 Good morning everybody and welcome after a great first day of the conference. Welcome to the second day We have much in store today. So we have a third invited paper session, which is this one We have a policy panel coming up the young economist session We are looking forward to yet another Nobel Prize keynote lecture And as I heard the supply bottlenecks at lunchtime have been alleviated. So we look forward to meeting you also there So this is session three on monetary policy and financial markets So we have a pleasure to have two of the most popular academics in the especially in the yield curve community And I would say also especially in the central bank yield curve community and clearly way beyond So we have Dimitri Vajanos of London School of Economics He had to sammel work on integrating the supply demand view of bond pricing with the asset pricing view So we owe him a lot in our field and a lot of central bank models are based on what he has done with co-authors And we have Cynthia Wu of University of Notre Dame with similar impact on what we do I mean many of you may know her for the work on shadow rates But she has done way beyond that that field of research. We have discussions We have Andrea Vlado from the European Central Bank from the Monetary Policy Department and Stefan Dupras from Bank de France also from the Monetary Policy Department So I'm happy to have this impressive lineup here for this session I think we have to stay in time because the Policy panel is shortly afterwards this one. So Without further ado Dimitri the floor is yours, please Thank you very much for this very kind introduction and thank you Organized you and organizing committee for including our paper in the conference. So this is a joint work with Pierre Olivier Goure and Sha and Walker a Okay Okay, so just to motivate in general kind of the question we can have in mind is How does monetary policy whether it's conventional or unconventional? Transmit them domestically in turn internationally now If we think in terms of the standard open economy macro model the Answer is kind of are kind of very clear and simple. First of all The expectations hypothesis of the term structure holds up to constant risk premia constant over time same for a uncovered interest parity and as a result Expectations hypothesis implies that the yield curve in each country is determined fully by current and future expected short rates UIP implies that the exchange rate absorbs any diff deviations between short rates So there is this kind of it insulates one country's yield curve From movements in the other countries short rate and yield curve and finally quantity type of interventions like a QE or Foreign exchange interventions have no effect now this kind of simplified view of Bond prices and exchange rates has some Problems so first of all on the finance side There are some important violations of expectations hypothesis and the UIP so And some important kind of predictability patterns or anomaly kind of pattern financial anomaly pattern So has been documented them extensively since a Bilson and Fama that the currency carry trade is profitable So borrowing low interest rate currencies and investing in higher interest rate currencies is on average Abnormal abnormally high return Then so the high interest rate currencies do not depreciate enough to cover the difference in interest rates and sometimes even appreciate on average then Expectations hypothesis is also violated in important ways So starting with the word of the work of Fama and bliss and Campbell and Schiller It has been documented that the slope of the term structure predicts positively Excess bond returns so times when the term structure is upward sloping or is significantly upward sloping Long-term bonds on average do better than short-term bonds and vice versa long-term bonds do worse than short-term bonds when the term structure is downward sloping So this is the essentially the bond carry trade is profitable Boring on the segment of the term structure at where later rates are low and investing on this on the other segment where h Are high yields are high is profitable Moreover, there is a more interesting kind of a body of work that shows that there are these some interesting connections between pre-prime and bond and currency markets for example people have shown that the Differential between term structure slopes in the in two countries is predictive of the of the profitability of them of the currency carry trade and And finally on kind of if we go leave the finance world and go to the macro world We have kind of it has been documented that QE has been having an effect both domestically and also internationally so QE affects the Domestic bond yields, but also affects exchange rate and affects even foreign bond yields so we'll try to propose a different view of kind of this Kind of international Macro that or let's say for now international finance primarily there's macro element is very limited in the model that I will show you That tries to address some of these phenomena kind of things. So we're going to I would show you a two-country model with a partially segmented bond markets and currency markets So there will be some investor clientele in each market and there was there is this Some integration that is going to be achieved in a partial manner by risk averse arbitrage ours So essentially we're going to distinguish between kind of the outside of the financial market and the core of the financial markets that are these arbitrage ours and Okay, so And these arbitrage ours can be different types of financial institutions So broker dealers can be hedge funds That somehow that through their activities they integrate markets But do so partially because of capital risk aversion capital constraints Etc. So the main results are First of all, we're going to speak to the Predictability patterns of bond and currency returns. We're going to show that that this view of markets Can generate realistic predictions for the predictability of returns and speak to these finance puzzles And at the same time we're going to show that the implications for monetary policy are quite different than those of the standard model So the open economy, let's say a new Keynesian macro model So QE we are going to show that it transmits So QE peters Q purchases lower domestic bond yields They also transmit internationally to foreign bond yields They also lower foreign bond yields and they depreciate the exchange rate conventional policy is also transmitted to foreign bond yields Although we find quantitatively that the transmission is weaker than for QE and More broadly, I mean the the message from these two kind of results is that Floating exchange rates provide limited installation. So monetary Policy at home whether conventional or unconventional affects yields at foreign and I should emphasize affects yields at foreign holding Foreign monetary policy constant Maybe monetary policy also can foreign monetary policy can change in response to Home policy. We're not going to speak to that but we're saying that even if it does not change. There are these spillover effects Okay, so so that's the summary of the paper and then now let me go through the model and Okay, so the model is a builds on my earlier work with villa on the for a closed economy kind of term structure so and Extend this to go to a two country. So there are two countries home and foreign. There is we call the nominal exchange rate by ET So is the home price of foreign currency. So when ET goes up means the foreign currency appreciates then there is a continuum of zero coupon bonds in each country and The We did we call the bond price in country J at time T for maturity tau by PJ T tau we define the yield to maturity in the standard way and We take the short rate quite importantly is exogenous So the short rate is the limit of these yields when maturity goes to zero tau goes to zero We take this as exogenous and specify some process for that that it can be the result of conventional monetary policy so There are as I announced earlier There will be two types of traders in the model that will be arbitrageurs and there will be these clientele of investors in different markets So they are the more somehow sophisticated and interesting agents in the model will be the arbitrageurs We will take a more reduced form in modeling the the investor clientele So arbitrageurs have some wealth WT We use the home currency is the numerator the Invest endogenously some wealth Wft some part of their wealth WT in the in country F so they can be there in the in the cash I mean in the in the in the foreign countries currency or can be in foreign country bonds and They also invest some wealth in Bonds in each country so X so for the amount of wealth that are invested in foreign bonds is included in WFT But we call it format for foreign bonds with maturity tau. We call it X F T tau okay We give them a very simple objective. They are rational. They maximize this mean variance preference over instantaneous changes in wealth And this is the law of motion that is important to understand. It's very in some sense It's very simple conceptually. It's kind of it has a bit of continuous time stuff but the the ideas that arbitrageurs get this kind of basic Let's say default kind of rate W Sorry, I the home short rate on their wealth and then they can get some extras which reflect the Returns of the three trades that they can perform in a kind of in addition to this basic rate which okay first of all they can in The currency carry trade so they can invest some of their wealth in foreign and get the interest rate differential between home and foreign and Of course, there is also the exchange rate appreciation Then so that's the currency carry trade. This is the home bond carry trade So they can invest some amount X HD in home bonds And they get whatever this bonds yield in excess of the short rate of the home short rate and they can also invest in foreign Bonds and again, they are going to get whatever this bonds yield in excess of the foreign short rate All right, so and then so that's one class of agents and of obviously arbitrageurs are going to Optimize over this they will choose this WFT and this X HD and this XFT optimally given opportunity kind of Opportunistically given how high the returns of these trades are and the returns of these trades are going to be endogenously Depending on how much arbitrageurs are investing in them. This would be kind of they would be determined in equilibrium Okay, and then there will be some this client investor clientele's we call them preferred happy that investors and bond investors and currency traders Will take a more reduced form approach for them so So we will assume that them The there is some demand for bonds in country Jane maturity tau that is a downward sloping function of price That so essentially the demand for this for bonds with maturity tau Depends on on this negatively on the bonds price. Also, we're going to put a demand shifter this beta JT this beta here So this will be a demand factor We allow the demand for bonds to be shocked kind of for exogenous reasons to the model and also there is a demand like likewise for foreign currency That We allow it again to be a decreasing function of the exchange rate In fact, we allow it to depend on the real extra to be decreasing the real exchange rate. We allow we introduce kind of in a Exogenous kind of form inflation in this country and we allow it to that this constant over time will now Kind of we essentially specify the demand to be for currency to be a decreasing function of the real exchange rate And we also in have a this gamma T which says a shock to the demand for for currency for okay So so this is them. This is essentially the model. That's it. These are the two set of agents and Given the interaction between these agents. We're going to determine prices the exchange rate and bond prices so markets clear we have We're assuming that bonds are in zero supply This is not this without loss of generality because we can put any kind of positive supply into the demand of Preferred habitat investors we can call it a net demand So demand of arbitrageurs for home bonds plus demand of preferred habitat investors is zero Same thing for foreign bonds same thing for currency okay, so now We're going to start a last thing before kind of closing the models as some notation. So we have we have five risk factors We have short rates to short rates one inch country We have bond demand factors this intercepts in the demand functions and we have currency demand factor So we allow in principle these factors to follow some me reverting dynamics so these factors are all exogenous in the model and There are these matrices gamma and sigma five these five by five matrices that In the last part of the paper where we do the calibration of the model We'll try to get a handle of how we can Quantify the elements of these matrices for now just general alright So we'll start with simple cases and progressively kind of enrich the model So first of all what happens when arbitrageurs are is neutral then we're back in the let's say the classical model so the expectations hypothesis holds expected return of Bonds is going to be equal to the short rate so bonds don't offer and appear returns to relative to the short rate and Obviously, there will be any QE on the yield curve and Also the whole the yield curve in every country will be independent of Shots to the foreign short rate. I mean to the short rates in the other country as long as these shocks do not affect future expected short rates in the country in question okay, then also you IP is going to hold and This will give us the the Mandelian kind of insulation shocks to a Short rate so the shocks to the short rate one country do not affect that does not affect the term structure in the other country okay now Let's now Make arbitrageurs back risk averse So to get kind of our main results and but let's keep them all simplify the model in some other way Let's get rid of demand shocks in bonds and in currency So the only shocks are going to be shocks to the short rates Home and foreign we take the short rates for simplicity to be independent and Specify these simple processes the simple AR one process for the short rates. I mean it's me reverting processes So this is a case where we can get lots of analytical results and kind of in full generality so okay, let us start with Kind of The case where arbitrageurs arbitrageurs are segmented even are not only the Bond and investors and currency investors segmented but even arbitrageurs Let's break our arbitrageurs in some arbitrageurs who trade bonds some arbitrageurs who trade currencies just for To get a handle kind of on the additional effects that will The cross trading of arbitrageurs in different instruments will generate let's first of all assume that arbitrageurs cannot do this cross trading Some arbitrageurs just trade home bonds and just do arbitrage across the home term structure and vice versa for foreign and vice versa for currency okay, so These are the words I will show you two basic results in this model in this segmented kind of arbitrage model so first of all on the exchange rate so the exchange rate kind of underreacts to shocks to to short-trade shocks and What this means the flip side of this result is that the currency carry trade is profitable So when for example, there is an increase in the foreign short rate The expected return of the currency carry trade Increases in other words, it becomes more profitable to invest in the in the in the in the for in the foreign currency so the intuition here builds on a Gavexel majority and also dating back to curry that is is as follows Suppose that there is an increase in the foreign short rate So monetary tightening tightening at foreign then arbitrageurs in principle. They want to invest more in the currency carry trade they want to borrow from home and invest in foreign because they see okay foreign short rate Has gone up. So we want to pocket the short-trade differential now as a result of their pressure of their Buying pressure they want to buy foreign currency to invest at foreign foreign currency is going to appreciate As a result because foreign currency appreciates what happens in equilibrium The currency traders reduce their holdings of foreign currency because now foreign currency is more expensive arbitrageurs find themselves holding more foreign currency and to being used to buying to hold more foreign currency well The currency carry trade has to yield a higher expected return in equilibrium. That's the only that's the only kind of Way for arbitrageurs in equilibrium to be bearing more currency carry trade risk the expected return of the currency carry trade has to go up Okay, so result number one result number two there is a counterpart of this in for the bond market so Foreign sorry Let's say yields in a bond yields in country J home or foreign and underreact Relative to the expectations hypothesis to shocks to the short-rate in that country. So and as a result the Profitability of the bond car carry trade Decreases in the home short-rate. So let me explain. So this in the intuition here builds on my paper with villa. So Let's say that the home short-rate goes down so monetary loosening at home what's going to happen now arbitrageurs are going to Find advantages to borrow short and invest in long-term bonds so This will push prices of long-term bonds up This will As a result the this habitat investors are going to hold less of these bonds because the price of these bonds have gone up They don't they they have this price elastic demand. They don't want to hold many bonds if bonds become more expensive arbitrageurs will find themselves holding more bonds therefore bonds have to have a higher Offered to have to have to offer a higher Expected excess return relative to the short-rate So this means that the current the bond carry trade becomes more profitable Boring short investing in long-term bonds becomes more profitable. It also means that bond yields and they react to the to the To this to the short-rate so bond yields do not go down as much as the expectations hypothesis would say and What okay, so additional implications there are some quantity implications from that from the segmented arbitrage model So quantity changes have effects, but only in the market where they happen. So QE In country J reduces yields in country J does not affect the exchange rate does not affect the other country and same thing for currency interventions, so So essentially we get some insulation results So changes in the monetary conditions at home do not affect foreign So it's a full installation, but this full installation happens because of segmented arbitrage not because of it was of segmented capital markets in a sense not because of And of the standard kind of reason why of this floating exchange rates So now let's now that we got this kind of simple case Under our belt. Let's go to the to the case that we're focusing on in the paper Which is arbitrageers now can trade all the different instruments all bonds and currencies and how they're going to manage this portfolio and kind of Trades trade of these different types of risk will affect the spillover will determine the spillover effects All right So our main results are in this slide and in the next slide. So what we show so Let's say when home short-rate goes down Then okay as we have already seen currency carry trade becomes more profitable It becomes more profitable to invest in the foreign currency. It also becomes more profitable to to do the bond carry carry trade at home Moreover the new element is that a Reduction in the home short-rate pushes even foreign bond prices up and reduces the profitability of the foreign bond carry trade so there is this transmission of Monetary conditions at home to foreign Even holding the short-rate at foreign constant and the path of expected future short-rates constant So what's the intuition here? This has to do with the essentially cross currency hedging by arbitrageers, so okay Home short-rate goes down Arbitrageers are investing hold now in equilibrium more currency foreign currency and more home bonds now Arbitrageers by holding more foreign currency now they find themselves exposed Not only to home short-rate risk, which they demand the compensation for but also to foreign short-rate risk So they are exposed in particular to there is that the foreign short-rate will go down Because this will reduce the value of their currency holdings. What do they do to hedge that risk? they a Good way to hedge that risk is to buy foreign bonds because foreign bonds go up when foreign short-rates go down so as a result of this This trade generates this hedging trade generate price pressure causes bond prices at foreign to increase and the Profitability of the bond of the foreign bond carry trade to decline. Okay Number one now. Let's follow this logic of this hedging of by arbitrage or this kind of Global hedging and let's see what this implies for quantity interventions so Now let's say that there is okay. So QE and at home country J home so the This okay QE at home is going to cause the home currency to depreciate and is going to also Not only cause bond yields at home to go down But also bond yields at foreign to go down will transmit itself to the foreign bond market as well So why is that? arbitrage are going to accommodate QE by Let's say in the extreme case going short bonds. I mean they're going to reduce their position in in in home bonds. So now the Okay, now arbitrage are short in home bonds means that they have a they are exposed to the risk that home short-rates are going to Decline Because if home short-rates are going to decline there are short bonds are going to home bonds are going to become more expensive They are going to lose on their short positions. So what was what trade hedges the risk that home short-rates decline? Well to go long in foreign currency because foreign currency goes up when short appreciate when home short-rates decline Okay, so arbitrage so They buy foreign currency foreign So foreign currency appreciates so domestic currency depreciates more over we have already seen from the previous slide that when That the when they have a position in the foreign currency they hedge this by buying Foreign bonds they has the foreign short-rates risk by buying foreign bonds. So bottom line QE at home causes foreign currency to foreign currency to appreciate and raise and lowers foreign bond yields not only bond yields at home same logic the modest in intervention in the currency market if let's say there is purchase of foreign currency then Okay, the foreign currency of course will appreciate but also Bond yields at home are going to decline and yields at foreign are going to go up And the logic is essentially the same as the hedging logic that we explain in the context of QE If arbitrageers hold a smaller position in foreign currency, let's say they go they then they hedge this position The the hedge let's say short position in foreign currency they hedge this position by buying home bonds Is that kind of the law the same similar logic with them hedging of the for the QE? Okay, so the implications for open economy macro is that So home so the essentially the spillover result conventional or unconventional policy Spills and affects yield curves in both countries holding and Monetary conditions at foreign the same and So and there is this imperfect installation results even though exchange rates are floating There is kind of the standard kind of a installation results fail the standard dilemma fails Okay, so that's these are let's say the main our main results and we show analytically now We can I can tell you a bit more about kind of a our attempt to quantify the model in the remaining of my time, so Remember we'll have this monster five factor a model that we have all that we have this In addition to the two short rates we can have these shocks to the demand for bonds and currency So Okay, so the model still has a simple kind of a fine solution for bond yields and and and for currency So we can Essentially adopt some simple parametric forms for the demand of habitat of bond of a preferred habitat investors in the bond market and for currency traders We also need to parametrize to to make some parametric choices about all these gamma and sigma matrices All these spillover kind of effects of the factors So We try to make it take a minimal structure don't not kind of choose too many parameters different from zero Essentially, we will choose this matrix gamma to be diagonal the for the dynamics of the factors except for one Term that is going to the data tells us that it should be there, which is as feedback from How shocks to into interest rates affect demand for currency somehow that we're going to allow for that And then we're going to take also sigma the shocks to be diagonal except for the correlation between home and for insured trade which the The data tells us that is there is positive and quite significant. So this is for we do we calibrate between the US and the euro area Okay, so we use a bunch of moments of yields and exchange rate changes so second moments so Variance of yields variance of yield changes variance of yield Differentials between home and foreign For we do this for short for the one year yield which would take to be the short rate We also do this for the entire term term structure. We have variance of exchange rate changes We have the ui p moments essentially the regression coefficient of the yield of the yield differential between the two countries on the exchange rate changes Have some moments on training volume so and we try to minimize the distance between the model generated parameters and the parameters that are in the data They are in the moments in the data. So the Okay So the of course we have to fit many more moments in the data than there are parameters in the model We get a decent fit not perfect, but decent. So here we can see the variance of yield changes at home and foreign the the variance of yields the kind of some covariances between the yield changes for one year and for Kind of any maturity anyway, so there is not that far and then what we try to we get the model to try to match some Untargeted moments. So that's the first step. Then what I will tell you talk a bit about monetary policy Some untargeted moments which are first of all the predictability of in the bond market essentially the bond the Relationship between the profitability of the bond carry trade and the slope of the term structure So this is the result from famine bliss that are in the in the top Row and then also the related results from company and shielder essentially both papers do a similar regression kind of just in different form so essentially we find that there is this and As our model predicts we find that them sorry and as is the case in the data that we find that the the When that I mean that the slope of the term structure is positive related to excess return of long-term bonds relative to short-term bonds Not the data tells us something that we don't find in the model that this coefficient of the predictability is Increases with maturity we find kind of fairly flat but anyway, but at least is Significably strong and within the bounds of the data then on the on the currency market. Okay, so We generate the uiP Violations as in the data. Okay, we have put some moments on the uiP in the calibration. So So in some sense we have this is not completely untargeted. However, we also generate this interesting Predictability pattern that I mentioned at the introduction, which that the slope differential between the two countries Is predictive of their excess return of the currency carry trade and this is the last bottom right graph Which is untargeted so Now let's now talk about policies pillovers and then I will conclude so what we do is we do a we look at the shock to conventional policy and Which an anticipated one of 25 basis point decrease in the short rate with a half life of one year So fairly quickly me reverting and also we do a QE shock. That is 10% of GDP But a positive demand shock for bonds and this is more slowly me reverting so So here are the results from the calibration so the So we find that okay, even though theoretically there is a Significance there is a spillover from conventional policy to foreign bond yields holding again foreign From spillover from home short rate movements to foreign yields So holding foreign monetary policy constant we find that this bill over is small so This is the this so essentially the essentially is almost non-existent Obviously the term structure. Let's say the top curve is the shock to the home short rate Affects the term structure at home That's the blue curve and but the foreign term structure barely moves now This is partially due to the fact that we Find a small demand elasticity for currency which matters for the spillover But also partially because we find that the person because of the correlation that we Estimating the data between short rates. So this hedging effect that generates a spillover is not as strong Now we find on the other hand bigger spillovers for QE essentially the effects on of QE at home and foreign are of roughly or same order of magnitude and And also their significant also even for the exchange rate And why there are significant the effects of the QSP lovers are significant partially because of the correlation between home and short rates home and foreign short rates are positively correlated and therefore Kind of portfolio changes of changes to the composition of arbitrageers portfolio Let's say of home bonds can have a significant effect on home on foreign bonds, which are correlated. Okay, so let me conclude So we present an integrated framework to understand term premium currency risk kind of emphasizing their interrelationships so we saw that This framework can help us to make sense to some extent of the predictability patterns in bond and currency markets and the Flip side of the same coin is that this predictability patterns also relate to the transmission of monetary policy so the So we find that the monetary policy transmits Whether conventional or unconventional transmits internationally through precisely the same changes in risk premium that we find that Kind of matter for this predictability patterns. So Now there are two extensions That I wish I I Could tell you more but still kind of his work in progress one is to consider deviations from covered interest parity In this model covering interest parity holds so but If arbitrage with more severe kind of constraints on arbitrageers who could find Violations of covered party interest parity and the other kind of big extension and this is where we have made quite a bit of progress But it's still not at the presentable stages to embed this Setting into a new Keynesian open economy model and to derive many of these things that I took as exogenous in this model like Short rates or like these demands of currency traders and the preferred habitat investors derive them for first primitives I guess since I will tell you more about this at such an effort in the in a closed economy model But anyway, so this is our next agenda. Thank you very much Thanks Dimitri very clear presentation spot-on timing also by the way and Andrea, please And you can think of your questions also online, please So I will receive them and if you online participants have questions, let us know via webex, please So I'm honored to to discuss the paper by Dimitri, Pierre Olivier and Ray Thank you very much to the organizer for the occasion and the usual disclaimer applies so I must refill from the beginning. I find the paper seminal in great part because it also extends the paper by the previous paper by Dimitri and Vila to an international macro setup and For those of you who are still young in in the room. This paper Has actually laid the theoretical foundation of the interplay between preferred habitat investors and arbitrageurs in Explaining bond markets and that has allowed us to understand Like why is QE working in practice? And how does it work and I got to know actually that the paper was written before QE was actually Started after the great financial crisis So in the current model that Dimitri presented today, the arbitrageurs go globally so They are allowed to invest in in foreign currency and in foreign bonds And they are risk-averse. That's very important. They maximize the mean variance utility objective function the second clientele in in the model our preferred habitat investors of both currency Because for example of economical reasons exporters importers or bond traders either in the long end of the curve like pension funds Or in the short and like money market funds. So The two interact and in equilibrium the impartial equilibrium exchange rates and and yields are functions of five state variables the short rates of In the domestic and in the currency these are exogenous But I understand that there's work in that direction of making integrating macro and their foreign currency demand factor and two bond demand factors And the key ingredient of the model is the risk aversion of arbitrageurs and that allows to the model to capture the empirically Violate like evidence for the empirical violated expectation hypothesis. So yields The yield curve does not reflect just future expected short rates But also term premia as well as the uncovered interest rate parity Which has also empirically been shown to to be violated namely that Differentials in short rate are not I mean the exchange rate reflects something more than just the differential between the short rates So the main findings are Domestic conventional a monetary policy does not impact the foreign yield curve or The fact so it's a very tiny impact unconventional monetary policy, however has a sizable impact both at home and Similar to a similar extent. I would say also on on the foreign yield curve and the third I'm finding is that spillovers from both conventional and unconventional monetary policy are very strong on the exchange rate and comparable across the two types of of policy so in today discussion I Will bring a bit of Empirical implications I would cross check the empirical application of the model across more parsimonious non-structural term structure models high-frequency identifications or VR models so Let me Replicate so the first chart is showing the the impact of conventional monetary policy from the model on the current and the the foreign yield curves and As you see the reaction in the non-originatic country is is low. So I Raise the question. Is this something that is for example in line to what we have seen what we are seeing in the current easing Sorry in the current hiking Cycle is it for example reasonable to expect no impact of US Hike expectations and then on on the Eurya yield and for that I Used so one empirical evidence comes from a from a model by Brant et al Who are estimating a sign restricted beaver and you see they and here is plotted the composition a historical decomposition of the 10-year OS rate into your area policy a US policy and your area macro another and US macro and you see that the green part which is showing the US policy impact on the change of 10-year OS so risk-free your area rate is significant now this actually captures Both conventional non-conventional monetary policy in the US, but I would think it's Tendent evidence that also conventional monetary policy would spill over to to your area. So You Dimitri mentioned there is some empirical Influence of the correlation between the two short rates has that maybe have changed over time with the economy turning more global and that could Maybe let's say speak to or suggest that currently there are stronger spillovers from Domestic which is US and that is actually a second point. Is there something? That given that US is documented to have to play a very large role in In the economy like a global currency could that maybe Be twisted so like an asymmetric in from a spillover from domestic US Currency conventional monetary policy to for example your area like smaller economies. So maybe this This first empirical finding could be Tested differently so could be empirically different Then Let me turn to the impact of a rate cut so conventional points monetary policy on on the yield curve Dimitri explained that Then upon the policy rate cut bond yields do not do not drop all the way to the value implied by the expectation Hypothesis and this is something that for example one one US paper that also Looks into a term structure model Abraham Zendal also confirm That would mean that so term premia tends to increase upon rate cuts And then I thought of okay So how do things look like in the in the euro area and in the euro area we find there's a model that by by Geiger and Schup that actually find that When accounting also for the lower bound and I'll get to that a rate cut is actually estimated to to lower expectation component but also the term premia so I would I mean it's not an additional thing but when you show So when you have here like the the blue line or the red like the big the big reaction How about decomposing how much is expectation component behind the curve and how much is the reaction of the term premia such that we would be I think We would be we would like to know where is the ratio of the two components? how is the ratio of the two components going across the curve and another point We tend to think that easing like the the previous easing cycle was associated with lower even negative premia And the current hiking cycle in the current hiking cycle for term premia tends to generally increase And and turn positive if they maybe have were positive So I know that you mentioned in the paper that there is you don't want to talk about Unconditional moments, and it's all about the reaction, but I think it would be informative for the literature to also mention be like science and Of premia in and various hiking or easing cycles last thing here Maybe you can discuss Whether they can be their scope for implementing an asymmetric Impact of easing versus tightening cycles So if this can some is if there is a channel in the model that could allow to to explain this possible asymmetry now a smaller point relates to the to the impact of of estimating the data on a sample that basically covers also the low interest rates period or even Negative interest rates in the euro area and that by now has turned to be a third to cover a third of the sample that's using the estimation So I would find it interesting if the authors would Comment on what is the caveat of not accounting for the for the lower bound because in the paper by vaccine jam previous paper by Xinjiang and Dora you see here them The there's a slight difference in the impact of conventional monetary policy on the short trade and I think that that's worth a bit discussing whereas So the impact in the lower bound period in the in the US according to the to the paper by Maxi is found to be smaller. However in the euro area because we There there was the the negative interest rate policies The reaction in the second in the lower panel which shows the impact of a ten basis points rate cut On the on the yield curve is actually found to be similar if not slightly larger in the non in the negative interest rate a policy period and for example one channel that is That could explain that is the seek for yield Channel so maybe the lower but not accounting for the lower bound plays a smaller role in the euro area than in the in the US but I think it's it's a point worth discussing and You also mentioned that there is the Nominal and real version of the model. So maybe One way to go about it would be to calibrate the model on on really it curves so next moving to the impact of unconventional monetary policy on the domestic and foreign yield curve the impacts are are sizeable and large and the authors do a good job in explaining the sensitivity of these spillovers and the impact on the risk aversion parameter So the smaller than the risk aversion the smaller also them the relative impact on the on the non originating country The correlation of the short rates also here plays a role As well as the elasticity demand of of currency traders and the way that is calibrated But there are I think that there are a few other points that are Worth discussing because they are not they are left out from the current calibration. So for example, it might be reasonable to assume that Quantitative easing or quantitative tightening Comes in in a cross-border positive correlation. This is exact I mean, this is the empirical Our experience from from the past decade and it would be interesting to see if one allows for this additional flexibility in which direction Do the results go? At the same time also conventional and unconventional monetary policy shocks may be positively correlated And that is also one aspect that I think it's worth mentioning and discussing the implicate the possible implications And a separate point it refers to the to the shape of the of the reaction especially when we think about the euro area you see here a decrease of the slope but Especially given that the model is estimated on on on bond yield It's might might be worth mentioning also the heterogeneity of the impact of of QE in the euro area across the various across the core and the periphery countries because for example the pep experience has shown that There are also other shape other reactions empirical of empirically of the of the yield curve. So the curve can also Basically shift parallel and I think it's it's worth discussing or briefly mentioning also that in in your exercise and lastly the last point I would like to make it it relates to to the path of QE and QT and how do how do the authors and how is in calibrate that versus how it is Implemented the fact of my policy makers. So In in the model the QE Shock is amounting to 10% of the euro area of the US GDP it is unanticipating occurring at time zero and unwinding with a half life of seven years. So I think that it's it's the same amount Like 10% of your of US GDP that represent also the surprise in the euro area And that's why I think it's I mean I wanted to compare that with the initial a pp package which was announced in in January 2015 and that one actually amounted to 12% of the euro area GDP. So it's a similar overall envelope that the authors are considering that speaks to the first a pp Announcement yet the profile as you actually see in in the right-hand side chart is different than Then the typical I mean the way that the authors have implemented in the model though And so though the yield impact which you see in the left-hand side chart We find in this paper to be similar overall like so decrease of the slope and Close to 50 basis points for the big four countries. The profile is of conducting QE is Is not happening at time zero. So if there's a gradual build-up of the portfolio and then there's also Then the unwinding pace may be maybe differently so Concrete comment would it make a difference if you were to implement a more realistic Profile of QE and how about QT? We know that QT also in terms of of profile can be different than QE It's anticipated. It's the pace of one unwinding is Slightly different. So what will be the implications of the model for the QT phase, which we are? Thank you so much Lot Andrea for the discussion a lot of smart economists working on yield curve topics So looking forward to have your questions, please Yeah, thank you very much for a great presentation and wonderful discussion also And my name is Claire Tuerwächter. I'm from Morgan Stanley research And I would have one clarifying question to The author I've been in the initial Presentation you said that there were in assumptions on the short term rates of home and foreign to be independent of each other And I wasn't sure if that was then dropped along the way of Reducing the assumptions and if it hasn't been dropped my question to you would be what you have thoughts are on like Synchronousy and global cycles thinking of your co-author Helen Ray. I'm sure you have some Thoughts and discussions on this among yourselves as well. Thank you. Dimitri. Is it okay? If we collect and then you respond to Andrea and the questions. Okay, good Very good. And if you need more paper, please I think it's quite similar my question, but I would go a little step further So I would expect that if contemporaneously in both countries you have a cute QE You don't have these incentive for our beat rogers to purchase The bonds in the other country So you would have no effects no spillover am I right and does this affect also? Then going back to the discussion if you allow for a positive correlation, then probably Defects when only one country do the QE are much stronger than what you obtain I have two questions from remote from from colleagues from the ECB So Alexander Jung is asking on the asymmetry QE versus QT I mean that already Andrea touched upon but he also asked about spillover from forward guidance specifically as a monitor policy Tool, so you had conventional and QE and he asked some forward guidance and our colleague tannily makinen asks Can them a model shed light on the relation between exchange rate movements and order flows Just the first question again. Yeah, so he just so you had conventional policy and QE And he asked on spillovers from fat forward guidance So if you have forward guidance policy, how would you think about the spillover of that specific policy tool? Maybe one last from the audience Jack more No, and another one. Yeah, fine. Very good. Yes, just a quick clarification I mean from the title of the paper one would be inclined to think that the prefer habitat investors play a crucial role now in your explanation a lot of intuition rests on Arbitrageers now what are prefer habitat investor actually doing in in this world? Yesterday question. So how much so one of the hypothesis is that? Investor hedge, so I'm trying to understand how how much do they really hedge? Okay, how much is relevant this hypothesis because for example pension funds might be So they have to hedge may well be but not all investors. So how much is relevant also when you do the empirical? The empirical exercise that you don't find so much hedging going on So I'm trying to understand whether you might dig a little bit Okay, on the different originator of investor that might will look and search for yields and might will have long-run Expectation instead of short-run expectations. Okay, so Also when I see a QE and I do expect they also the other country is going to do the QE What am I going to do so? Thank you Thanks for the questions. So please Dimitri Should I well as you prefer? All right, so First of all, okay, big thanks to my to Andrea for her Excellent discussion. So I think that it's kind of it's very very good and very constructive So I think that the number of things that she suggested as things that we could actually do like to kind of for some Look at the effects of QE when QE is Simultaneous in the book in both countries or when it's it's kind of correlated with Conventional policy and actually this relates to one of the questions and I will say a few more things Or for example, we can do this richer We can get through go through the mod get through the model of this richer pattern of QE that is suggesting that kind of It's like essentially forward guidance on QE essentially that QE is a kind of anticipated the pattern of purchase We can calculate that then The Anyway, and the few other things that she said so Yeah, so let me just let me just leave it at that I mean, I guess on the data on the I would put also question mark on the data on the On the big spillovers from the US to you to Euror area bond yields of the US conventional policy area bond yields I guess there are one thing that I would like to know and investigate more is whether US conventional policy may signal something about expectations about future euro rates and somehow whether this is controlled by the Empirical analysis because our model of course is a theoretical result and it's very stark and says that given that Taking everything about the path of not only current short rate in one country But also the path of expected future short rates to be the same the effect of conventional policy is weak, but I guess the question is What is being held constant? So that's that's something that I would I would I guess I Would try to look a bit more in the in the data and that's for us to do that work. So now on the questions So very many and very good questions. So on the correlation between short rates We okay. Yes, the answer to your question is in the In the numerical part that I showed you yes there we allow for a correlation between short rates and we estimate that actually from the data and actually it's quite high as you say and all these kind of calibrated results are under that assumption the analytical results that I show you in the main part of the paper are with independent short rates, but this is just to Kind of isolate kind of some simple effects now the On the contemporaneous clear cross countries I think that you are right and also this relates to a point by liana that that by Andrea so that the the If it's quite important use there should not be effect on the foreign currency in this hatching effect kind of should go away So whatever effect we get for international spillover should be because of the correlation the correlation due to the correlation of the short rates so indeed it may be the case that is weaker for the The spillover effects are the effect of QE at home is going to be stronger than the effect of QE The spillover effect at foreign. I think that is something actually for us something to investigate the Order flows and the exchange exchange rates. Yes, that's essentially the model is designed to deliver this effect and We have some actually even some estimates from that we from the literature that On what how big is the effect of order flows and exchange rates that we used to calibrate the model? So the kind of elasticity. Yeah, so This on demand elasticities, but yes, the model is essentially this is something in the built in the DNA of the model I mean, it's not something that kind of is very some very basic feature of the model Foreign guidance spillovers. Okay, so that's in there are spillovers of foreign or forward guidance although now I am unable to tell you exactly what the direction is but but yes, I Can respond to that a bit late, maybe but but but there are because essentially foreign forward guidance means that arbitrageers are going to take some position In in bonds in order to kind of essentially incorporate their the news about Kind of short-term expectations to bond yields and this will generate some some some trade Also on the foreign currency, but I cannot work it out through on my head now They're all of preferred habitat investors. What is their role? Yeah, so preferred habitat investors essentially I talked about them primarily in the context of the demand elasticity that they have Essentially, I say that arbitrageers buy bonds This means that the bond prices go up. This means that preferred habitat investors selling bond to arbitrageers This means arbitrageers are finding themselves with a bigger bond position and therefore they demand Higher kind of excess return from the bond carry trade That's a very kind of let's say in some sense rudimentary kind of role of preferred very primitive very simple role of preferred habitat investors But of course there is another role that I did not really emphasize so much which is that the preferred habitat investors in principle also Generate kind of changes to the demand for bonds. They can have the shocks to their demand that this this factor beta that I that was in the model Which I did not really emphasize very much in the calibration. We kind of make made this factor fairly simple But this factor might relate these changes to the shock the shocks to the demand of preferred habitat investors Which of course are exogenous in the model. I should be very clear. It's just a factor May account for example some of the response for example in of risk-premia that something that Andrea touched on her discussion that following a monetary loosening then risk-premia risk-premia decline in In the euro area bond risk-premia this might be because this loosening could cause it could this shock to are to the interest rate why I can Trigger a change to the demand to the demand of for preferred habitat investor various reasons Of course, this is a bit of a black box in our model But perhaps with some data one could discipline this I mean there are these very there are some very nice data on investor holdings and we can try to get some some Try to get a sense of how the demand how the demand of preferred habitat investors Changes in response to to monetary policy shocks and then we can say some interesting things about Kind of data basically discipline about Okay, how the effects risk-premia now and the finally, okay, so Laura's question how do much to investors head how do arbitrageur's hedge? I guess you want me to again, that's a bit of an I mean hedging is of course It's important in our model in some for a number of exercises that we do I Mean we observe and if we observe for example that many times Domestic investors invest in foreign currency, but they hedge this position In sorry investing foreign currency currency instruments in foreign bonds But they hedge this position by training in the currency market to get a kind of guaranteed return in in domestic currency now So there is kind of people who do trade multiple instruments and there is some risk management there But okay, how much I guess we have to look again at some data to discipline that thank you Dimitri and So we hand over to Cynthia another model that has QE in there as a common feature with what we just heard, please Cynthia Of course yours All right, I would thank the organizers for inviting me. This is a joint work with the Yan Xixi from Bank of Canada In response to COVID-19 crisis the Fed has double its balance sheet to nine trillion dollars And for the same period of time the Treasury has also implement a sequence of stimulus that includes a paycheck Protection program the economic impact payments and both of the two programs amount to about eight hundred billion dollars Now our research question is how do we compare all those emergency monitoring physical policy? and We answer this question with a tractable new Keynesian model in our model there are two types of households the constrained household and an unconstrained household and The financial market is also segmented similar to Dimitri's paper And we have two types of bonds a short-term bond and a long-term bond The financial intermediary plays some maturity transformation. They also face a leverage constraint There are four types of policy instruments in our model conventional monetary policy QE which is modeled as the central banks holding of long-term bonds and Two types of fiscal policy allow some transfer to the constrained household as well as government purchases So here's a preview of our main result We compare between QE and physical policy and we find as long as the physical policy is tax financed Those three types of policy will have the same aggregate implications Now when we compare between those three types of policies and conventional monetary policy We find the conventional monetary policy is more inflationary We also discuss the redistribution effect and we find QE and transfers have some redistribution effect Where as conventional monetary policy and government purchases do not And in our model recording equivalence breaks Specifically we find physical policies more stimulus when it is financed by taxes compared when it's financed by debt Now in our paper. We also study some optimal policy coordination, which I will not have time to talk about today So here's a plan of my talk I Will start with three and proco motivations we would like to speak to with our model and then I would discuss our linear New Keynesian model and talk about its properties and essentially connect back to the and proco motivations I start with and Finally, I will talk about the full model behind the linear model and see how it's derived So let me begin with the and proco motivations and here what I'm showing is a breakdown of the federal debt in the United States The first panel shows the amount of the federal debt that's held by the Federal Reserve in trillions of dollars now before QE starts it's about half a trillion dollars and this is after QE 123 and That's post COVID. So now it's above six trillion dollars The middle panel shows the amount of federal debt that's held by the US public. So you can see that while the Fed has increasing its holding Dramatically the amount of the you are the amount of the debt that's held by public has also grown a lot The reason for those two simultaneously happens because while the Federal Reserve is Absorbing the debt at a fast speed the Treasury has been issuing the debt at a faster speed Now here is the what I want to highlight in the and proco literature We typically focus on the central banks balance sheet, which is summarized by the first picture or variation of it From this aspect the and proco literature has argued that QE has been expansionary Whereas in theory especially in a theory about QE and for example that includes a work by Gerlary Cruddy and my own work In those theoretical model They show that we should focus on the joint balance sheet between the Federal Reserve and the Treasury and Specifically if we look at the middle panel, which is a total amount of the debt that's held by the public The conclusion from the theory would be the balance sheet policy Since a great recession would have been contractionary So the question I would like to raise here is does an and proco literature miss a dominant piece This part or does a rapid of debt growth of the Treasury not matter in theory So that's the question we would like to address with our model and I will come back to it The second and proco motivation is about physical multiplier So I would like to highlight two parts the first part is the estimates of the physical multiplier is a range So specifically I'm quoting the survey article by Valerie Raimi in JEP and in her table She summarized those Adam estimates from various papers and the ranges between three point three and point eight And the second part I would like to highlight is the physical multiplier is stay dependent and specifically I would like to highlight the dependence on the financing method so here I list two papers the first paper uses a micro data and Finds that when the government finances expenditure by debt it crawls out private loans and Consequently reduce the multiplier now the mechanism in our model works precise like how she describes in her and proco studies The second paper use a cross cross country evidence to show that multiplier is smaller in countries with higher debt to GDP ratio The third aspect I would like to talk about is transfers and I would like to highlight two parts The first is aggregate implication of transfers and Here's a sequence of paper defined the US government stimulus package during recessions increased household spending and the second part is about the distribution consequences and This paper find the responses are larger for poor household with lower wealth and income so now those are the three and proco motivations and Now let's talk about the linear models its properties and how we connect back to the and proco motivations So I will start with the textbook version of New Keynesian model with the IS curve and the Phillips curve Why is output I Is the one period nominal interest rate? Pi is inflation the head variables for log deviation from the study state and We have the IS curve on the top that summarized the demand side of the economy and The Phillips curve at the bottom that summarized the supply side of the economy the parameters here are all standard Sigma is the inverse of intertemporal Elasticity of substitution beta is a discount factor now in a typical model We just summarize those two parameters together and that's a slope of the Phillips curve and This one additional parameter that captures the study state share of the unconstrained household consumption in Output now in a standard model all the households are unconstrained. So this parameter is just one and it drops out Here's our model I Use a blue terms to highlight the new components So you can see the additional policy includes QE transfers and government spending and They enter both the IS curve and the Phillips curve now. How do they work? QE works by relaxing the financial Intermediaries leverage constraint so that allow the constraint household to borrow more Which allow them to consume more and as a result it stimulates aggregate demand Transfer works similarly it hands out checks to the constraint household So they are able to consume more and as a result it also stimulates aggregate demand Now one key parameter I want to highlight here is a parameter eta This parameter captures a fraction of fiscal policy. That's financed by taxes It's between zero and one and one minus eta is a fraction of fiscal policy. That's financed by that So here's the first result now in the literature typically we either study QE or study fiscal policy But typically we study them separately so our paper make the effort to bridge them together and we find the effects of QE and Tax finance the physical policy whether it's government spending or transfers have the same aggregate implications Now if you like to look at the equation Here's how we get the conclusion We impose eta equals one that means all the physical policy is financed by taxes and in this case you can see QE Transfers government spending enters the equations all this in the same manner and in three different locations And I want to highlight all of them enter both into the demand side of the economy as well as the supply side of the economy Now the second result is about inflation We compare those four different types of policies and we show that to provide the same amount of stimulus Conventional monetary policy is more inflationary compared to QE and fiscal policy And this result is consistent with the literature for example in my restart paper We make a similar comparison between conventional monetary policy and QE although that paper doesn't have the physical policy Now there's also some empirical literature show that physical policy is not inflation not that inflationary Now in our follow-up work We also look into more directly into the empirical implication of the model result in terms of inflation And the preliminary result supports this claim now Why is conventional monetary policy different? Because you can see all four types of policy could be stimulative That's because they all enter into the IS curve, but that's not the case for the Phillips curve Conventional monetary policy does not enter into the Phillips curve Whereas the other three types do and you can see if there's an expansionary QE transfer or government spending What it does is it puts downward pressure on the Phillips curve So this is an active sign now I'm not saying that those policies are deflationary What I'm claiming is that there's this additional downward pressure through the supply channel on top of the Inflationary channel through the demand through the demand channel, which is in a standard model So why these why this happens why there's a downward pressure? Because those policy crowd out to the consumption of unconstrained household so those households have to go out to work more which puts downward pressure on wage and That's put puts downward pressure on prices So now we can speak to the 2021-2022 inflation surge with our very Simple model at that period before the inflation takes off the logic order of the fact Titanic would be to unwind the balance sheet first and then raise a policy rate But what actually happened in response to the high inflation is the fact raise a policy rate from the zero lower bound To over 5% and during the same period of time they barely wind it down the balance sheet Now this is actually consistent with our model predictions such that tightening the policy rate is more effective in lowering inflation compared to the quantitative tightening For the same period of time the physical authority has provided another round of stimulative What they do is they hand out checks and to help our households to alleviate from the increased cost of living So for example in late 2022 17 states have done so Now our model implication of the combination of the two policies the tightening of conventional monetary policy the rate hike together with expansionary physical policy together Have the potential to lower inflation, but does not cause a major contraction Now we can talk about redistribution In our model QE and tax finance the transfers have this redistribution effect and it goes from the unconstrained household To the constrained household Whereas the other two types of policy the conventional monetary policy and tax finance government spending do not have such a channel Why that's the case? Well QE and transfers they both work to relax the constrained households budget constraints So the constrained households is able to consume more So the resource redistributes from unconstrained households to the constrained households Whereas for the policy rate and government spending although both of them could provide the aggregate stimulus They do not go through the constrained households Now we can go back to the empirical motivation in terms of transfers and Two of our propositions the proposition one and this proposition three speak to the Emporical results and proposition one says the transfers are not neutral not like in a typical standard representative agent you can see a model and proposition three says the transfer Redistribute wealth from unconstrained household to the constrained house and both of them are consistent with empirical findings Now so far all the physical policy are the discussion based on tax finance the physical policy What happens when the physical policy is financed by debt? We show that once the physical policy is financed by debt it doesn't have any aggregate effects now again to see that Let's look at the IS carves and the Phillips carves When we impose it out to be zero Those great terms drops out So transfers and government spending no longer enter into the two equations only QE does Now what's the intuition? Transfers and government spending themselves are expansionary But at the same time when the government needs to finance them they issue long-term bonds the fact that the issue long-term bonds itself is Contractionary now in our very stylized model those two effects just counter each other out at the there's a zero net result So now we can speak to the empirical motivation about the balance sheet policy When we talk about the balance sheet policy the question is should we worry about when the Treasury is issuing long-term bonds as The opposite effect of QE Now based on this result when the government issue in long-term bonds there is a contractionary effect However, this contractionary effect is countered by the expansionary effect of government spending transfers So those two effects are cancelled out especially in this very simple model So consequently what's remaining is a central banks balance sheet So essentially what our model does is supports the practice in empirical literature that we should primarily focus on the Federal Reserve's action and Not worry so much about the Treasury issue in long-term bonds Now let's talk about the recording equivalence in our model recording equivalence breaks because tax finance is different from that finance Specifically when a larger fraction of physical policy is tax financed government expenditure transfers or more stimulative So again the same equation the is curve The key parameter is this eta So in this equation when eta is larger that means we have more More physical policy that's financed by texas It is more stimulative So what matters is not total amount of transfers or government spending But there's a total amount of transfers and government spending that is financed by taxes empirical motivation on the physical multiplier So I described earlier in the data the physical multiplier has a range between 0.3 and 0.8 Now in our model This is still a range if you look at the y axis of the graph It's between 0 and 0.72 So in a very simple model and standard calibration We're able to generate a reasonable range in the model And also I want to highlight the state dependency in terms of this picture That the physical multiplier is an increasing function of eta And eta is the fraction of physical policy that's financed by texas So the more of the physical policy that's financed by texas the more effective they are And both of those results are consistent with the empirical findings All right So now I describe the linear model the properties and discuss how they relate to the empirical motivations I started with Now let's see what is the machinery behind the linear model And how do we start from the first principles to get those is curves and flips curves In the background there are six different sectors There are two kinds of households unconstrained household and constrained household The unconstrained household is pretty much standard They say via this short term that instrument is a we call it the deposits And they pay texas Now the constrained household is relatively non-standard What they do in our model is they issue long-term bonds to finance their consumption And they also receive transfers from the government The financial intermediaries stay between the constrained household and unconstrained household And they perform maturity transformation At the same time they face a leverage constraint The firms are totally standard Where they do they face a cowl sticky price and the central bank implements two types of monetary policy QE and conventional monetary policy The government implements two types of physical policy They make transfers to the constrained household and they make purchases And they finance their spending by either tax the unconstrained household or issue long-term bonds Now the the unconstrained household They're standard they maximize their utility and they get utility from consumption And they get this utility from labor supply Their budget constraint They can consume or they can save with this deposit going from t to t plus one So it's one period nominal saving vehicle On the income side they that's income from wage And those are total proceeds of the deposit from last period and that just captures all the dividends and transfers and taxes First order conditions are standard the first the first first order condition captures the labor supply condition And the second one is the order equation which Which is a A function of the nominal short-term interest rate here constrained household in our model the constrained household does not work And the reason we we assume they don't work is for tractability alone and to get the is curve and flips curve We need to make some assumptions now. We do robustness check numerically when we assuming they work in appendix as well They're less patient than an unconstrained household So equally in equilibrium the constrained households or borrowers whereas unconstrained households are lenders The constrained household finance is consumption by issue and long-term bond Now the reason we call them constrained i'd put a quote here is for the following First the constrained household the borrowing is limited because of the leverage constraint that the bank sector is facing So pretty differently the constraint is not literally on the constrained household themselves and their order equation still holds But the constraint is transferred from the banking sector Now another reason we call them the constrained household is because all those structure differently But they in terms of certain aspect for example transfers They behave similar to the hand-to-mouth household in the tank model That's probably what you have in mind in terms of the constrained household Although they are not exactly hand-to-mouth in our model. So those are the constrained household They maximize their utility only over consumption and the discount factor beta is smaller than the discount factor of the unconstrained household So this makes the constrained household the borrower Their budget constraint They consume Now that's a total coupon payment of the debt they issued previously The new debt issuance days period Those are the transfers from unconstrained household and from the government And they also have this first order condition the order equation now The difference between this order equation and unconstrained household's order equation are the two folds first The discount factor is different and two this order equation is used to price the long-term bonds So rt plus one is a holding period return of long-term bonds from t to t plus one So essentially we can imagine the unconstrained household is pricing the short-term bonds Whereas constrained household is pricing the long-term bonds. So that's a market segmentation in the model Now between the two types of households is this financial intermediary Financial intermediary lives for one period again. This is a tractability assumption That's similar to my previous paper and we need this assumption to reduce the system down But it does not really change the quantitative result of the model much They're balance sheet condition They can hold long-term bonds. They can hold the reserves issued by the central bank And they can fund their holding through deposits and their own net worth And their own net worth has two parts. One is a new equity injection from the household and the second part is essentially coming from the fact we assume the financial intermediary only live for one period So they're inheriting the wealth from the previous intermediary that exceeded previous period Now the key thing here is a leverage constraint So the amount of the long-term bond the intermediary can hold Is smaller than or equal to a factor Of their own net worth Financial intermediary maximizes the dividends Discount discounted by the unconstrained households to showcase the discount factor and subject to the leverage constraint So here are the two first order conditions For the first one omega is a Lagrangian multiplier on the leverage constraint. So what that means When the leverage constraint binds, which is a relevant case in our model In that case omega is positive Omega is positive means that the return of holding long-term bonds Is higher than return of holding the short-term bonds So this is how we generate a positive yield curve in these types of model And this is essentially how QE works in the model QE works by relaxes the financial intermediaries leverage constraint which flatten the yield curve And then transmit into the real economy through the consumption or the constraint of the constraint household Now the second first order condition is a standard what that says is essentially that Without any friction between the reserves and deposits Holding one period deposit is the same as holding one period reserves They give you the same return because both of them are short-term Instruments and there's no risk either way The central bank The central bank implements a standard Taylor rule Their balance sheet condition they they can hold long-term bonds and they finance the long-term bonds with reserves And we define QE as a real quantity of long-term bonds that are holding And Q is a price and B is a real quantity And the central bank returns surplus now it could return surplus to the fiscal authority or the households and their equivalent just like The only difference is how the math shows up physical authority What does the physical authority do they can make transfers They can spend They have to pay off the coupons from the bonds they issue previously Now on the income side the issue new bonds And that's tax income The tax revenue we break them down into two terms The first term is the key term we want to capture in the model Which is a eight a fraction of the total transfers and government spending The second term we add here, which has an interpretation of the in the literature for fiscal responsibility So essentially the physical authority have to tax a fraction of the debt previously to make the system sustainable And for the reason we need to have this additional term is so that the non-linear model has a unique solution in the equilibrium and the to guarantee the uniqueness We need the following condition, which is purely technical reason for our model The equilibrium conditions The goods market clears such that the output can be consumed by unconstrained household Constrained household and the government The SM market clears such that The bond could be issued by the government The constrained household Which can be held by either financial intermediary or the central bank Now we have made some additional assumptions to reduce for the system to be able to reduce and that's the last one We make an additional what we call convenience assumption And that's the assumption on the transfer between the unconstrained household to the constrained household As a result of this assumption The constrained household's consumption depends on QE transfers now again transfers and government spending So the system has a total of 24 equations and 24 unknowns and with some algebra We're able to reduce it down to the IS curve and Phillips curve Now after describing the full model we can talk about additional transmission mechanism and additional intuitions. First, I want to compare between QE government spending and the transfers So let's start with the constrained household consumption And we're based the discussion on the tax finance of eta equals one And in this case, you can see that the constrained household consumption Depends on QE and transfers, but not government spending How do they work? QE allows the constrained household to increase their consumption because they are allowed to issue more bonds Transfer works similarly the household the constrained households receive this additional check which allow them to consume more So because both of those two policies work through the constrained household and they have a distribution effect Now here's the aggregate resource constraint Unlike the constrained household consumption here, we have both QE and transfers, but also the government spending So government spending enters the aggregate resource constraint the same way as QE and the same way as transfers So The takeaway here is those three types of policy have the same aggregate effect But they have different Redistribution effect and the reason is because government spending although it enters into the aggregate resource constraint It does not enter into it does not affect the constrained household So that sets apart between the other two policies and government spending although in aggregate they look similar Finally, I will talk about the breakdown of the recording equivalence So when the physical policy is financed by that, which means eta equals zero Now in this case, you can see that only QE affects the aggregate resource constraint and both government spending transfers dropped out So that means debt finance physical policy have no aggregate implications Now why this happens? Physical policy itself is stimulative But on the flip side when the government needs to finance those physical policy by issuing bonds The issuing bonds part is contractionary in our model Now why it's contractionary? Because the total amount of the bond demand is Exogenous And that's dictated by the leverage constraint of the financial intermediary as well as QE When those are decided the government issue bonds As a result it crowd out the private sector. So the constrained household are not allowed to issue the amount of bonds they want And it lowers the consumption of the constrained household, which is contractionary And in our model because it's a very simple simple model the two parts of the effects just cancel out So the net effect is government spending and transfers drop out So to conclude we propose a tractable model that features four types of government policy And we show that QE and tax finance the physical policy have the same aggregate effects Now compared with those three types of policies conventional monetary policy is more inflationary Among the four types of policies QE and transfers have a redistribution effect whereas conventional monetary policy and government spending do not And in our model recording equivalence breaks And in our paper, we also discuss implications for optimal coordinated policies So our model reconciles with four with the three empirical motivations. I started my Talk with The first one is that the balance sheet policy should be summarized by the central bank's bond holding And that's a practice in empirical literature. And that's what our model suggests as well and two We show that fiscal multiplier depends on the financing method and as a result it could be arranged instead of just a simple number And the final one is that transfers are both stimulative and have redistribution effects And thank you That extends the scope to also the fiscal dimension. I see a couple of colleagues from the fiscal Division here also so good to cover everything Devon, please great Many thanks first to the organizer for the opportunity to discuss this paper Because it is after something which I think is extremely relevant Namely to update the baseline UK engine model to the recognition that the great moderation ended some 15 years ago now Because this baseline UK engine model was really the model of the great moderation And it's embed some of the reality and also of the policy consensus of the time number one Monetary policy is better suited than fiscal policy for stabilization purposes and number two Monetary policy is implemented through control of the short-term interest rate now The great moderation ended in 2008 because you know stuff happened Uh, and when facts change as kain said we should change all two equation model But we haven't really So what should we add really to this model if we want to take into account where we've learned in the past 15 years First of course, we need to add the importance of financial intermediation and a role for quantitative easing This is actually something that cintia already did in a paper with eric sims 2021 They Embedded qe in this two equation model But something else that we learn if not starting in 2008 at least 2011 and especially in the past six seven years Is the importance of household heterogeneity which creates a role for fiscal transfers And this is what cintia is doing here with yinxi they going to keep qe and also add a role for household heterogeneity and fiscal transfers So obviously if you want to fit into two equation Qe financial intermediation household heterogeneity and fiscal transfers that's starting to be a lot to put into two equations But the clever thing that cintia and yinxi realize is that Actually the sims rule paper is already a tank model. It's already a model with A saving household and a borrowing household You can think of it basically as a gertler Karate model instead that the firm this time is a household that borrows its consumption from the financial intermediary So if you want to add Fiscal transfer you just need to add fiscal transfer here. There's nothing else that you need to add to get the household heterogeneity part So you can sum up the past 15 years of macro research in two equations, but that's actually an understatement Because there are also new lessons from the paper that you will not necessarily find in either the quantitative easing literature or the household heterogeneity literature So for instance first for the same effect on aggregate demand qe transfer and government spending are less inflationary Than conventional interest rate policy. That's not something that have been so much emphasized in either literature Number two quantitative easing and transfer would distribute from savers to to borrowers households, but not government spending and conventional interest rate policy The first part is kind of intuitive, but the second one is actually quite at odds with the household heterogeneity literature. So it's already New and also a bit thought-provoking and third Deb fine and government spending is less stimulative than tax finance government spending Your system one brain might be thinking oh sure I know this from islam, but notice it is not the same thing It's actually the opposite from the islam result and that comes from a new eviction result From the lover's concern of the financial intermediaries Okay, so In such a model obviously you need to make a lot of simplifying assumption So i'm not going to ask for a secondary liquid asset irreversible investment and the like instead I just want to point at some assumption That I think are unnecessarily strong But matter for the result and that could be relaxed without losing much on tractability all right comment number one is About maybe embedding more of the insight from the Hank literature Into the model because right now the model kind of superimposes two interpretation Of its model the first one is interpreting this borrowing household as basically a proxy for firms In a full-fledged gertler charity model and this is kind of the interpretation in the previous paper on qe only But now that there is transfers to constrained household it also wants to Talk about this household as like the hand to mask constrained household of a tank model This is not a problem in itself of course, but right now it implies that it makes some assumptions That abstract from some important channel that have been emphasized in the hank literature The main assumption I have in mind here is the assumption That the constrained household doesn't have labor or capital income Its only income is coming from a transfer from the other household Or from the government if the government wants to do a transfer and that's kind of important because the hank literature has emphasized The importance of indirect income channel in the transmission of either monetary or fiscal policy So it's going to matter for instance in matters for result number two that A conventional interest rate policy don't have distributive effect and they don't affect the consumption of the constrained household The hank literature has emphasized and said that it can have even conventional interest rate policy can have Large distributive effect and it can affect even constrained household through indirect Income channel so the main one will be the canton cross, but there is also Adrienne Eau Claire interest rate exposure channel the fissure depth deflation channel This is for theory, but empirically The empirical hank literature has also find that those mechanisms are very strong in the data There are several papers that do this here. I just picked the one with the most dramatic picture. This is from Anna Arrogantini-Pinco here at dcb and our co-authors Based on swedish data and you see for instance for labor and gum At the at the top here that you have a very heterogeneous response of labor and come in particular that The response of the labor income of the bottom income distribution is quite strong so you can think of those income maybe as the constrained households of a tank and The the impact of monetary policy is actually even stronger on them because of those indirect income channel So incorporating those indirect income channel would be something super useful Second reason why it matters is for the result number three that the tax finance government spending multiplier is Sorry is larger than the debt finance government spending multiplier and here it kind of matters because The islm argument for the opposite result Is precisely very much based on the cajun cross on those on all those indirect income channels Obviously, there is a new channel here in the model, but precisely because they go in opposite direction That would be useful to have both In the same model to see how they interact and which one end up being the strongest Okay, comment number two are going to work backwards back to result number one on the Inflationary effect of qe and transfer. So the result here is that quantitative easing and transfer have a less inflationary effect than conventional interest rate policy and this comes in the model from the fact that The household that work are the saver households so that if for the same level of aggregate demand You have a transfer from the unconstrained to the constrained household The unconstrained household is going to feel poorer. It's going to be willing to work more This is going to put downward pressure on the wage on marginal cost and this is going to put deflationary pressure This is a meaningful channel, but Its sign may depend on the assumption that only the saver household Is actually working and is on its labor supply Curve so here I just generalize the labor supply schedule of the model to the assumption that both households work and they can work By different amount depending on their disutility of working and you can see that the new term Which is the one that depends on the consumption gap depends on the sign of this blue term here and this term can be either positive or negative depending on The differential disutility of work of the two here So if you want to have the result of the paper that a smaller consumption gap Decreases marginal cost it needs to be the case that savers Work a higher share of work than they consume That than their share in aggregate consumption Under the Q interpretation of the model that's reasonable But if we want to think of those constrained household as the hand to mouse household of a tank Maybe the opposite assumption is equally plausible So that would be useful to just try to tie that to empirical evidence on the Distance of this Income effect. So for instance the typical worry during the transfer in the u.s. During a covet was that those transfer to to households were Necessary, but that they could have put Upload pressure on wages because because they had had transfer they would be less willing to work So this is not empirical evidence is just a worry But it highlights that it's a theoretically possible channel. So it would be useful to have some empirical backing on the sign of this channel okay Comment number three I'm going to go back on the eviction effects. So there is this this cool new eviction effect in the model That implies that that finance fiscal policy Is actually fully neutral. This is in part due a bit to the formulation of the Full bailout from the unconstrained household, but mostly it comes from the leverage constraint of financial intermediaries So the story is the following Imagine i'm a constrained household and the government sends me a check of 1000 euro But it's going to finance that by issuing a new government debt Now the financial intermediary is going to have to buy these debts And because it is constrained he's going to have to reduce private lending by 1000 euros I am the private sector. I borrow so in the end for me. That's a complete wash. I don't increase my consumption That's Meaningful, but like the complete eviction result may depend quite a lot on the assumption that Public debt waits as much as private debt in the constraint of those financial intermediaries And that's a bit of a strong assumption and it can be I think easily relaxed For instance, just doing the same thing as in garter and charity in which they assume that Public debt does wait on the balance sheet of those financial intermediaries, but less than public debt In a garter and charity model it would be because it's easier to monitor The government for which you have a lot of information than a private lender And do glass diamonds yesterday made a similar kind of argument that it's not the same to To buy government debt and to finance the private economy By doing real banking activity Doing so should downsize the eviction effect that Cynthia and Yixi are emphasizing The empirical reason for maybe enlarging this is that empirically So we've seen obviously a massive increase in The federal debt in the u.s And also as Cynthia mentioned it in her talk In the size of public debt held by domestic By domestic also so here in blue I'm plotting federal debt minus what's held at the Fed and minus what is held abroad It did increase but in terms of In terms of intermediation spreads those are the ic bank of america spreads in red and green We didn't see such an increase in those spreads and they are basically back to where they were Before even though public debt is now higher. This is just the same thing for the euro area Let me just skip that to move to comment number four final comment Which i'm going to start from that from just an extreme version of the previous argument Um, which is that there is one view on this gertler karadii Channel of quantitative easing, which is that it only works in times of financial stress This is something that is a bit present in gertler karadii 2013 already Which is definitely very present in kurdi abut four 2011 And peter and enter never recent paper 2021 in which they make this this argument and look at this argument in much more detail The idea here is that in normal times the leverage constraints of financial Intermediaries are not necessarily binding so that you would go back to the qe irrelevance result For this paper that would mean that obviously quantitative easing would be neutral, but also some of the fiscal policy implication Would be Different depending on whether we are in financial stress time or not Now this is not the end of the argument though because even if this one channel of qe Doesn't work as much in normal times. It doesn't mean that other channels cannot be at play. So for instance qe could alternatively work through Uh, what dimitri just presented A model of portfolio balancing and direction extraction to be sure Dimitri's model is also a model about limited arbitrage, but it is quite different from a gertler karadii thing It's not that the central bank is stepping in and doing the job of financial intermediaries in its step Uh, alternatively qe in normal times could work through the signaling channel Obviously and since we are in the euro area It could also work in part by affecting self-fulfilling expectations of sovereign defaults. That's for instance Uh, a recent paper by bank of spain economists coasting union and thomas and here again That's quite a different channel. It's not that the central bank is taking risk away from the balance sheet of financial intermediaries It is that is just eliminating this risk altogether from the economy so I'm starting to be willing to You know put a bit too much into just two equations. So i'm going to stop here But you know even this slide is just a tribute to uh, The to the paper and the objective of cincha and yinxi in this paper Which is really to update this baseline model to uh, what has happened in the past 15 years? And I think that they are on a very good track to doing so Thanks a lot seven Okay for you. We try now to collect first questions from we try collect first questions from the audience like last Please Yes, but Please say your name first many people know you but maybe not all online. So yeah, thank you Jesper linda from the international monitor fund a very interesting Paper and the discussant also by Stefan. I thought You know, I just wanted to first sort of make one comment and two questions The comment is that I think it's sort of well known from wood for its baseline models say that Inflation output gap elasticity is the same for Monitoring fiscal policy, right you you if you take the woodfordian model your engineer You know more through the monitor policy or fiscal policy and alternative output gap path You're gonna get the same transmission on inflation. So that's gonna be the same But it's gonna differ for fiscal. No, if you go to the what you do shinto and Stefan talked about inflation output elasticity No, because fiscal policy also has effect on potential output So that's gonna lower the effectiveness Of fiscal relative to short term policy if you talk about the inflation Output elasticity. So I I in sense. I just wanted to I think it's your result is well known there The comment is then then you said on slide 32 That you're gonna have in your model that you do debt finance fiscal stimulus or Consolidation you have basically no inflation effects But I wanted to have your thought on how that relates to evidence No, because I mean in reality, of course all fiscal stimulus that we have measured I guess has debt effects No, and they do not have zero Zero multipliers. So I wanted to have your thoughts on that. Then I just wanted to have your thoughts on Because you have no endogenous labor if you allow for endogenous labor How do you think about the inequality effects of fiscal and monetary in in a case where you allow for endogenous labor? Thanks Leo Leo from Titan from the ECB I found this paper super interesting But given that we here are at a central bank. I have a very simple question namely what exactly If you bring your findings at some point or to optimal policy choice Choices what is the motivation of qe? And when do we need it my understanding is, you know, you stress this a couple of times qe has very strong distribution Implications, but for this alternatively you could also do it through something that is more conventional from a fiscal perspective Transcess for example while you said As long as you talk about inflation issues controllability of inflation you use as a conventional short-term rate Now my question now I think is the following. You have not mentioned the term lower bound I mean the way qe came into practice practice is because Many central banks had the feelings they were constrained by the lower bound So they had to think about something else but in these complicated circumstances then of course qe has a double task right because For most it had we needed to to stimulate inflation knowing that it may not be first best You would like to do it through the conventional channel while If you do not talk about the lower bound I would say we can essentially stick to standard assignments There is no need then to go so strictly for qe if you have you know fiscal policy makers do their job taking care of our distributional concerns central banks do their job controlling inflation I see no other questions so Cynthia if you can reconcile within two minutes that would be excellent. Yeah Thanks very much for the discussion and other questions So let me respond to Stefan's questions first So yes, we make some stark assumptions in order to reduce the system down and specifically for the assumption about The constraint household does not work So we have sort of like the way we answer in a paper is we put it in appendix where we assume the constraint household also work but also obviously that's given the Prematrization or calibration in our paper and in that sense quantitative result does not change now If we change the calibration as you mentioned it might change now The way we think about this like where we started is from the qe literature So essentially we model the constraint household as a shortcut to model the Firm sector without getting involved in capital accumulation investment So in that sense like we didn't model it were the the constraint household to work The the second question is so you are saying that the short-term interest rate does not affect the constraint household Now this assumption is actually coming from the market segmentation our model seeing our model the Unconstrained household it has access to the short-term interest rate Whereas the constraint household has access to those long-term bonds So this is how they break down like the short-term interest rate does not affect the constraint household now in principle Because of the yield curve there could be transmission from the short-term interest rate to the long-term interest rate through the yield curve and It could be the case in the simplified model that's shut down But in general there is a spillover, but it's just not as much as a Standard model you will imagine because those two interests are the same right the spillover the beta is a smaller than one From the short-term to the long-term interest rate, which is consistent with the empirical data So that's why there is a breakdown The next question you were saying is the inflation So the result on the inflation depends on the how we model the constraint household Right, so that goes back to whether the constraint household work or not again So we try to indirectly tackle this question by a follow-up paper by looking into empirically Is there an active sign to on the The the Phillips curve and what is the general equilibrium effect of the different types of policies right Is the conventional monetary policy work the same as QE as transfers in terms of the aggregating in terms of the supply side So we are trying to answer this question not from justifying like our assumption But trying to see if our results actually make any empirical sense consistent with the data um And your point about the government the debt and the private debt We yes in a big model you could model them differently and we thought about modeling them But just for simplicity we dropped them But I think we could add them and it might It might still be able to reduce to the equations we have but now with the delta floating around with eta Right, so I think that would be the change so the math should work out Yes, it's a stark assumption that because in the The difference between the government bond and the private bond in the girder karate framework as well as my in my previous Model in the medium scale the difference is quite mechanical It's just the scaling effect as you write down as a delta So we didn't think there's anything that's more intriguing there. That's why we drop it But just also for simplicity we could look into like Investigating more Oh one last comment You were saying for the leverage constraint only binds for the extreme periods but for not for a normal time Actually, I think that the reason we assume it binds for both times is if we think about in the study state, right on average The yield curve is spreading up Right, there's a positive spread between the long-term interest rate and the short-term interest rate on average in the data Right in this model If we assume the leverage constraint doesn't bind for normal times and if we only assume the zero lower bound is occasionally binding Then the spread will be zero So in order to make that be Empirically relevant to generate a positive slope of the yield curve on average We basically need an assumption of the leverage constraint to be binding now It could be something else could be other friction, right? But that's one friction generated slope So in our model, I think it makes sense to always have it binding and qi will always be relevant Now whether the central bank wants to implement qi during normal times that's a different question But it should be always relevant now going back to the question by jasper So you were mentioning about the relationship between inflation and output gap in this model the The coefficient between the inflation and marginal cost Is the same as in the standard model, but the elasticity between inflation and output is different for qi So we actually discussed that in in our restart paper And you were commenting on the debt financing whether the multiplier should be zero literally now in general It's not zero, right? And we all we did that quantitative exercise by Relaxing some assumptions, but in general in Even we relax all the additional assumptions we make to reduce the system down The key takeaway is that financing is less Stimulative compared to tax finance in our models that is consistent now whether it's literally zero I don't think so like once you take away some of the assumptions And also this result is consistent with the empirical findings of the paper One of the paper I cited the micro study paper And she actually finds that the tax finances has a smaller Output multiplier. So that's consistent now whether it's a literal zero It's a it's a like artificial fact of the small scale model Induction stable, I think I talked about that and the question from leo the optimal choice So both the optimal we discussed the optimal policy in a paper I just didn't have time to mention it. So in optimal policy, we derive the welfare function It's depending on not just the output gap and inflation variance, but also depends on Cross sectional consumption variance So it does depend on this welfare function of the distribution between the different kind of households And so we talk about that we can talk more about the results and zero lower bound we do the zero lower bound in the restart paper and precisely because like qe is Zero a result of the zero lower bound now. It just involves more math and the Qualitative result is the same except for we cannot implement conventional monetary policies So we have to do qe to counter some part of it. So that it That's why we didn't do it in this paper, but we did some of the analysis the end of the paper Excellent. So please join me in thanking again Dimitri Andrea Cynthia and Stefan for this excellent seminar. Thank you now We are going to bridge science and practice even more intensely in almost Only 10 minutes in the policy panel. So we're looking forward to it. The coffee break is very short So please be back in time at quarter to 12. Thank you