, but once you assume these things, you know already that what this, the first assumption gives me is that whenever I have a fundamental shock, quantities are gonna increase in the long market, but fall in the short market. Why? Because the demand responds more than supply within the day, okay? And the second one is gonna allow me to say that whenever I have a shock, a liquidity shock in the short market, it doesn't spill over to the long market vice versa within the day. And with that, I can now identify the system. Okay, so I have three identification tests, okay? I'm excited, I don't know. I have three. Why? Because, and note, they're not just three, you know, you like one more or the other. One is exploiting the cross-sectional distribution of assets. One is exploiting the month-to-month variation. And one is exploiting the high-frequency nature of how the trading happens in this. So they're really fundamental exploiting the three dimensions of my panel. Okay? Why is that good? Because they're gonna essentially do different layers of the data I'm gonna explain. I can see, do they give me the same answer? And here's the thing that's really striking. Here's the correlation between my fundamental shock across the three strategies. It's really, really high. Moreover, I can do something that I think is somewhat innovative in econometric terms, which is, well, I have one identification strategy, or I have two identification strategies. I'm gonna use the estimates from those to check, to test for the identification strategy of the third one. Namely, give you an example. I can look at the identification strategy from the granular IV or the atrocious toxicity and say, do I observe that indeed the hedge funds respond more than the dealer banks to whenever shock arrives, to the shock identify that way? Or I can look at the ones from the timing restriction and say, do I observe that once I look at the identified shocks from there, do I observe that indeed it turns out that you have the orthogonality restrictions of the GIV, I can test them? Or I can, to talk about the last one, look at the GIV identified shocks, do they have a much higher variance during the RPI release dates? Nothing identification of that imposes it, and yet it is definitely the case that the variance doubles in those dates. So I can really verify all of them very well. All the results I'm gonna show you now, just because otherwise the plots kit is gonna be using the strategy three, the time restrictions, but everything's very robust. Again, I'm gonna find the same shock, so I'm super confident that I have the same, that I've done this right. So now I can turn to the last part, and the third result. So results, financial, macro, and behavioral. So those of you who are finance people, pay attention now, then you fall asleep, macro comes later, then comes behavioral for the three audience if you want. Financial markets, here's just to start with just the VAR responses of quantities and prices to a fundamental shock identified by VAR mark, okay? When you get an increase in fundamental inflation, you get an increase in short-term and long-term inflation swaps, as they should. You get an increase in the short-term quantity and a fall in the long-term quantity. That's the differential reactiveness, right? Because again, because if the hedge funds respond more, you get that there's more demand for, for the pension funds respond more, there's more demand. The hedge funds respond more, there's more demand for protection, the quantity goes up, because the dealer banks respond more, there's less demand. Very striking fact from these pictures, which I wasn't expecting, after three or four days, actually, really after two days, everything is horizontal. And this is true across, these markets are really efficient. Now, I can reject inefficiency because the effects still lack for a day, but remember, in an efficient market, it should be strictly horizontal. All the information gets absorbed within the day. Here, it's true that tomorrow you can trade on what happened the day before, but two days later, you already can't trade. The prices have all gone as they should into horizontal lines. So, strikingly efficient market. No, this is weak for efficiency, not strong for efficiency, just that you don't have any serial correlation to price after shock after two days. But, here are my slopes of the demand functions. Having estimated the response of demand, quantities and prices, I can estimate demand functions. Interestingly, if I look in either the long or the short market, those slopes are really similar. Turns out that the hedge funds are a little more price sensitive than the pension funds. And so, it's a little excited. I was like, oh yeah, maybe they're more price sensitive. I could start theorizing about it, but really, it's not such as lucrative. What's very significant is the following. And that is that if you look at the dealer banks, in the short-term market, they have a nice upward sloping, or maybe not so nice upward, if you're a fine and efficient market guy, upward sloping supply, but in the long-term market, it's incredibly horizontal, okay? Why forget the econometrics? What are the data drives this? Long-term inflation swap prices, for those of you who had to stare at them, and who Isabelle has from her speech, don't move a lot. But quantities in those markets move a lot from day to day. How can a quantity move a lot when a price doesn't move a lot? The supply curve has to be horizontal. As simple as that. You look at the short-term market, quantities actually move less than the long-term market and prices move a lot more. That's basically what's driving this result. After you do all the fancy econometrics, that's basically the key thing that strikes you in the eye and why this is a very robust. What this tells me is that, of course, if I'm into the IO of financial markets, there's something about the fact that if you're starting, on the one hand, it must be that dealer banks either A, have more capacity to absorb long-run risk to short-run risk, and that's why they're more price insensitive there. They have a long-term supply curve, or my hunch, they have more market power in the long-term market, and remember monopolists have, indeed, supply curves are horizontal at the monopoly price. In dealing with the pension funds, the dealer banks are able much more to dictate it's a take it or leave it offer, whereas in the Louisville hedge funds, there's more of a negotiation leading to an upper-slope supply curve. What this implies as well, though, is the following, and now moving a little bit more from finance to macrofinance. When you do the various acquisitions and focus here on the prices, what you see is that the fundamental shock, or let me start the other way around, the liquidity shocks of the pension funds are gonna have a big role on the quantities, but no role, pretty much, on the prices. Why? Because shifts in the demand with a horizontal supply curve don't change the price. That means, therefore, that the long-term inflation swap price is gonna be driven by either fundamentals or liquidity shocks of the banks. And our identification says that it's roughly 80% fundamentals, 20%. Liquidity shocks. That says that, again, when Isabelle puts in her speech a 10-year swap, it's actually pretty reliable. It's 80% fundamental. However, when you look at the short-term market, it goes the other way around. Actually, fundamentals are incredibly small, 10%, 15% of the variation in that price. Most of when the short-run one, two-year inflation swap market prices move, it's almost always just a liquidity shock rather than the dealers or to the hedge funds. So very unreliable to look at it through your swaps if you're a macro economist trying to figure out what's going on through inflation. Important question, which we can't answer. I think we need to write a new paper on it. What about the five-year, five-year? Because now you're having the 10-year minus the five. And you have a lot of noise in the five? Well, again, in my paper here, when I say short, here it's a three, two or three, and the long is the 10, because we want it to be very sharp because there's a lot of it in between. We're writing a new paper with it in between so we can then tell you about the five-year five. That's the fine insight, macro. Here is, if you want, I think the picture that President Schnabel and others would like is, okay, I can give you, the red one is the one that she showed in her speech. The blue one is the one after all this work that I can extract the liquidity shocks. You can, again, see a version here of the 80% of the variance. It's not radically different. Indeed, the actual price of it. But it does change somewhat. And what is the main way in which it changes? Which is interesting and surprising. If you regret, let me just, let's do the simple back of the envelope if you don't want to do all the work. Regress, blue on red. Cardo, I can't do all of this machinery. But whenever I see an actual, I'm gonna multiply by the coefficient of that regression to adjust what it is. That coefficient is 0.8. In other words, if you see, like we started the stock 35 minutes ago, the expected inflation in the last three months, according to the swaps going up by another 40 basis points, really it's 30 basis points on average. It overreacts. That 10-year rate is overreactive, okay? That's a simple one-liner in it. But let me look a little bit more carefully to show you why that is. And here, again, remember, this is UK. And so these are UK facts. What happened during COVID, 2020? Remember, when we got the beginning of the COVID recession, we had that. We went into lockdown, March 2020, I know, painful memories, apologies. And we had a big decline in expected inflation. Marcus thought we're gonna get into a big deflation trap, if you remember. Around the same time, there was the liquidity premium also fell. Okay, that is what you had was that the dealer banks selling protection against deflation, or the other way around, the demand for protection also declined. And as a result, the, together with the fundamental liquidity premium co-moved positively with it, and therefore the swap price exaggerated the fear of the inflation. When we had the invasion of the Ukraine, again, big increase in the fundamental shock because energy prices went up, inflation had to go up. What we saw at the same time again is dealers at this point withdrawing from supplying it into this market. And as a result, we had an increase in liquidity premium. And so that again, the jump in the inflation swap then overstated the inflation danger at that point. Okay, so that's the case by case version of that. The short-term price on the other hand, moves a lot. Now, again, I was reassured that when you look at the low frequencies, the short-term price isn't just silly. Remember, this blue essentially is 50% of the variance of the red is what the variance competition told you. And yet it kind of tracks the low frequency movements going back to the paper in the panel before, between the difference in the low frequency and the high frequency. But it really overreacts quite a lot. The overreaction here is much, much bigger. It's not just 20%. How does this liquidity premium compare with bid-ask spreads? It correlates with them. Not perfectly, but correlates with them, providing some external validity that what we're measuring seems to be capturing something interesting and important. Let me turn now finally to my third application, the behavioral application. So far, I've really been focusing on the sectors and exploiting the micro, when I was doing the granular IV in terms of heterogeneity across. But now let me explore a lot more of the heterogeneity within the sector. Here are, I can't show you their names, here's eight of the major hedge funds that trade in the inflation swap market in the UK. Remember, I already told you these guys are very jittery. They changed position from day to day, either very informed again or very overreacting. But a very interesting thing is that when you look at them individually, is how they often trade on different sides. One guy's buying and another guy's selling at the same date. It's not that they're all, I showed you already, they all went from on average and an aggregate, selling inflation protection to buying inflation protection. But actually, but now some guys are selling, some guys are buying, some guys are taking big positions, some guys are taking small positions. In other words, and from the perspective of expected inflation, there's a lot of team that's very dear to my heart, a lot of disagreement out there, even among the sophisticated investors. They just have very different views of whether they should be going long or short in this market. Here's what we can do. Since I've identified a fundamental shock, I can first, and since I have the individual quantities held, I can regress individual quantities now or project on my shocks. And in doing so, get a price or a quantity impact. Remember I had those, the beliefs get weighted by your assets times your responsiveness was mui? Well I can recover that, right? Because I can see how much is a quantity response to a shock and get the jitteriness of the different agents, which then would tell me a little bit about their price impact in terms of who's shifting the prices. And so I'm gonna show you that and I'll come back to this. And here is the nifty fact, and here we even threw in all the identification assumptions, okay? And the nifty fact is that if you look, this is for the short term market, it's a little more striking, is that it turns out that there's four dealer banks that really have a huge impact in this market. There's a few hedge funds, a couple on the one side, a couple on the other side, that have huge impacts as well. So we're talking about the inflation swap market. In the short market, we're really talking about often eight or nine major players four on each side if you want, okay? They have just a very large amount of price impact. One because they disagree a lot, and again that drives a little bit of why we identified 20% of the liquidity shocks. But two because they also again respond very strongly to what this identification is. I can't tell you who they are, but again, the Bank of England people now know that there's who are some of the key people that are driving. So when the inflation swap price goes up, and you're trying to make sense to immobilize your point eight or not, you cannot look at the trade behavior, you know which four guys are getting very jittery in some ways or not. And try to assess whether you think they're just crazy or whether you think they're onto something in some ways. So it's really quite striking. And again, another validation from the GIV is how you really have a few guys here and a few guys here, two guys there. In the long market we have 200 pension funds. And I guess I didn't have the picture there but then I won't be able to come back. It's much less the case. It's much flatter. You don't have like three pension funds driving everything. It's much, much flatter, which contributes to the fact that you get a lot more fundamental in the long term swaps. But the next thing I can do with this is that, because I know who these people are, they have identities, I can look at a completely different source of data now which is that Bloomberg called these people and asked them what you think inflation is. Now it's not the same people who sell inflation swaps. It's usually the chief economist, the professional forecasters, but they call them what you think expected inflation is. And that is a measure in some ways of what their fundamental subjective, non-risk adjusted, not the risk neutral one, actual expected inflation is. So, so far every I told you about fundamental was risk adjusted. But now I can say to what extent that those movements that I showed you, are they movements in risk premium or in subject expectations? If I have a measure of expectations that correlates with my measures, then it's probably most of the expectations. If it didn't correlate, then maybe it would be risk premium. Okay. And so here it is. I'm gonna show you the same impact from regression again of the subject expectations on our fundamental. One source of data is Bloomberg. Another source of data is the guy buying and selling and taking positions. One is I just talked to this guy in Bloomberg. The other one is I put my serious money. It's my job on it. So for those of you who also kind of interested in the braille question of, should I just ignore surveys because I, you know, I say some stuff on the phone, but actually I trade something completely different. You would hope to see a very positive relation here. Right. For those, for those from a different angle as I put it earlier, don't worry about how much of movements in market prices are moving subjective expectations versus movements in risk premium. You hope that it's also a very positive slope so that it's most of the expectation instead of risk compensation. And lo and behold, it's a really nice positive slope. I mean, I only have 18 of these, but it turns out that you call a bank, pachy bar or something. And if they say I expect inflation, I go and look and these guys have taken positions in the long and short market that it's consistent with expecting a lot more inflation and therefore buying more inflation protection as opposed to selling it. That's a pretty striking indictment of why surveys are good in some ways, but also why swaps are about expecting inflation, not just compensation for risk. So to conclude, all I've done is first show you some facts that at short horizons we have hedge funds and they alternate with dealers on whose negative and positive and longer rises the dealer sell inflation protection made good fee money for 10 years, lost a bit of money in the last year and a half just like an insurance company does. So we should call them the insurance dealer banks not as more than the dealer banks. Second, I propose three separate identification strategies that exploit a completely different dimension of the data frequency, concentration and time series and they all gave remarkably consistent results. At the short horizon, the supply curve is steep, therefore liquidity shocks drive prices. At the long horizon, supply curve is flat, fundamentals end up accounting for 80% of the price variation. Our new measure is more anchored. So it's a little bit of reassuring message to the central bankers. When you see the long-range inflation salt moving around, it's overreacting. It's more anchored to this. So again, for the UK now, we have that expected inflation is closer to the percent target than if you just read the measure. And that, and finally, I show that the risk to expectations from the market positions match with the subject expectations inferred from the survey answers, either giving Valia to the surveys or Valia to the risk neutral as telling me something about expectations. Only four minutes up here. Thank you very much, Ricardo, for the very, very clear presentation as always and for adjusting so well to the time constraints. So it's time to collect the questions from the floor and also from the Webex, Wolfgang or Este, we can pick up a few of them together. So, Wolfgang, here in the first row. Yes, thanks, Ricardo. I mean, I saw in your presentations along these topics a couple of times and somebody whose job is to interpret these market measures using models and drawing policy conclusions. It's always super insightful, especially this combination with quantities and prices. And at that parallel, I think what we have done for a bit longer was nominal yields when we interpret QE. So that's very cool. By the way, good to see some of that chiming with what we did internally. So 2019, we looked at the EMEA data and one of the key results, what you show today, namely hedge funds being the ones active in the short term and pension funds in the long term. So I just checked, rechecked the same same result. So I'm a bit comfortable with that. Very good. And also, I think take away from our type of people here is especially short term. Don't look at every basis point and try to over interpret it. I mean, I think you made a strong point here. I have one concrete macro finance question. So taking what you show today, so we need to think about say the ILS 10 year, five years. Genuine expectations, inflation risk premium in the narrow sense, plus the supply demand stuff that you came up with today. So we have three components. If I then enter into macro sphere and compute rear rates based on these swaps, how should I do it? And how should I think about what matters then for the macro economy? Because if I just take a nominal say 10 year swap rate, subtract the ILS rate, I consider this a rear rate, but now you showed there are three components in my inflation, which was deflate. So which one matters then for essentially, eventually for the real economy because these things are traded. This question one, question two. I mean, of course, I'm either said it probably this morning, we look at a host of indicators, not only five year, five year surveys, what have you. What about tips or tips derived break evens? How do they speak to your to swap based analysis? These two questions. Thank you. I have a couple of more, but I don't want to steal that. Thank you. Thank you Wolfram. Please identify yourself for the records. Next is Orestes over there. Thank you. So yeah. So Orestes Tristani from the ECB. Ricardo, very interesting this segmentation, these evidence segmentation. So maybe we should no longer call these market expectations. We should now be specific about which segment of the market is forming this risk adjustment expectation. And related to this, I was wondering, so for this long end of the market, you emphasize this evidence of market power or these dealer banks. And so I was wondering whether you can also tell, okay, the liquidity shocks play a smaller role. But if I think of the liquidity shocks also related to some financial constraints affecting the financial market, whether it is right to think about these as constraints affecting more the dealer banks rather than the pension funds that the ones that demand compensation to start with. Thank you. Hi, I was wondering, ah, Gaitaro Gaballo from Ashos. So I was wondering how we can think about expected inflation versus inflation uncertainty. And whether you look at how in press conferences, we could have a measure of higher inflation uncertainty after some speeches. Thank you so much for those questions. Let me answer all of the four of them to come and get one from you. First, on the real, the liquidity premium part is really a part about the financial market plumbing, if you want. Now, of course, I can always tell a story in that that financial market plumbing correlates with something that I care about in the macro. But the first approximation, I would say, you wanna clean that out, therefore use my fundamental as opposed to the actual prices when you go and subtract from, say, an interest rate to come up with a real interest rate. So the first answer is, you wanna take out the liquidity premium. That's the part that you wanna take out, I think. Because it's really about the supply and demand here, not being horizontal as issued in the efficient market. Second though, what about between the subjective and the risk premium? Which note, I have made no progress up until the very end when I show there's a correlation between the subjective. So all I can say is that it's not all the risk premium and the fundamental, but beyond that, I can't separate them. That requires a model of risk compensation, of course. On that one, do you wanna separate it or not? I believe you do not want to, but it depends very much what the R-star is that you're trying to calculate it for. If we're trying to ask what is, for instance, the I-star that you wanna add to percent to your R-star from, what is the long run value of the nominal interest rates, say, as a guideline for policy data for what the neutral rate is and not, you want to include the risk, sorry, you want to include the risk premium, attention. Why? Because this is a shift in, precisely the compensation that people are asking in order to hold inflation risk. That's going to feed in in a model towards agents using that discounting it from the inflation risk premium and going behind it. Second, if you wanna use this because you're coming at it from the fiscal policy perspective and you're trying to figure out what is gonna be my expenditure with paying for the government debt, you absolutely want to include it because you're paying for this. This is now a cost, the fact that this inflation risk premium is up. However, from a third perspective, if you instead you are trying to hear, figure out whether the Phillips curve is vertical, for instance, or whether, again, we want to take that out to back out the anchoring of the subjective expectations, crucial for pricing, then you wanna take it out. So it depends very much of what is it that you're trying to do with your real interest rate? Are you trying to assess a guideline for monetary policy in terms of what the neutral rate is? Are you trying to figure out what is the right value for me to do public finance perspectives? Or are you trying to figure out what's the subjective inflation perspective of anchoring and pricing? So it's gonna depend, I think, on the question of what you wanna do about that. Second, on the break evens, yes, of course you've looked. The break evens move in the UK. So the UK luckily is a country where the break even is very liquid because it's a very liquid tips marketer. They call them linkers there. Tips in the US, linkers in the UK, index bonds everywhere else in the world. The guys who are trading in the swap are the same guys who are then going to, of course, take a position in nominal and real bonds. So what's moving in the inflation swap price reflects very closely to the break even from the other one. We know that there's a gap, there's a famous work, as you know, that looks at that gap. Can you make some money out of it? Maybe just a little bit or not. We have looked at whether the gap between them correlates all liquidity shocks. It does, as it should, not very strongly, but it does as it should. So basically, to a first approximation, everything I told you about the swaps applies to the break even from the linkers. And there's some interesting things from a research perspective on linking to liquidity and the bid asks across such markets. Third question was on, I can't read my handwriting. Oh, yes. The market power and financial conditions. Note that what I showed you a rest there was that because of the horizontal supply curve, liquidity shocks to the pension funds didn't matter for the price. They matter a lot for the quantity. They're just not much for the price. And then I also told you that the liquidity premium or the liquidity shocks are only 20% of the fundamental. That is, of course, a claim that the dealer banks have not had huge shocks between 2019 and 2023 to their financial conditions and the ability to supply here, right? That is not to say that if you had a large shock to the dealer banks, and indeed again, with more time, I would have shown you during this LDI crisis where the dealer banks had to unload a lot of the linkers that fed into their ability to supply on the inflation socks, you saw massive distortions if you want. In the inflation soft price being driven by this liquidity fact. Meaning a lot of movements in the actual soft price, they were not changing the expected inflation. So the 20% is not a claim that they can't actually, because it's horizontal supply, precisely financial concerns in the dealer banks are going to have a very big impact on this measure of expected inflation. It's just that not many of them happen in our sample. Fourth question on the disagreement or uncertainty. First, I have this other work that's on their agenda where I've used swaps as opposed to swaps to try and measure inflation disasters and tales, try to start some information on that. So a brief answer is, see my paper with Halon, and, sorry, Raveev and Hilscher, which is exactly what I'm trying to get answered. But you point it out to something I think very useful, which is the extent to which with our clean series, we can now revisit some of the events studies that we had done around conferences, press announcements, and others, and see now whether in these we get sharper results than we did before. In particular, having separate the liquidity premium from the fundamental, we can now say that when, again, Isabel made a speech three hours ago, and the inflation saw price imagine where it could have moved, how much of that was the liquidity premium versus the fundamental? That answers a pretty fundamental question, but I should do it the first round, a pretty important question for us as modelers, which is how much do speeches move perceptions about the economy and expectations, or how much do they just move constraints in financial markets in terms of these guys jumping on the news and hitting constraints in different ways and the liquidity in those markets? That is, we tend to think too much of speeches and announcements as affecting the macro fundamentals, but there's no reason why it can affect the point of rest it was making, or even Wolfgang, that they may be affecting precisely faster conditions in this market, and have an impact on the finance side of it. And at least now, I hope with our series, which will as usual kind of make available, and you can download, you can at least write that paper and see how much is one percent. We know that small prices move when speeches happen, how much is liquidity versus the other ones, and neither of them is noise. Both of them are interesting just in different ways. We have some minutes left for another round of questions. No? Okay, let's see. One over there and then it's available. Yes, please, go ahead. Great, thank you very much. I'm Katja Peneva from the Federal Reserve Board. On point five, Ricardo, I was curious. On market positions, you have very high frequency data. The Bloomberg interviews are once a month, once a quarter. So how do you match the two, was the microeconomies that they, you know, was the response based on the analysis from the night before or the last month? I was just curious there about the frequency. Thank you very much, Ricardo, with fascinating speech. So you talked about short-term, long-term, and you mentioned in, I think, one sentence that it's more complicated when we are looking at things like five-year, five-year. But of course, we are looking at those a lot. So can you elaborate a bit on how you see that? Thank you. Note that what I measured was I used the daily date on the trading to get a market impact reaction. Then I used a monthly regression on the Bloomberg, monthly, because that's the monthly data, to get the reaction or responsiveness of the Bloomberg guys. And then the plot I gave you was neither monthly nor daily, was then across 20 people. I think it was not even 20, I had like 16 that they lined up very carefully. That's what I showed you. So I did not show you that these guys' expectations could move, the traders and the dealers. What I showed you is that these guys respond and react within an institution with the same reactiveness across the traders and the dealers. That's what I was showing. Then the other two questions are, Isabel's question, I need a new paper. I mentioned briefly that one needs to write a little bit of a new paper because what happens here is that I exploited, maybe abused the fact that 10 year and more is only pension funds. Three year and less is only hedge funds. Three to five is mostly hedge funds but some pension funds. And starting around five, the pension funds start coming in. So there's a gray area, the three to 10, which in this paper I threw out the data. I just did less than three and more than 10. The three to 10 is very interesting. And again, we're writing a new paper but we'll take a little while, which is you just need a much smoother model of market segmentation. Essentially what we're doing is doing a Vianos Vila-type model where it's much smoother the segmentation across the whole horizons because then I can do three and five and seven which is what I need to do the calculation that you want. As of now, we were just very sharp. I threw away the three to nine basically because it was just so sharp and I could do it at the level I did now. The next step is of course, to do a much more continuous thing and we're gonna continue a little bit. As of now, the hints, as I suggested are a little bit hard to interpret because on the one hand, the 10 year is moving but the five year is overreacting but the 10 year overreacts but the five year overreacts more but the five year doesn't just overreact fundamentals, it also just moves because of liquidity. So depending on whether in a period you had a liquidity or a fundamental shock, both or not, the five year, five year could go up or go down. If all I had was fundamental shocks then I can tell you that the five year, five year is gonna underreact. There was only fundamental shock but I already told you there's a lot of liquidity shocks and so given that, it could actually be a little bit. To be clear though, I could do the 10 minus three, sorry, the three seven, sorry, not the five five with my beer and I'll do that, I'll send it y'all. We'll do that. That we can do and I can tell you what I can do is do the counterfactual of that and the adjusted one that I can do and we can see what it looked like but lessons are going forward, there'll be much harder and then the last question I can answer in 10 seconds, that's a great thing, you should write that paper or maybe I should write it but someone should write the paper. I think yeah, looking at QE and seeing the extent to which as you change the supply of bonds, I mean the only objection is that of course we would have been great if the central banks had bought some mix of index link bonds and non-linked and they had changed that mix as it went around then I'd have the variation to really test this in some ways. They didn't do that for the most part and so as a result you don't have as much variation as you otherwise that would be a really great paper but yes, absolutely. To what extent, and that goes back to Wolfgang's question, to what extent when the central bank is intervening in the bonds market, that changes the capacity to take on risk when it comes to not just interest rate risk but also inflation risk, that changes their ability to supply here and there. Are we distorting the liquidity shocks beyond the fundamental, that channel of QE we've studied very much in people here in the room have studied but it is also liquidity channel there and that's just the paper to be written. So we are looking forward for your paper for next year's session, Klaus. If we invite him, I... Pause in the program, I'm in the program with a five year, five years. Good. Very good, thanks a lot, Ricardo and thanks for the very good questions. So next Titan in the agenda is the launch actually which is going to be served nearby, I guess in the foyer and don't forget the posters session which will take place now during the lunch break and we will reconvene here at 2.30 for the second session. Thank you.