 So Charles you will say a few words, ten minutes for the panel, followed by Jim, Christine and finally Philippe. Okay, well me too I want to thank the organizers for inviting me. It's a great honor and a great pleasure as well. Now our chairman put to me three questions following up on the previous session. What are the sources of low inflation? What are the implications of persistently low inflation for macroeconomic performance in the euro area? And what are the implications for the monetary transmission mechanism? So I'll try to deal with these questions in my own way. Now I think this morning we had great presentations that built up on a very large literature on why we have had such stubbornly low inflation. And my reading of my understanding of this literature and what we heard this morning is that the general agreement that cyclical conditions, so the slope of the Phillips curve, matter less and less for inflation, but they matter. We'll listen later to Christine, who is highly up more evidence that foreign, what's happening abroad elsewhere in the world, or exogenous things that are not Phillips curve, matter more and more. And we had the Yuri's presentation, which to put in a very simple way, tell us that normal people don't care about monetary policy. It's a shortcut, but that's what I got out of it. And that is important and interesting. Now I've put here the inflation performance, you know these curves by heart. It's annual inflation in blue, it's the euro area, and in red it's the country where I live, but which is interesting because it has had stubbornly low inflation. Actually Switzerland had had negative inflation for much of the last eight years. And we all have been trained to think of deflation, negative inflation, deflation as a total disaster. And I see Governor Kurota seeing what I'm talking about. The experience of Switzerland is exactly like Yuri, people just don't care. People don't know that inflation is negative, they don't think it's a problem at all, and they're happy to live with it forever and have many children at the same time. So this sort of goes very much in this line. Now what I'm just looking at the data, but there's a lot more formal work, is that inflation was stabilized before the recession, well the great crisis around the slightly above the target for the ECB, and smacking the middle of the range for the Swiss National Bank, and then boom there was the change of regime, the crisis came, Chop pushed everything down by a big notch, and since then it's been fluctuating on average significantly below what it was before, and that's sort of the external shock that comes in, and that is very, very powerful. So it does, intuitively the interpretations above seem to do a good job in explaining. So the conclusion, and I'm going to paraphrase what I said earlier, I think by Lucrezvia, the Phillips curve is alive, but it's not kicking very much. And that I think is what we are witnessing. So the theory is that inflation should be driven by expected inflation, by the small cap X, which is the output gap or any measure of activity we've seen Jim presenting this morning, and then the capital X is a bunch of external things, or exogenous things, everybody has his favorite list, and there is a big literature estimating these things, and as Christine will explain, yes, that's happening, and beta is stronger than alpha, and increasingly over time. Now, the problem is that even when you take account all of that, this kind of equation generally don't do a good job empirically, and what seems to be doing a better job empirically, is the approximating expected inflation with past inflation. That's sort of what I think the literature has showing. Now, so if that works, that's the long surviving Phillips curve. The output gap or any activity measure has an increasingly weak and unstable effect. What's happening elsewhere or what is exogenous is beyond reach of central banks, and if the past is what determines inflation, not expected inflation, then we are stuck with the past. And so that's, again, looking at the data I showed before, there was a big negative X, sometimes around 2008-09, we just got down, and the effect of this cyclical condition is weak, and as long as X is zero, all that's happening is inflation. Today is inflation yesterday or last year, and there are little bubbles here and there, but not much is happening, and that's sort of my simple-minded summary of why inflation has been low and is staying low. Now, what are the implications? So the question was what is the implications of persistent low inflation on macroeconomic conditions in the euro area, that was. So it goes back a little bit to what Larry was talking yesterday about monetary policy effects on real variables. Here the question is the impact of inflation, low inflation on economic performance, and again my understanding of what we seem to be agreeing upon is that if it's stable, so when inflation is stable and has stabilized and remains stabilized and expected to remain stabilized, then it doesn't have any real effect. So it's a variation of the neutrality. When it moves and it kicks, we discuss that abundantly it has effect, but when it stays at a low level, it doesn't have much on effect or on effect on the real life growth, employment, whatever. Now, of course, one exception of that is that if central banks try to deal with too low inflation given their target and they start doing all the things that they can do, then they do it more and more of it because the Phillips curve effect is weak. So if they want to raise inflation, they have to do big time-monetary policy, then things can happen. And of course the role of fiscal policy is also in the background. I think we've seen fiscal policy pretty weak until recently in the U.S., but certainly in the U.R. I don't think much has happened. So there isn't much happening on the real side except maybe of the effects of the very explanatory monetary policy. So all the discussions about distortions, risk on the financial market and so on and so forth. That may be an issue and everybody is aware of that. Now, there is an interesting question of the exchange rate, so it's part of the, in a way, it's what's happening abroad, or it's our relationship with the rest of the world. I think the evidence here is that the effect is increasingly muted, but still it's this permanent temptation, the beggarly neighbor temptation. Now, one of the things that we have observed, at least among the developed countries, is that the temptation to play around with the exchange rate has been largely resisted. They have been movement in the exchange rates, but they were not really attempts by the authorities to play with that and have an X, the capital X, that helped achieve the inflation target. So that is, I think, is important for our understanding of monetary policy at the global level, or at least among the developed countries, assuming they can do something about the exchange rate, which wasn't discussed this morning, and that the exchange rate can have powerful effect on inflation, although there is evidence it's becoming also less powerful. There has been some implicit or explicit discussions among the saddlebankers here, and it has avoided possibly negative, bigger-than-neighbor policy. Now, there is a very special case of the Euro area member countries. They don't have exchange rates, so that, in a way, backfires because there is hardly anything they can do if they would like to do. They cannot have monetary policies that are bigger than neighbor, but there are fiscal policies that could be used, and this hasn't been used, but it's a permanent source of tension. So this aspect is one of the underlying consequences of this low inflation. Now, finally, I was asked implications of monetary policy on the transmission mechanism, but I didn't think that was an interesting question. I wanted to zoom directly to monetary policy. So my understanding again of what we have seen and much research is that inflation targeting remains the best strategy. There hasn't been much of a challenge to that. A number of people like to say it hasn't worked, but I don't think it's true. There is this sort of bubbling return of a suggestion to have nominal GDP targeting. So it comes now and then over the last 15 years, and I think it's a smart idea, but it raises more questions than it answers of, first, how do you measure GDP, how precise, how soon you know it, and so on and so forth. If this morning we were told that the consumer price index is badly measured in real time, suddenly measuring nominal GDP is open to 10 times more difficulty, and I think that's a good reason why it's likely we should stick with inflation targeting. Now, the interest rate instrument remains the logical instrument in the history of the ECB. It's not an obvious conclusion. Early on there were deep debate about what should be done. I think the evidence is that it remains the right instrument except that there is a lower bound and except that we've hit the lower bound and we are at the lower bound. Now the answer is QE, and I think that's the proper thing to do. There is the Japanese fine tuning on shaping the yield curve when you are at least zero lower bound. So my impression, and that's also what I think I heard from Larry Summers yesterday, we are more or less in terms of the fundamentals where we were before the crisis. There has been, of course, a lot of critical discussions, but at the end of the day, the things are okay. There is this new instruments of QE and possibly a long-term, affecting the long-term interest rate directly, not just through expectation, and that's probably the best we can have. Now is everything fine? Certainly not, and there are good reasons to worry. And I want to take a particular line on top of all the things we've heard, including the dangerous impact of very low interest rate for very long and bubbles and all the other things which are well-known. Just following on what we heard this morning, what is clear is hitting the inflation targeting has proven much more difficult since the crisis than we thought. Here again, I bring back this data, so the above one are the U.S. and Switzerland. The other one is in grey, the U.S. and in green, the U.K. With the target, well, Switzerland has a zero-two target. Now, what we see from the U.S. and the U.K. is assuming the target is 2%, which wasn't in the Fed, and there was no target, but assuming that the inflation target is 2%. What we have seen is in contrast to the ECB and the Swiss National Bank, in the U.K., the average inflation over the whole period has indeed been 2%, so success except that the variations from year to year are not considerable and just measuring the standard deviation is about 1%. So if you start asking how often are the central banks likely to be near the target, the answer is not very often because there is just so much variability. So now that matter, that matter because precision is, if the precision of the things that central banks can do are very imprecise, even in the medium run, because we never get in the medium run as we know, then there is a question of how central banks should go about doing the inflation targeting for these reasons and also because of the financial stability mandate, which is one of the other innovations in most countries that central banks have sometimes to forget about the financial inflation, price stability mandate and worry about financial stability and do all sorts of things. So all of that in my mind is not challenging the inflation target strategy, but it is challenging the way we have been going out of it. Now the natural implication is that if we are heading to try a target with a very imprecise instrument, it would make sense to have a wider margin and so that's an ongoing discussion. Yesterday Larry said because of the very low real interest rates, you want to move your margin, that's not the argument. My impression is that by trying to stay within the margin that it is impossible to stay within, central banks create trouble for themselves. It's not just a question of credibility, it's a question of constraint on policy deliberation and the way it's acting. Now the second thing and it comes directly from this discussion this morning, that I mean Yuri's paper makes the important point that firms and consumers respond to information when they are given the information, but they don't listen to central banks, they don't care, they don't even know what is the inflation rate. Now to me the way to improve precision is to actually talk to these people in their language, which is again what Yuri developed this morning. And here I would like to try to do a little bit of a provocation, which is time. Okay, I don't have time to do my provocation. Minus seven. Okay, just very quickly. My perception is that again what Yuri said, talk a simple language. Now central bankers talk a very complex, sophisticated, convoluted language because they talk to financial markets. And as a first approximation financial markets don't matter beyond the very short run. I know there is this transmission thing, they don't really matter and I would submit that it would be a great progress in monetary policy if central bankers would stop talking to financial markets, let them take the risk and probably reduce the risk taking because they would have less spoon feeding and talk way more to the people in the streets who mistrust the central banks, who don't understand anything, but who matter for inflation. And that's the end of my provocation. Okay, thanks very much. It's a pleasure to be here. I'm a big supporter of this conference. I think it plays a very important role in bringing together a lot of people who think about monetary policy in exchange. I titled this the case of the disappearing Phillips Curve, but we could also do this as a murder mystery who killed the Phillips Curve and I'm afraid the suspects are in this room. So the slope of the Phillips Curve was negative in the 80s but has been drifting towards zero in the inflation targeting era, which I'm going to date from 1995. That's an approximate date, but the inflation hit 2% in the U.S. and I would say implicitly the Greenspan Fed adopted a 2% inflation target. We often call this a flattening Phillips Curve. We've already referred to it basically in all the comments today. Now, at the same time, monetary authorities have improved monetary policy during the inflation targeting era and I'm not going to provide direct evidence of that, but I'll just state that inflation has been lower, closer to target. The variability of inflation has been lower, the variability of inflation expectations has been lower in the countries that have inflation targeting. So that's what I'm calling the inflation targeting era and that's in stark contrast to the earlier pre-inflation targeting era, the pre-1995 era. I'm going to argue that these two phenomena are linked, that the improved monetary policy in this sense has led to the flatter Phillips Curve and I'll draw up the implications for monetary policy after making my core argument and I'm going to talk in very dense, complicated central banking language. The empirical evidence has been shown already in several different ways, but I like this particular presentation, which I'm taking from the BIS Annual Report 2017. Data is for a panel of G7 economies. The coefficient is estimated, the coefficient, the key coefficient, the Phillips Curve coefficient in a preferred specification is estimated for rolling 15-year samples from the 1980s to the present. So we're going to see how this coefficient evolves in the picture from the 1980s to the present and the point estimate will be the weighted average across economies GDP weighted. So this is the picture you get in the 80s. You might have had a coefficient here on the order of minus 2 in this regression and as you move forward with the rolling samples, it gets to be minus 1 in about 1990 and it goes all the way to zero at the end of the sample and even the point estimate is slightly positive at the very end of the sample. So I'm calling this the disappearing Phillips Curve over the period since 1980. So let's write down a model and I'm going to use a simple and standard model. It's probably not totally convincing because it's too simple but it is a version of more complicated models that underlie almost all of modern analysis of central banking and it's the beloved three-equation New Keynesian model with an IS equation number one, New Keynesian curve number two and monetary policy conducted using a Taylor-type monetary rule number three. Why is the output gap high as an inflation gap? Why is the policy instrument rose the real interest rate which is subject to a shock? There are two shocks here, the Epsilon and the U. There are some structural progress, Sigma Kappa Beta which are all positive and critically there are two policy parameters, the Phi Pi and the Phi Y which are the policy matrices in the policy rule. So if you figure out the rational expectations equilibrium of this you find out the inflation gap evolve according to a linear function of the shocks in the model, the Epsilon and the U are in the numerators in these expressions but also that the policy parameters enter these expressions. So now let's turn to monetary policy and we're going to try to frame something that gets at this idea better and better monetary policy over time. So I'm going to look for monetary policy but I'm going to do so in a special way. There are many ways to talk about optimal policy in this model. I'm going to do one particular way. I'm going to constrain the policy maker to stick to the Taylor-type rule that I've given the policy maker and the policy maker gets to fix any Phi Y that they want and then we're going to minimize a quadratic function here, Equation 6, by choosing just one parameter Phi Pi. And you notice that this minimization has an alpha in it. The alpha is the relative weight in the objective function on inflation stabilization but it turns out if you solve the problem this way it doesn't really matter what the alpha is. The solution is actually to send Phi Pi to a larger and larger number which technically to go all the way to infinity. So that's going to be very useful in the next slide here. So optimal policy would be interpreted as that policy makers should promise to react very aggressively to deviations of inflation from target in conducting monetary policy. I'm going to use that as a summary statement of what has happened since the 1980s in the G7 economies and broader across the OECD economies that policy makers have become more and more obsessed and more and more aggressive in responding to deviations of inflation from target. And you might say, gee, Jim, this doesn't include unconventional policy but I want to think of the unconventional policies all encapsulated here in that policy makers were willing to do extraordinary things when the crisis hit to make sure that inflation did not go too negative and did not go too far from the target. So there again I think this captures the spirit of a more and more aggressive monetary policy, inflation targeting policy over time. Okay, so now I'm going to put on my nutrition hat. I'll be Jim Stock here for a minute and run in regret of the inflation gap on the output gap inside the model, and we'll call that estimated coefficient the slope of the empirical Phillips curve. And because this model is so simple, this empirical Phillips curve slope is going to be able to be calculated directly as this formula number seven here. And if you look at this formula a little bit, you see these structural parameters, Kappa Sigma. You see the variances of the shock, Sigma Epsilon and Sigma squared U. But you also see the policy parameter, Phi Pi in the denominator. And in fact, if you take the limit of the right-hand side of this, you will get zero as Phi Pi goes to infinity. So as I've defined monetary policy, Phi Pi is going to infinity. As Phi Pi goes to infinity, the slope of the Phillips curve literally goes to zero in this model. So this is the main point I wanted to make. Now, is this empirically relevant, you might say? So this is a very simple model, a stylized model. What happens if we go to a more serious model? So we did something, we did two things about this. One is this Lubick-Schorfeide estimates, which are from 2004. So that's before the crisis. They estimated a model very similar to this. And we took their numbers and used Okon's law to get to the unemployment version. And if you do that with their estimates, you'll get, and let Phi Pi get larger and larger, you will get this picture, which is very similar to the one I was showing you in the beginning. As Phi Pi gets larger and larger, the slope coefficient goes toward zero. In this model, with Phi Pi equal to one or less than one, you'd have a very negative slope of the Phillips curve. Similar to the one we might have seen circa 1990 or 1988 or 1985. But as Phi Pi got higher and higher, this would go toward zero. So this is one way to say that maybe this is empirically relevant. There are other, I'm certainly not the first person to make this point, and this has actually been hinted at by people in the audience already today. But I'll just do a very partial review of a few pieces of literature that would support this point. This has been published literature over the last, well, you could say since Lucas, but certainly over the last decade or more. Bovina Giannani, for instance. This paper, Del Negro Giannani and Sharfaiti actually tried to address the missing deflation after the crisis in a DSG model. They made an argument that's very similar to the one I'm making here. And then this last paper, McClay and Tenereno wrote, they actually do, this is a very recent paper, this last one's all promoted here a little bit. They do a better job than I'm doing in these slides of analyzing this more extensively and pointing out the identification problem that would arise from all kinds of versions of this model and this kind of exercise here. So I would recommend, if you're interested in this exercise, I would recommend that paper. So they would say that Phillips Curve can't be easily identified in the data because of the problem that I'm outlining here, the Lucas critique kind of problem. So what should we do as policy makers? This is my last slide. I think we have to look for a different signal. Ultimately, successful monetary policy can push the empirical Phillips Curve slope all the way to zero, according to this analysis. Of course, the structural Phillips Curve is still there. The friction is still in the model. It's the empirical Phillips Curve that's disappearing. But what that means is that as policy makers, we probably can't take a signal, a reliable signal based on empirical Phillips Curve relationships that we estimate. So we're going to have to look elsewhere in order to get a signal for monetary policy. Thanks very much. Thank you, Jim. Can I get the first polling question from the technicians there? I hope you use the application for the answers. It's a bit more difficult this year. You get the question? Yeah. Can the Phillips Curve, i.e. the relation between slack and inflation, be described as alive and well in advancing economies today? Press number one, if you think that yes, but slack and or inflation is not measured accurately enough. Press two, if yes, but it is not linear, and hence inflation reacts with the leg when slack is large. Press three, if no, it is flatter or weaker than in the past. And number four, no, traditional Phillips Curve is dead. The fluctuation in inflation can largely be attributed to factors other than domestic slack, such as external factors, shifts in expectations, sound spots. So, Cast your vote. You have 15 seconds. I think many didn't download the application. Yeah, if you don't have the app, you can download it from the App Store or Google Play, or if you need some help, you can reach us at the booth downstairs. So, can you see the results? Here's the results. There it is. Mixed Christina. I think it's time to go. Well, thank you very much. Actually, I will argue that more than one of those answers is correct. So, that's probably why it's been hard to get consensus. Could I have my slides up at some point? Thank you. So, it's a pleasure to be here, and yes, Peter, I know that the topic of the panel is not mariner instruments, but I thought I'd, given that we've had a long intense morning, come at the questions we're looking at from a slightly different angle. So, hopefully you have had some time to do some sightseeing in the area. If so, you've probably learned about Prince Henry of Portugal, also known as Prince Henry the Navigator. He's one of the great mariners credited for the age of discoveries in Portugal, and he, as many of the Portuguese mariners, relied heavily on the Astrolab to get assessed latitude when they were sailing and how far their ships had come. So, imagine if your Prince Henry in your Astrolab starts to malfunction, and it consistently over-predicts how far your ship has gone. Somewhat like central bankers, all of our models are consistently over-predicting the return of inflation. What should Prince Henry do? Should he get rid of this instrument, get rid of our inflation models, and go for a new model or a new instrument to predict where his ships have come? Or should he give up on instruments, give up on models, as some have suggested, just sail by his gut, by his landmarks, by the wind? Some people suggested maybe central bankers should rely less on our models. Or third, should Prince Henry try to fix the Astrolab, tweak it, make it work better, figure out why it's not working well? Similarly, should we try to fix our inflation models? And that's going to be my point today. I don't think we should throw out our basic inflation models. I think there is still some grain of parts working there, but we do need to adjust them and fix them to make them work better. And I'm going to argue at least one of the key factors that's missing that will make them work better is to better account for globalization and changes in the global economy. It is amazing how domestically focused our models of inflation still are given the substantial increases in global integration that have happened. But before I get to that and before I make those arguments, I want to just take a step back and think about whether inflation really is too low. That's a key premise of the panel today, key premise of these two days here, that inflation is lower than it should be given the stage of the business cycle and the falls in unemployment and closing about put gaps around the world. But of course that assumption, inflation is too low, comes from our models that we're all questioning how accurate they are. So I thought it's useful just before I get into how to fix these models or how we can improve on them, and really how low is inflation? In there I found it useful when I struggled with these questions at the Bank of England is a cross-track to do trend cycle analysis of inflation. This is a way to take a step back, Lucretia introduced us to this this morning, take a step back, don't make assumptions about how you measure slack, lag structure, the whole set of assumptions that go into these models. Instead just look at the statistical properties of inflation and separate inflation into two components. A slow-moving and persistent trend, which would be the target of central banks in monetary policy, and then the temporary cyclical movements around the trend that you can largely ignore, but can drive headline inflation rates and even core inflation rates. Like this framework, minimal assumptions needed, no need to parameterize, and it allows the models to automatically be flexible over time. It's a few different ways to do this. I'm going to use a technique I developed at the Bank of England with two colleagues there, called an ARSP model, and I'm going to use models built on the seminal work by Stock and Watson, which use a UCSV model to basically do the statistical decomposition. I'm going to basically use the Stock and Watson model, but add an autoregressive component in the error term, which is something Steve Cicetti and others have suggested doing. So no time to go into the technical details, but basically break down inflation into a slow-moving trend and then a shorter-term cyclical movement. That's what I do. I'm going to start with the two countries where inflation is about at target or above target, the U.S. and U.K. And I'm going to show you a bunch of these graphs that are all set up the same way. Black is headline or CPI inflation or core inflation in the U.S., headline inflation for the others, closer to your targets. This is quarterly inflation annualized and seasonally adjusted. The blue is the slow-moving trend and the red is the cyclical component. These graphs are interesting to look at. There's a lot in there about what drives inflation, how much is slow-moving trend, how much is cyclical. But given the time constraints, I want to focus on what does this suggest about where trend inflation is relative to targets. Especially given that headline inflation is picking up in a number of countries. How much of that is underlined inflation and how much is more temporary effects of oil prices. So what you see for the U.S. is that underlying core inflation or trend inflation is pretty close to the target. My sample ends in the fourth quarter of 2017 and it's 1.9%. It's gotten up a bit since then. You see that some of the weakness in inflation in the U.S. in 2017 that got a lot of attention is identified as just cyclical. In this, UK is one of the few countries, actually one of the only countries in my sample where underlying inflation is actually well above the 2% targets. It's at 2.7% at the end of 2017. So these are countries where underlying inflation is basically at or a bit above inflation targets. Now let's go to the rest of the world. Low inflation is still an issue. So let's start with in honor of the president of the ECB, Italy and in honor of where we are, Portugal. What does this say about trend inflation? These graphs are pretty typical of most countries in the Euro area. You see big cyclical drags on inflation during the global financial crisis with concerns about Euro debt problems. But overall you see this pretty steady downward shift in underlying trend inflation. So trend inflation at the end of 2017 is still well below the 2% inflation target. If you look at some of the other economies in the Euro area say France and Germany, you see again headline inflation is picking up CPI inflation is picking up but according to this decomposition a good part of that is cyclical is the red. Underlying trend the more permanent component of inflation is still quite a bit below the 2% inflation target. Moving outside the Euro area do the same thing for other countries you see some very different experiences in other advanced economies. So here I picked two countries where unemployment is probably at or below Nehru, pretty solid growth but yet countries that have struggled with getting inflation up. Sweden inflation has picked up has recently been about 2%. But some of that pickup is underlying trend inflation but still got a ways to go to get to 2%. Japan if you squint you see that that trend inflation is coming up but there's still a long way to go to get it up to a level consistent with 2% inflation. I could show you a lot of these graphs but I think that you get the point. Underlying trend inflation in most countries around the world is still quite low. So people organizing this conference you might have been worried that now headline inflation is getting up there. It looks like we don't have a problem more in a lot of the world but if you look at these decompositions a good part of that pickup is cyclical and there still is a reason for monetary policy accommodation in a lot of the world. Underlying trend inflation has still not come back to levels you'd like to see consistent with target. So what's going on? So there is a disconnect. There is a disconnect between pretty solid growth and unemployment rates, output gaps getting close to closed or closed in many economies and this low underlying trend inflation. So we've heard a number of stories about what's going on. Is it how we measure inflation, how we measure slack inflation expectations we just heard about credibility of central banks. There's been some nice work at the BIS suggesting global slack should be included. My read of all this literature is that there is something to each of these arguments. I think each of these factors has played some role is important. Each of these does help improve our models and understanding of this underperformance of inflation. But what I think is also missing which is more and more important over time is a better incorporation of changes in the global economy. It really is amazing when you think about how much globalization has proceeded since the Phillips curve was developed, how integrated the global economy has become and yet how little we make any adjustment for that in our basic Phillips curve models. We look at the models we've looked at today with due respect to the authors who presented they did what's typical in a lot of the literature. There's no control for what's going on in the rest of the world. Some estimates that are a bit more progressive have one variable which is oil prices or import prices is their way to capture what's going on in the rest of the world. And there's this argument in the literature that if you control for domestic slack and say oil prices or import prices that should be a sufficient statistic to capture changes in the global economy and how that affects inflation. But my read of the evidence is that there have been more fundamental changes. We need to better think about what's going on in the rest of the world when modeling domestic inflation. So for example trade flows have increased so if a country imports more then imports are larger share of the CPI index are just mechanically changes in exchange rates or imported prices will have a bigger impact on inflation. If countries export more then changes in global demand are going to have more of an impact on how firms set prices. Emerging markets play a much greater role in global growth and global demand. They're driving sharper volatility in commodity prices which could affect pricing especially if effects are non-linear. There's also quite a bit of work on the greater use of supply chains. Much easier to shift parts of production to where it can be done more cheaply and that's going to affect bargaining power of local workers for example. So just a couple examples but there have been some pretty fundamental changes in the global economy that are going to affect inflation and we really are not working them into our models well. So what happens if you do put some of these factors into just our simple models of inflation. So I'm going to show you a few pieces of evidence that show that adding them will help improve our models not solve everything but go some way. So first I'm just going to estimate a simple Phillips curve with all of the problems and I'm going to estimate an unquarterly inflation for a pooled sample of countries about 40 countries 1970 to 2017 for CPI inflation and core inflation and I'm going to include the standard Phillips curve variables at the top inflation expectations, lagged inflation in the domestic output gap and I measure the domestic output gap is a principal component of all sorts of different measures that could go into Slack to try to get around the issues with that measurement. And what you see at the top of this the coefficients all come in positive and significant which is what you'd expect higher inflation expectations higher lagged inflation larger positive domestic output gap correlated with higher inflation. So standard Phillips curve variables still work in a pooled cross section of countries over time. But then I also include a set of global variables or global factors the exchange rate, world output gap, oil prices, commodity prices, energy and a measure of PPI dispersion to capture pricing pressure and supply chains. Each of those coefficients also comes in significant in just about all cases with the expected sign. So exchange rate depreciations larger world output gap higher world oil prices, higher world commodity prices and more dispersed producer prices are all correlated with higher inflation. So this is just one piece of evidence these global variables do seem to matter at least in a pooled cross section of countries. They do significantly improve the adjusted r-squares. But if you estimate these for separate countries one caveat, the model doesn't look quite so good. If you estimate the same equations, just one country at a time rarely do you get these great results where everything comes in with the right sign insignificant. Instead in most countries you get three or four of the different variables coming in with the expected sign insignificant. Usually you get at least one of the global variables but usually not all. But when you look at the differences across countries it is pretty intuitive. For example, if you do this just for Germany you find that inflation expectations lagged inflation in the world output gap are all significantly correlated with inflation. If you do it for Iceland you get totally different results. None of the variables important for Germany are consistently significant but instead for Iceland the domestic output gap, oil prices and the exchange rate are the ones that are significantly correlated with inflation. So you can do this country by country. You get pretty intuitive results. Different variables matter in terms of determining inflation in different countries but that also does make it hard to generalize across countries and results. But overall global variables do matter in most countries around the world and improve the explanatory power of the aggressions. It's just different variables matter more for different countries. Will that put out matters more for Germany, exchange rates and oil prices? So that's one challenge in estimating these models. The other challenge which we actually just saw in the last presentation is the coefficients on these variables can change quite a bit over time. The role of these global factors as well as domestic factors do vary. I'm going to show you a couple of pictures to make this point but this also is why estimating these models can be hard. What I'm going to do is estimate the same model same cross section of countries 1990 to 2017 but now allow the coefficients to roll over time, over eight year rolling windows. And then I'm going to take each of these coefficient estimates and look at what happens to it for the median country in the sample. And let's do that say for the real exchange rate. So this is the coefficient on the real exchange rate. First half of the sample coefficient averages about minus 0.05. So what that implies for the median country a 10% depreciation is correlated with a 0.5% increase in CPI inflation over the next two years. So that's sort of logical fits with estimates of past through. Suggest exchange rate movements are important in affecting inflation in many countries around the world. But then you look at what happens around the time of the global financial crisis. The coefficient goes from as expected negative to basically 0 and then briefly positive and now it's coming back closer to the sort of pre-crisis average. This changes in the effect of exchange rates on inflation could be explained by different shocks causing the exchange rate movement. For example, if the exchange rate is moving more due to demand shocks during the crisis that would cause less pass through. But the bottom line is the effect of these global factors does vary over time. Here's another one to show you how the effect varies over time. The impact of commodity price, X energy movements on inflation. Movements in commodity prices didn't seem to have much effect on inflation outside of oil before the global financial crisis. Since then movements in commodity prices seem to be having a much bigger effect on inflation in countries around the world. Especially CPI inflation. This might be because commodity price movements are more correlated with demand in emerging markets. So this could capture global demand effect. But the bottom line is this is a global factor now having much more effect on inflation that didn't matter in the past. So it understood why it wasn't included in models. We've talked about the decline of the Phillips curve. So I also, here's the coefficient on domestic slack when you do these rolling regressions. Get similar results to Jim where the coefficient seems to become less important. Well, you get the expected positive coefficient, larger positive output gap correlated with higher inflation pre-crisis. Then it falls to just about zero. But it is starting to come back at the end of 2017. So maybe the Phillips curve is re-emerging which would be consistent with theories that there are non-linearities. So bottom line of all of this adding global variables to models trying to explain inflation do seem to be important. Global variables do seem to play an important role. But their effect varies over time making it hard to measure in why you can get some different results and different papers on whether these global variables matter. Their weight does change over time. So as the last graph I'll show you is how important are they? Okay, so I can show you nice significant coefficients, lots of stars, it suggests that they're significant, but are they meaningful? How much will adding these global variables actually reduce the errors in our inflation models? So as a last test for that, what I did is I took the same rolling coefficients of this model where I have the standard domestic variables plus the five extended global variables, estimate the model for each country trying to predict inflation, quarterly inflation again using rolling regression coefficients so you allow the variables or affected the variables to change over time. Then you take the predicted coefficient for each country plug in the actual values for each country and then look at what predicted inflation is using this model and then look at the deviation of predicted inflation from actual inflation. And I'm going to do that once with a full model with all these variables in there and then look at the gap between actual inflation and predicted inflation if you just include the domestic variables don't include the global variables. So it's sort of a rough test of basically what's the error from the model with and without global variables. And if you do that this is the squared error terms you get a graph like this and again what you want is these numbers to be smaller. Close to zero means the model is perfect that it exactly predicts actual inflation. Higher up means your model is not working terribly well and you're getting some bigger errors. And the black is what the model predicts when you only use your standard Phillips curve variables, your domestic variables the red is the errors when you include your global variables. And what you see is adding the global variables isn't going to make the models perfect there's still some misses you still get some numbers above zero but it does reduce the errors by a meaningful amount and it particularly reduces the errors from the crisis and most recently from about 2012 to 2016. And this is an error where the global variables do seem to play a more important role and really have affected the inflation process in a more meaningful way. So adding them to the inflation models it's not going to make them perfect but it does meaningfully reduce the errors you get. Last Peter had asked so which global variables matter I wish I could have said given showed you this graph and said one, just add the world output gap just add the exchange rate and you'll close this gap between the black lines and the red lines unfortunately it's not that simple different global variables seem to matter more for different countries so you really do need a more comprehensive set of global controls there's no one magic bullet you need to incorporate different aspects of globalization and global integration but I will skip the regression results in the interest of time and just sum up what comes through when you do this more formally is Jim said, tech it up with more formal regression analysis the results do suggest that global factors should be included more comprehensively in inflation models but you also do need to allow parameters to evolve over time many of the global variables were not as important before 2009 more recently they're playing a bigger role which factors are most important exchange rates do seem to matter across the whole period in predicting CPI or core inflation over the last decade commodity prices and global slack seems to matter quite a bit more and particularly improves the explanatory of models explaining CPI inflation domestic slack seems to be less important in predicting inflation especially in advanced economies with their own currencies the bottom line back to Prince Henry in his Astrolab I'm in the camp of don't throw out the Phillips curve models don't throw out our old models we need to adjust them we should, it is past time given the globalization that has occurred to explicitly and comprehensively incorporate global variables global factors into our inflation models similarly the Portuguese supposedly their innovation to make their Astrolabs work better was to find stable land or at least a very calm day and then their Astrolabs worked a lot better it seems like a small innovation seems pretty obvious in latitude in similar I think this adaptation to our Phillips curve type models adding global variables not a huge innovation but it can significantly improve their accuracy thank you as the last speaker I predicted to be a lot of slides already produced so I decided no slides and also means I can maybe respond a little bit to some of the previous contributions but my order from Peter actually talk about how central banks should respond because there might be temptation listening to some of what we heard this morning basically to be a bit nihilistic saying who the hell knows inflation is low we can't find factors that that would help to track inflation and so on and I think we could have had a lot of these this discussion a few years ago so if we had this we went back to central 2014 for example maybe a lot of this discussion would be pretty much the same but then of course between 2014 and now and I'll just talk about the ECB maybe Jim can come back and how the Fed has responded in the last four years is action was taken and it'd be interesting to go back and rerun your online survey and ask how optimistic are you in summer 2014 about the ability of the ECB to deliver policies that can bring inflation to use a phrase I've heard before on a sustained path to the inflation target and what we've seen over those last four years is that some of the messages we had this morning from Yuri and others which is and I think you had as well like aggressive action if you take aggressive action if you roll out the package of measures whether it's going below the zero lower bound, whether it's the APP, whether it's forward guidance we now four years later do have evidence that that package does help so if your initial condition is a lot of slack far away from what you might consider a busy European economy and if you can see that some of the obvious transmission mechanisms are far away from what might be considered normal so funding conditions are high if lending rates are high compared to what you might expect then having an effective monetary policy can do quite a bit and what we've seen over those years is a big change of financial conditions so unlike Charles I do think no matter what for me think of the financial system the transmission mechanism does work through financial conditions so communicating clearly and persistently about the effect of these policies explaining to the market is very important in moving the funding costs that matter for financial intermediaries so I do think we now have a lot of evidence from that that we saw a big decline in financial conditions I agree with Kristen that the role of factors changes over time so in 2014 early the first quarter 2015 the exchange rate moved a lot then for a couple of years the exchange rate did nothing it was pretty much flat so there was a big initial response from the external sector but then that provided time for domestic parts of the economy to take over and what we've seen by and large since then is a broad balanced recovery where the decline in lending rates has allowed private consumption to take over and more recently business investment and even more recently household investment in terms of construction and so on so the transmission that monetary policy is effective in moving financial conditions in turn financial conditions can move the real economy and then increasingly we're confident that it also moves the distribution of inflation and all of that is lining up so I went back last night and I looked at Peter's talk to the ECB watches conference and one summary measure that Peter talked about there was if you put all of that together and if you think about the cumulative impact over a core part of the program from 2016 to 2020 it's adding 1.9 percentage points to your area GDP and 1.9 percentage points to the cumulative inflation part so this is saying central banks are not ineffective it's not just cases being able to move the real economy as Larry Summers said last night it's also a case that the distribution inflation has noticeably shifted I can very much remember in summer 2014 going to conferences where the belief that that deflation risk was actually going to take over in the era area that that tail has been eliminated the skew has been eliminated we you know I would agree that we're not a target yet and I think from those who paid attention the limited number of people who paid attention last week to the ECB policy announcement in that it's crystal clear that still a lot of accommodation is going to be needed in order to make sure we persistently get back to target let me you know that's I think my basic point here is to solve communication because to extend your working truth financial conditions improving the real economy and in turn the real economy through the model Phillips Curve not necessarily the empirical Phillips Curve then building up wage and price pressures that takes time and in that gap which is can be quite a number of years in that gap communication policy is so important saying listen we're persistent we're patient we're going to keep on going we need guidance to fill that gap but of course when you get headwinds like the movement in oil prices it doesn't help and that's why it's so important to be able to have a and it's not just communication by talking it's communication by acting in terms of policy but also by research so this is in the end what we do in the governing council totally relies on what the staff is doing in delivering evidence and I can tell you from my point of view on the governing council what convinces us is not kind of staring at the raw data it's the fact that the modelers both in the ECB and the national central banks are delivering many models and you said this this morning Peter many models and it's not putting your bets on one interpretation of the data or one structural model it's cutting it in many different ways will give you that kind of increasing reassurance that inflation is on its way let me emphasize that this session is about price and weight setting we haven't heard a huge amount about weight setting this morning and I think maybe some people did talk about it but you know I think this is maybe and Philip Lowe actually I think had it this morning I mean where inflation expectations come in big time is in that kind of union negotiation so I think the communications is also I think very much you said you had an information campaign in Australia I think in some of our member countries something similar is going on because it would be a mistake to have backwards looking expectations which will just slow down the adjustment towards targets so I think that's quite important so I think I would say also just on this trend cycle debate John Mulebar said this morning that we had a heart attack in 0809 I mean that sounds right to me I mean if you drive the economy you know that wasn't a normal cyclical shock if you drive the economy so far away from normal it could take a long time to get back to normal and so how you handle that is that a trend break is that a dummy of a multi-stage model where you're transitioning between stages so this is why I go back to the ECB the strategy here among the many models we would have models like that which would allow for those kind of one-off factors or very slow recovery type models so I mean that's my personal view is that this is not a typical business cycle episode and until we're closer to full employment we won't get a very strong labour market needed to push up inflation Jim Stock this morning had something which we talk about in Europe which is in that furniture and home services anyone who's living in a busy city knows that the cost of getting a plumber, the cost of getting a carpenter that's shooting up because these guys they don't care about inflation expectations they just say what can I charge and in a busy economy you can charge a lot more so we also know that negotiated wages are probably moving more slowly but the opportunities for overtime and so on are coming in so I think I'm going to be economical here I see the clocks as I have a minute but I'm not going to use it so you're very much the other side from what Charles was saying we are too much preoccupied with markers and experts you conclude Charles so you insist very much Philip on the financial conditions in general which I agree with you you want to react Charles on this yeah I mean you took a very different view from what you were saying no I understand and the question I think is is there inconsistency between talking to financial markets and talking to the plain people because the argument Phil put well the argument why you want to talk to financial markets right you agree no I don't I don't because financial markets react extremely strongly to whatever you say so you tend to develop a communication mode of extreme precision and extreme care now if you want to go down and talk to the plain people you'll have to talk way more openly and more simply it doesn't have to be incompatible but it will be occasionally incompatible that's something else that's a big difference that's a big difference I take questions now with great pleasure yes I know you but give your name finally very good who are you by the way Richard Portes the Kristin's evidence was quite clear the trend component is down Charles said that we're back to the pre-crisis situation and this whatever has happened it doesn't matter much for economic performance I would put in one qualification there's a big difference that is debt burdens are much higher and with a lower trend rate of inflation it is going to be harder to deal with those debt burdens and I think we have to address that issue it's a very important one yes I wouldn't kill the cycle so easily I agree that the trend is the big component but since this is a policy panel less maybe perhaps is useful to start from our mistake I mean mistakes that we know there were mistakes exposed the ECB rates were increased in July 2008 and then twice in 2011 and this was when oil was very high but we now know exposed that the recession had slowed down the economy was kicking in so the single extraction is super important so don't be complacent that we have killed it and we can relax because inflation is anchored I mean we still that may be true at the long horizon but at the short horizon is super important and you know if you just look at those two increasing interest rates actually make the difference between touch or most deflation and it took a long time to reverse the cycle let's continue to pick up some questions thank you my question is to Christine one thing which I think is completely compatible with the Jim's presentation is indeed when you showed us the change of parameters over time you may tell a story that yes real change rates are very volatile but central bankers are now much better in sticking to the target and in that sense you can tell a story that inflation targeting is now much more aggressive in a sense and even when oil price goes up and down by factor of two oil exporters maintain inflation of two or three percent which is quite amazing if you think about this my question is to you did you try also to look at migration global mobility of labour as a variable it actually matters for many countries in centralist in Europe population is decreasing but also because of out migration and that affects of course wages it affects inflation and again over time they manage it much better in a sense that wages grow fast but inflation is still under control yet another global variable to look at as labour has become much more mobile than it used to but one in the back I don't see where you went and then Michael thank you Angelou Bidet from Goldman Sachs question for Jim but for everybody did you just restate good heart slow essentially by saying that when inflation becomes a target it stores being a useful measure and if so how would you run monetary policy then that's definitely not and then we stop we ask for the answer I enjoyed both these interventions yours especially because you point out that the changing exposure international trade changes price behavior and you see that in Germany where Germany has become very globalized in the past 20 years and Jim's point is another fundamental point the Lucas critique and I wonder why no one has really said that the reason why the Phillips curve doesn't work anymore is because the relative variances have changed parameters have changed we know that's happened so I mean shouldn't we really not be too concerned about the econometrics of the Phillips curve we know we've known since 40 years that it depends on the regime Luigi Zengales University of Chicago I would like to link what Christine said with what Jim said earlier this morning because it seems that in Jim's analysis the three things that were working were real estate, hotels and entertainment all three are very local and have a huge real estate component in it so this is a bit against what Jim is saying because if the Phillips curves would have disappeared for everything but it's not disappear for real estate as disappear for labor and so maybe is internationalization or the loss of power in the labor but that makes it one disappear and the other not okay so actually let me start with Luigi your comment I actually I listened to Jim Stock's presentation earlier not this Jim the other Jim's I came to a very different conclusion that our papers were shared actually came to the same conclusion but from completely different angles so he found the Phillips curve still works for the components of the price index that are more domestically focused but yet then he took out all the other stuff that's determined globally and what I say is instead of taking out a big chunk of the price index let's actually try to explain what's happening in prices in that big part of the price index and you need to explain that now with not with domestic factors but with global factors so in some sense I see my papers a way to actually get at build on what Jim said but then get at that big component of the price index he isn't able to estimate with his purely your more domestic focus so I think that that's a nice link actually then couple there was a couple comments on adding migration I haven't added that yet that probably is important for some countries I was hoping to do all this for wages but the cross country wage data across time where I just wasn't comfortable enough to push it and that's why I think migration would come in most important there were also a couple comments on inflation expectations so just like one result I didn't have time to show if you do these rolling regression coefficients with the global variables it's interesting one variable coefficient that moves a lot is on inflation expectations I mean this would also fit with what you were arguing Jim especially around the time of the crisis and after the crisis when we saw very active central banks you see the estimate of that coefficient on inflation expectations increased quite a bit suggesting that it's stable inflation expectations linked to what central banks are doing was very important in supporting inflation during that period when you had large global output gaps and domestic output gaps so it would be consistent with that story also okay a lot of lots of great comments here on the question of debt burdens rising across the developed economies I'm very sympathetic to this some of you know I have research now that has realistic models of debt and in that model nominal GDP targeting is actually the thing to do so I think this is a very important issue to think about going forward in the world of monetary policy on the question about goodhearts law I think this is very much exactly that that once something becomes a target it becomes that you're going to change the correlations around that how to run monetary policy when this is the case given the discussion this morning you won't believe this but I actually think we should focus more on market based expectations of inflation I consider those very relevant to day-to-day monetary policymaking because market I like it that markets react to the current news I like it that markets have their own bets on the table about future inflation I like it that markets take into account all available information so in the US they're thinking about what fiscal policy is going to do over the next couple of years they're thinking about the low unemployment rate they're thinking about also what the Fed's going to do over the next couple of years so I think it's a great signal for what I do on a kind of day-to-day basis to be looking at market based inflation expectations because these are people that have real money on the table so and that was the whole idea in setting up these kinds of tips markets and I think it's a good way to run monetary policy right now those expectations are low they remain somewhat below our target and that gives me a lot of comfort that we're in a good position for US monetary policy on the question of whether because of the micro if I decompose the inflation into micro components then I find some correlations with the state of the economy so the Phillips curve returns for some components I guess I'm still not quite convinced about this argument but when you go to build a price index you've got a whole bunch of goods and then you've got to think about how much is being spent on all these goods and then one of the prices changes and all those shares change and then now you say well I'm just going to focus on one good maybe without saying too much about how these shares are changing I'm not sure what you're getting there so I guess what I'd like to see in Professor Stock's paper is a toy model that would show me here's how I'm going to construct the index and then here's how I'm going to measure things let's say I only see candy bars that's the only prices I see in the whole economy can I just look at candy bar prices and infer the Phillips curve from that or is that way off from what the whole index would be telling me or under what conditions would that tell me the right signal about the overall index because these shares are moving around all the time and that's a huge part of the price index Philip and then I will take two last questions let me come back to this issue about the role of globalization because I think I would fully agree that elastices change and so on but the particular historical episode we've been experiencing is maybe not a perfect guide to the future which is a lot of the global impulses have been towards low inflation so maybe a lot of investment in manufacturing capacity in emerging Asia maybe not enough domestic spending in those economies for a while so essentially there was a kind of deflationary pressure on say manufacturing prices around the world but as domestic spending goes up in emerging markets as these economies switch from investment towards that consumption maybe without the opposite maybe there would be actually inflationary pressures in and out of some other economies equally cross-country regressions but of course in many economies they're constrained by low inflation overseas so I'm pretty sure as your area inflation goes up and ECB policy normalizes the ability of Sweden to meet its inflation targets will change the ability of smaller economies equally for the US vis-a-vis dollar trackers there's a synchronization issue here which has been here where we've all had low inflation that doesn't necessarily follow for the future but what is for sure is that if inflation goes up in other countries it's going to be easier to meet it at home and everywhere even Iceland is so important a lot of what we consume is local whether it's rents hotels, entertainment whatever personal services in the end you have to have that domestic activity level hot enough to have significant domestic inflation two last questions James Stefan Gell Jim Bullard's refreshing our recollections about the problems with these reduced foreign estimates and this good arts law type issue is extremely useful that's a relevant point the challenge is where that takes you is going back to thinking about estimation of those structural parameters and of course that's a laudable activity but those open up an entire host of econometric problems too and they are basically unidentified you'll find a lot of evidence of that in the literature so now we find ourselves in a circumstance where those structural parameters used in the more sophisticated structural models or the DSGEs for monetary policy are essentially just calibrated objects relying on market expectations they're not going to be able to identify those any better than our excellent staff at the banks so I think it opens up a whole host of new complicated really severe problems that's safe that's safe Stefan before we leave the room so on this issue of communicating with the markets as opposed to communicating with households as Charles brought up it seems to me that households inflation expectations are pretty constant they don't respond to central bank announcements but markets they do respond inflation expectations change and therefore long bond yields change and therefore expected long real interest rate change in the current setup and that sort of strikes me pretty neat I want to hear your views on that one question and a comment to Christine here if you estimate a Philips Curve for any small open economy may that be Sweden or Switzerland or Portugal you would always have some measure of import prices in there you would always have oil prices in there so perhaps we're not so far apart can I invite you to comment on your confidence bands that you plotted because I think that's what a central 30% of the distribution and it seemed to me that if you used any conventional confidence band you know a 95% confidence band there would always be in the confidence band per person I was wrong thank you maybe on this Christine maybe you answer okay okay yes no I plotted some of these so the graphs I showed were the medians when you do this country by country and what stands out when you do this is you get very different estimates for different countries so I just limited it to sort of the medians to get sort of the median of the countries you always have some outliers where global slack doesn't matter import prices don't matter, oil prices don't matter and so I was just trying to focus and that's why I set slightly different bands so the graphs were still readable but I also your comment that of course when you estimate the Philips Curve you add in the import prices or oil prices some countries are much more disciplined by this and do add at least one control for global factors but what surprises me is it's still only it's one variable to capture everything going on in the global economy supply chains trade, import prices, oil exchange rates but yet maybe it's partly because I'm sitting in the US it's amazing how many people who work in the US who are leading macroeconomists who study the world still don't include anything about the rest of the world in their models and that's the one message I want to drive home even if you're a US based economist yeah you need it Jim? Yeah on this question of communicating directly to the masses as opposed to communicating mostly with financial markets I guess my judgment listening to everybody here is it's going to be an uphill battle to if you're really trying to communicate to the general public and that has not historically been how central banks have focused we do a lot more outreach than we used to we do talk to a wide variety of groups we do listen to all kinds of people talking about the economy and try to take their concerns on board and try to learn about the economy but I don't think it's you know I don't think it's going to be a realistic substitute for talking to financial markets so at least in my head we talk to financial markets they set the prices the private sector has to take those prices as given when they make decisions and this is how things the natural order of things I don't know if I can write a model down to get that but that's how I think about things The polling question and then we can go for lunch pick up your application, your smartphones what could central banks do to ensure that inflation expectation of firms and households are aligned with the inflation target press 1 if you think that the emphasis on communication should be shifted from financial market participants to firms and households press 2 if you think that improve financial literacy and awareness of monetary policy 3 for frame communication including the use of social media to target specific population subgroups for all of the build here I have 15 seconds cast your vote I didn't know where that one would come at everybody vote here's the results all of the above wins thank you let's go for lunch now