Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable.
=Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable.
=Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable.
=Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable.
=Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable.
=Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable.
=Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable.
=Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable.
=Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable.
=Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable.
=Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable.
=Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable.
=Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable.
=Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable.
=Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable.
=Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable.
=Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable.
=Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable.
=Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable.
=Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable.
=Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable.
=Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable.
=Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable.
=Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable.
=Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable.
=Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable.
=Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable.
=Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable.
=Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable.
=Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable.
=Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable.
=Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable.
=Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable.
=Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable.
=Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable.
=Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable.
=Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable.
=Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable.
=Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable.
=Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable.
=Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable.
=Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable.
=Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable.
=Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable.
=Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable.
=Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable.
=Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable.
=Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable.
=Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable.
=Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable.
=Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable.
=Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable.
=Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable.
=Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable.
=Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable.
=Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable.
=Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable.
=Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable.
=Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable.
=Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable.
=Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable.
=Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable.
=Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable.
=Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable.
=Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable.
=Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable.
=Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable.
=Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable.
=Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable.
=Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable.
=Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable.
=Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable.
=

# Exchange Rate Puzzles and Policies (AFA lecture)

Authors:   
Discussant: None  
Video: https://www.youtube.com/watch?v=-e-5AL0rWYs&t=0s


## Talk (00:00:00 – 01:39:38)

[00:00:08] So, welcome everybody to this year's American Finance Association lecture 2023.  
[00:00:15] It's my great pleasure to introduce my friend and a former colleague Oleg Itzhoki for to give this presidential address.  
[00:00:24] Oleg is this year's John Bates Clark Medal winner. So, he's the top economist under the age of 40.  
[00:00:31] Uh just I got this prize recently.  
[00:00:35] He's working and he's one of the leading experts on international macroeconomics, on exchange rates, international finance.  
[00:00:43] Uh and you know, he's the voice to listen to.  
[00:00:47] Um he got a a Sloan Fellowship. He got many prizes, was in review tour and so forth.  
[00:00:54] Originally, Oleg came is out of Russia. He studied his undergraduate studies in the New School of Economics in Moscow.  
[00:01:02] Then he went to Harvard, did his PhD there and then came to Princeton and we became colleagues for 11 years and I hope this were his forming years at Princeton.  
[00:01:15] And and then he moved in 2016, he moved to UCLA and um is um shaping the field from UCLA.  
[00:01:24] He's also very engaged in PhD students. I see several of his PhD students in the audience. So, he's a really also working on the next generation. And Oleg is is is fun to hang out with. You will see.  
[00:01:38] Um he has always a positive attitude to all to his surroundings. So, it's um it's always nice to be uh close to him.  
[00:01:49] He's also he became a dedicated biker and I was told he's now surfing also being in in Los Angeles.  
[00:01:55] Uh but today we will talk about work and that's we will focus on today. So, today we'll talk about exchange rate puzzles and policies. And Oleg, the floor is yours.  
[00:02:05] We're looking forward to your presentation.  
[00:02:07] Thanks. Thanks very much, Marcus, for the introduction.  
[00:02:14] Um yeah, I was going to say um I I didn't realize that econ and finance are in different hotels. So, I had to run from a session uh in at Hilton and so I had this uh delightful memories of the job market  
[00:02:29] of running between the hotels, which is you know, no longer here, but in my times that's how we did it.  
[00:02:36] So, um I want it's it's a high bar, I guess to keep people uh entertained here. It's it's a macro paper in a finance association.  
[00:02:48] Uh so, what I'll try to do is I'm going to try to talk about some recent development in international macro and finance. They have very clear overlaps with finance topics, with macrofinance topics. But nonetheless, the work that we've done so far, it will  
[00:03:03] be largely an overview of the work that I did with Dima Mukhin, who's here in the audience.  
[00:03:08] It's really international macro work. But it relies very heavily on themes from international finance. And so, I'm going to try to emphasize the interaction between the the macro and finance here. And so, this is the  
[00:03:22] set of papers that we started working on them a while back. Uh but this is kind of the recent drafts of of these papers. And so, I will try to emphasize kind of how one paper led to the next one. And sort of like we  
[00:03:36] started with basic thinking about exchange rates. How can we write down model a macroeconomic model of exchange rate, which is not failing you know, which is not uh failing miserably badly when it's confronted with the data,  
[00:03:50] right? So, we didn't set a very high bar in the beginning. And as we kept working on it, it's sort of like it's kind of we had to peel the onion and new new themes arise. And then eventually, like the current project is the project about  
[00:04:02] exchange rate policies. Um throughout this research agenda, finance plays a very important role, but we kind of have not really went into finance topics yet, but I'll show you the entry points.  
[00:04:16] Um Yeah, so you know, exchange rate just proved to be one of the most pervasively puzzling variables for macroeconomic models. I guess for finance models as well to a  
[00:04:29] large extent. And kind of the situation here is that as soon as you write down a model with more than one country whether you want it or not, exchange rate is there. You can write down a model where exchange rate is trivially equalized to one all the time, but it's still there and this model still gives a counterfactual  
[00:04:44] prediction for exchange rate then. So, ignoring exchange rate is you can ignore that exchange rate is a variable in the data and not look at its empirical properties. That's a possibility. But once you write down a model with more than one country, you cannot really ignore exchange rate as a theoretical  
[00:04:57] object. It's it's it's right there. Furthermore, it turns out that um uh a lot of the policies in the open economy, right? If you do closed economy analysis, obviously, you don't care about the exchange rate. But as soon as you want to do optimal policy design in  
[00:05:11] open economies, exchange rate is kind of a key object, right? And we oftentimes talk about why whether countries should float the exchange rate or should they adopt a peg, be part of a currency union, what are the costs of being part of the currency union. And so, you kind of really need a model of exchange rate  
[00:05:26] to, you know, even begin to think about those questions. And if you don't really trust your model of exchange rate, its macroeconomic properties, then it's very difficult to know whether you should trust the policy implications of such models as well, right? Especially, like how much  
[00:05:40] should we trust you know, our conclusions about the costs and benefits of Eurozone if we don't really have a model that's consistent with a bunch of properties of exchange rates under the float, right? And so, this is kind of the starting point. Um  
[00:05:55] And you know, most of the literature this this a long-standing puzzles in the literature and there is a large literature trying to address one puzzle at a time. So, our approach with Dima was really to try to think about this  
[00:06:08] puzzles as a combined description of the data generating process for exchange rates. So, you don't want to ignore any dimension of the behavior of exchange rate, right? You don't want to focus on comovement with consumption and ignore the comovement with interest rates,  
[00:06:21] right? And sort of what we try to do is write down a framework that can simultaneously talk to a lot of well, in fact, all of these puzzles at once. And then we didn't know when we started, but as we kept working  
[00:06:35] on it, it offers a very nice policy analysis framework, right? And so, that this are kind of the steps that I will try to go through in this lecture.  
[00:06:45] Yeah, so here I tried to list the set of empirical properties that I'm going to call facts about exchange rates, but most of them have some puzzle name to it, right? Like So, for example, the first fact is called exchange rate  
[00:06:58] disconnect puzzle and narrow version of it is often times referred to as a Meese and Rogoff puzzle.  
[00:07:05] Uh and the idea is, well, exchange rate is a random walk. That property is not so difficult to generate in models if shocks are persistent, you get processes for a lot of variables that are close to random walks. So, getting a random walk in exchange rate is not that hard. But a  
[00:07:19] much more pervasive and difficult puzzle to address is even if you knew information from the future, if you knew macro variables tomorrow if you knew what happened to monetary policy, to interest rates, to output, output gap, inflation,  
[00:07:33] consumption, this doesn't really help you predict what happens to exchange rates from today to tomorrow. And that's a very surprising property, right?  
[00:07:41] You're trying to improve a forecast of exchange rate over random walk, even if I give you information from the future, it's quite difficult to improve that forecast over random walk forecast. And that's really the core of the Meese and Rogoff or the disconnect puzzle, right?  
[00:07:55] And so, you want to write down a model where correlations between exchange rates and macro variables are very weak, close to zero, while a lot of the models predict in fact strong correlations between macro variables and different notions of exchange rate. Furthermore,  
[00:08:09] exchange rate is about an order of magnitude more volatile than macro variables. So, typical standard deviations of, you know, inflation, consumption, output is about 1 to 2% a year.  
[00:08:20] Uh floating exchange rates standard deviation is 10 12% a year. And so, the question is why do you have this disconnect in volatilities as well, right? So, the finance version of that is in fact the statement that exchange rates are too smooth, right? They're too  
[00:08:34] smooth relative to the stock market. They're about 2/3 as volatile as the stock market. Uh and so, this is an important puzzle to address. I'm not going to talk about it today. This is part of the research agenda too. I mean, we kind of have  
[00:08:48] a sense of the explanations there, how to write down a macro model that's simultaneously consistent with, you know, smoothness of exchange rate relative to the stock market. But I I I'm not going to cover it in the slides at least today.  
[00:09:01] The next puzzle and in fact, one of the famous ones, the ones from which this whole research agenda started many, many years ago, you know, people been working on this for decades, you know, going back probably 50 years, is the PPP puzzle.  
[00:09:16] And it's the fact that you can look at a nominal exchange rate, which is a relative price of currencies and you can look at the real exchange rate, which is a relative price of goods, of consumption baskets. Real exchange rate is also a deviation from the purchasing power parity, right? So,  
[00:09:31] it's a very kind of real object. And exchange rate is a very monetary object.  
[00:09:35] It's about the relative price of currencies, right? But for some reason, for for each countries that control inflation uh nominal and real exchange rates are nearly perfectly correlated.  
[00:09:46] It's sort of not that surprising if you thought prices were very sticky, but the problem is that that near perfect comovement uh persists to horizons way past beyond any degree of price stickiness to 5 years out and even longer. And so,  
[00:10:00] whenever you try to kind of see uh is the uh you know, real exchange rate maybe mean reverting and nominal exchange rate is not, does the gap open between the two? Well, the answer is in finite samples that we have, you you don't have really much evidence that the two, you  
[00:10:14] know, decouple very much, right? They just happen to be highly correlated even at horizons above uh 5 years.  
[00:10:21] Uh the next puzzle, and this is the one that really links finance and international macro, it's kind of probably a core puzzle in both, is uh Backus-Smith puzzle. So, uh um uh Robert Kollmann and Backus-Smith kind  
[00:10:36] of documented in the data that if you look at comovement between real exchange rate and relative consumption, the correlation is negative. Well, if you think about perfect um risk sharing, it predicts a positive perfect correlation  
[00:10:51] of plus one. So, the data is close to zero, slightly negative. A lot of models predict plus one, right? So, the question is how do you explain that? And it turns out, in particular in international macro, it turns out to be a very challenging model that even giving up on  
[00:11:04] complete markets does not really help you solve it. And I'm going to kind of talk about it a lot but a little later.  
[00:11:11] I need to build up to get there. Uh again, another puzzle that comes from finance is the forward premium or UIP puzzle, you know, the famous Fama 1980 paper, huge literature both in finance and  
[00:11:24] uh international macro. Uh often times the emphasis is that there is a wrong correlation uh between interest rate differentials and uh for- forward-looking appreciations, right? Like that basically  
[00:11:37] from a point of view of some sort of ev- equalization of expected returns, high interest rate must mean that you expect a devaluation of that currency to compensate for the high interest rate.  
[00:11:48] Otherwise, there is a uh profit opportunity that opens up, the the basis for carry trades, right? Uh so, a lot of emphasis in the literature was on the sign of the coefficient. We kind of think that that emphasis is somewhat misplaced. The coefficient is close to  
[00:12:02] zero. The the emphasis should really be that R squared in that regression is about zero, right? It's it's kind of 0.01 R squared. And so, it turns out it's not very difficult for models to match that R squared. It's  
[00:12:15] just not uh in the class of models that we'll be working with, it's just not a very powerful moment for identification, it turns out.  
[00:12:23] Uh but, it's also an important feature of the data that carry trade returns are there but not very high in terms of the sharp ratio, and the R squared in the Fama regression is close to zero. This would be our And you know, if you want the coefficient is negative as well, but  
[00:12:37] it's not as important in our opinion. And finally, there is a Mundell-Fleming puzzle. And it turns out that these puzzles, they kind of crucially uh crucial statistical discipline and devices for a macro model of exchange  
[00:12:51] rates under the float, but as I will show you, it basically does not give you an entryway into thinking about policies, right? But, the Mundell-Fleming puzzle turns out to be a crucial additional feature of the data that allows us to  
[00:13:05] get into the policy analysis, right? And so, what is the Mundell-Fleming puzzle?  
[00:13:08] Essentially, you can think of let's look at all the statistical properties under the floating exchange rate regime, and let's look at the statistical properties under the fixed exchange rate regime, the way they change. And it turns out they change in a very peculiar way. I'm going to talk  
[00:13:21] about it in a couple of slides. And this provides a very sharp identifying tool for us for thinking about the right financial model of exchange rate.  
[00:13:29] And knowing that allows us to go to policy, right? And so, hopefully it will at this point it's not supposed to be clear, right? At all.  
[00:13:37] But, hopefully it will become clear in a few slide. Um Okay. So, let me show you a couple of things in pictures. How I think about exchange rate disconnect in pictures. So, this is like a conventional picture. This is GDP per capita PPP adjusted for a bunch of  
[00:13:50] countries, so that each country is here. Switzerland, United States, uh UK, Japan, Australia. These are the rich countries that I plot. And you kind of can see this orderly progress, right?  
[00:14:01] Like you see this, you know, fairly smooth developments, kind of constant, fairly constant growth rates, right?  
[00:14:08] Nothing crazy in this picture. But, what will happen if we plot this picture in dollars, right? Not in current dollars, not PPP adjusted but in current dollars.  
[00:14:18] And so, I I I plot here Switzerland, uh US, and Australia. And so, you see that back in 2000, uh Switzerland was a typical Swiss citizen was as rich as typical American citizen according to this measure. You  
[00:14:33] don't want to do comparisons over time because there is inflation over time.  
[00:14:36] But, in a given point in time, it gives you a ranking of countries. So, basically if you say, let's take a typical Swiss person and a typical American and put them in a duty-free shop in Frankfurt, the purchasing power of a typical Swiss person and a typical American person was about the same in  
[00:14:51] 2000. Let's scroll forward 15 years, and suddenly a typical Swiss person is two times richer than a typical American person, right? And so, the question is how can you, you know, think of a macro model in which relative purchasing  
[00:15:04] ability of a Swiss person relative to American person can increase two times over a period of 10 years, and then you look at the domestic macro aggregates and nothing happened to them, you know, normal GDP growth, normal inflation,  
[00:15:18] nothing abnormal. But, this humongous swing in the purchasing ability of a typical Swiss person and a typical American if you put them in a duty-free shop in the same place. Obviously, they don't This is not where they buy their consumption baskets, right? But, nonetheless, you can do that, you know,  
[00:15:32] uh thought exercise, right? Same thing with Australia, it was half as rich in 2000, and suddenly becomes 60% richer than America uh 10 years later. Of course, you can tell me that, you know, Switzerland was a uh boom of demand for  
[00:15:47] safe assets that were produced by Switzerland, and Australia was a commodity boom, and both of it is true.  
[00:15:52] But, nonetheless, you still want to write down a model where this picture could be, you know, an outcome without any kind of crazy movements in the domestic macro aggregates, right? Like a well-being of a uh typical Swiss person, if we measure it in income per capita  
[00:16:07] adjusted for cost of living, right? Did not increase as dramatically as this picture would suggest.  
[00:16:13] Well, this picture, you know, UK gives us a lot of interesting data to think about because it's a floating exchange rate and lots of shocks, right? So, this is in the end of the sample there, it's a Brexit shock.  
[00:16:26] On the day when the Brexit vote was announced, the pound moved 11%.  
[00:16:31] And of course, it's a forward-looking variable, so it anticipated a lot of things that might happen. So, it's not surprising that the pound moved by 11%.  
[00:16:38] But, if you look at any macro variable, there is no discontinuity before the vote and after the vote. So, the amazing thing here is not the jump in the exchange rate, it's a forward-looking variable. But, like when exchange rate jumps, why the rest of the macroeconomy  
[00:16:51] not responding to that, you know, in next months, 3 months later, 6 months later? They you would not be able to detect any sort of structural break in the macro series. And so, this is another notion of the disconnect here, going from a jump in exchange rate,  
[00:17:05] which is easy to understand, to the lack of any movement in macro variables, right? Uh so, this is the recent UK event when the pound again moved by around 10%. At first, it depreciated and then it appreciated. So, this one is in a way less surprising because it was not  
[00:17:20] a persistent movement in the exchange rate. The Brexit one was a persistent movement. But, this one is interesting.  
[00:17:25] Our model will be able to kind of I mean, I'm building up to saying that I have a model that can can talk to a lot of these pictures. And this picture would be an interesting one for thinking about policies.  
[00:17:37] Uh this is Abenomics in Japan. Uh so, Shinzo Abe, when he became the prime minister, he announced Abenomics. It had a number of uh policy components to it. One of them was QE.  
[00:17:49] And QE was very successful at depreciating the yen. So, the yen depreciated about two times. But, nothing happened to the rest of the macro. There was no inflation. There was no change in output growth.  
[00:18:01] And so, you can write down models where nothing moves, right? These are the models of liquidity traps when, you know, your uh monetary easing and just doesn't help to anything. But, in those models, exchange rate will not be moving as well, right? And so, here you need sort of a model where the policy was  
[00:18:15] successful at moving exchange rate, but not successful at uh you know, uh doing anything to inflation or to the output gap, right? And that's a challenge. If you try to do it with a macro model, you're just not going to be able to calibrate it that way.  
[00:18:29] Uh you know, conventional macro model, I should say. Uh I'm not sure I'm going to have time to talk about this, but this is what happened to the ruble after sanctions after war in Ukraine started and uh uh sanctions were imposed on the Russian economy. And so, the amazing  
[00:18:44] thing is the same framework that we're going to write down for thinking about very financial kind of questions will kind of also talk to this path of the ruble exchange rate. So, we can basically use the same equations and plug in sanctions in those equations  
[00:18:56] instead of financial shocks, and it will uh spit out something that looks like the path of the ruble exchange rate in 2022.  
[00:19:05] Okay. So, this was kind of the motivating facts, kind of the set You can characterize them as unconditional moments that I showed you on the first slide and like bullet points, but then the pictures were meant  
[00:19:18] to illustrate how those facts manifest themselves, right? And uh particular episodes of time, right? So, from now, it's really the analysis, right? So, this picture is a little more information about exchange rate. So, what I plot here is different notions of  
[00:19:33] nominal and real exchange rates and terms of trade.  
[00:19:37] And so, if you look at Well, there is it's a US trade-weighted, I think, nominal exchange rate against its main trade partners.  
[00:19:44] And then I plot a bunch of uh comparable real exchange rates against the same trade partners. So, it could be a consumer-based real exchange rate. It which says, if I looked at a price of a consumption basket in US relative to a price of consumption basket in other  
[00:19:59] countries, how does that relative price evolves? And this is, I think, a blue line. Black is nominal, blue is consumer-based real exchange rate.  
[00:20:08] Red is let me look at producers. What is the producer-based producer price-based real exchange rate? And again, it co-moves together. Then we can look at wages and we can say, well, let's look at typical incomes that people get as  
[00:20:21] workers. And let's look at a wage-based real exchange rate, right? Like uh uh you know, a unit wage in US relative to a unit wage in those other countries.  
[00:20:30] And amazingly, even the wage-based thing uh at high frequency co-moves.  
[00:20:34] Obviously, the gaps opens up more persistently between the wage-based and price-based because all the accumulated labor productivity differences should be in that gap. So, it's not surprising, but what's surprising that at high frequency, all those notions of real  
[00:20:48] exchange rate co-move. And then you have this uh lone relative of this variables is the terms of trade, right? So, terms of trade is green and it is not doing the same thing as the exchange rates, right? So, in a lot of models, we're going to confuse real exchange rate with terms of  
[00:21:02] trade because models say they should be very closely correlated and co-move very closely. But in the data, terms of trade is this well-behaved macro stable variable moving a little bit slightly positively correlated with exchange rate, but nothing like the craziness  
[00:21:17] that we observe uh in the exchange rate. And this is this is also part of the moments that you want to capture in the model.  
[00:21:25] Okay. So, finally, I can get to the Musa puzzle. So, Musa puzzle is this uh thought exercise that I'm going to do uh in this four sub plots.  
[00:21:34] So, the first plot is the real exchange rate.  
[00:21:37] Uh '73 was the year, you know, February '73 is when Bretton Woods system collapsed. US refused to exchange dollars for pounds of dollars for gold.  
[00:21:50] And uh basically, the big countries went into a float against the dollar. Before '73, it was a nearly perfect peg. The last couple of years were associated with some jump appreciations and devaluations of certain currencies, but  
[00:22:05] this were like isolated events. Other Outside of this isolated events, this was a perfect peg of all G7 countries to the dollar, basically. So, in nominal, if I plotted you the nominal exchange rate, it was completely flat before '73  
[00:22:19] and then became a volatile variable after '73. What's interesting here in the top panel, I'm plotting the real exchange rate.  
[00:22:26] I mean, it's not so surprising that nominal exchange rate changes its property when you change a monetary regime. But why should the real exchange rate change its property? And in fact, what happened is that real exchange rate looked like a normal macro variable. So, I plotted on a scale of, you know, the  
[00:22:41] same scale. So, in the bottom, it's the consumption. I could have plot output or inflation. And so, real exchange rate behaved sort of like a normal macro variable during the peg with a standard deviation annualized about 2% a year.  
[00:22:54] But suddenly, you go to a different monetary regime and uh real exchange rate starts looking like nominal exchange rate under the float with an annual standard deviation of 12%. Right?  
[00:23:04] And there is a clear discontinuity. You you you you you look at the data, you immediately see Oh, and by the way, this is just raw data. I didn't do any transformation of the data. It's really just the evolution of the dollar trade-weighted exchange rate against a  
[00:23:17] bunch of rich countries. And you kind of see immediately where the structural break happened. You look at consumption or you could look at inflation or you can look at, you know, output.  
[00:23:27] You cannot eyeball any kind of structural break. And even if you use statistical methods, if they will find a structural break anywhere, you know, changes in the properties of consumption were maybe by 5, 10, 15%. Change in the behavior of  
[00:23:42] exchange rate was 10 times. So, it's an order of magnitude difference, right?  
[00:23:45] And so, this is the basis for the uh Musa puzzle. So, Musa famously said, well, look, you know, if real exchange rate changes its properties, it cannot be that prices are flexible, right? It cannot be that money is neutral because the only thing that happened was a  
[00:23:59] change in monetary regime. Why would the real variable change its behavior so drastically? Right? But then Backus and Smith they say, well, let's look at other macro variables. And other macro variables actually did not change their behavior when there was this massive change in the behavior of exchange rate.  
[00:24:14] And so, here I plot, for example, a rolling rolling uh window standard deviation annualized standard deviation of different variables. And so, you can see how standard deviation for nominal and real exchange rate, for nominal, it goes from zero  
[00:24:28] to 12%. For real, it goes from something like 2% to 12%. And nominal and real exchange rates become essentially indistinguishable during the float.  
[00:24:38] While if you do the same for inflation, consumption, output, you don't see a structural break. I mean, yeah, like if you use some some sort of statistical tests, occasionally, they find a structural break, but basically, you know, the red lines before and after  
[00:24:52] tell you the average volatility in the first period and then the later period, right? And so, kind of this is really uh this the sense of the Musa puzzle. And this provides us a lot of information for identification.  
[00:25:07] Okay. And so, going a little closer to finance, well, on the financial side, so we could look at Backus Smith correlation. And we could look at the fama regression coefficient. And what's interesting is both of those things, remember the benchmark is plus one for  
[00:25:21] both of them. The data is, I told you, under the float is close to zero, slightly negative. And so, this is what you see, right? Like in the data, during the float, this are the red bars.  
[00:25:32] It's negative. Sometimes a lot negative, sometimes mildly negative. Uh during the peg, the puzzle is kind of less uh strong, right? The coefficients that tend to be positive sort of in the right direction, but again, they're not plus  
[00:25:47] one. And so, you want to kind of build a model where a lot of this puzzles um are particularly strong in the floating period, right? So, this is another kind of over-identification  
[00:25:58] uh set of moments for for the theory. Okay. So, from now on, I'm just going to go into equations and try to give you a flavor of how this models work. It's it's really one model with different applications.  
[00:26:13] So, the starting point is the PPP puzzle because a lot of people tried to get them to thinking about exchange rates from the point of view of PPP puzzle, right? Because we can measure nominal exchange rate, we can uh measure inflation and so calculate the real exchange rate. And it already offers a  
[00:26:28] very interesting puzzle to solve. Why is nominal and real exchange rates are so closely uh correlated? So, what is a real exchange rate? Well, essentially, you look at a uh consumer price level in one country, consumer price level in the other country, convert it to the same  
[00:26:42] currency. And if they co-move, well, you say something like PPP holds at the aggregate in a dynamic sense, right? And the intuition sometimes is given like the law of one price, right? That the consumption basket in one place should  
[00:26:56] be the same value as consumption basket in other place. Of course, when you live in the world of non-tradables, the consumption baskets are not the same.  
[00:27:03] So, it's PPP is a different concept from the law of one price, right? And so, a very important paper is Engel '99. So, he said, well, let's do that decomposition. Let's split real exchange rate into tradable part and the  
[00:27:17] non-tradable part. And see what component drives it. So, if it's about consumption baskets not being portable across space, right? If it's about non-tradables, then maybe the big deviations come from the non-tradable component, but the tradable real  
[00:27:31] exchange rate is actually kind of pretty stable, mean reverting, not very volatile, and so on. And so, when he did that analysis, uh I'm not sure if like I was a student later, so I already took it as given.  
[00:27:43] So, I wonder back in '99 how surprised he was to find what he's found. But essentially, what he found was that 90 to 95% of variation in real exchange rate comes from the tradable component.  
[00:27:54] So, it's not about relative prices of non-tradables for which we have theory, Balassa-Samuelson model was a theory of, you know, relative non-tradable prices.  
[00:28:03] But he actually found that it's all about the tradable component. And that kind of set the literature on this trajectory to try to find law of one price deviations because conceptually, like people thought that if it's a tradable component, it's got to be that prices of tradable goods are not  
[00:28:16] equalized. So, it's got to be that the law of one price doesn't hold. So, maybe it's sticky prices that explain it or maybe it's variable markups that you set different It's called price into market.  
[00:28:25] You set different markups at home and abroad and that's why the prices are not equalized for the same good. And this is where the literature went uh for a long time. A lot of my early work was, you know, working with models of law of one price deviations. But a very surprising  
[00:28:40] finding, I I'm not sure who to attribute it to.  
[00:28:44] We discovered it with Dima, but I guess it existed uh in Atkeson-Burstein as a quantitative result. And perhaps a lot of people knew about it, but I don't know where it was actually written down.  
[00:28:54] The the the quite remarkable finding here that law of one price deviations do not really do anything for understanding uh real exchange rate, purchasing power parity deviations. Right? And so, it's  
[00:29:07] very easy to make that case when you work with variable markup models. It turns out that if some people some firms increase markups, other firms reduce markups. Very intuitive, right? Like somebody gains competitiveness, somebody loses competitiveness. You cannot have  
[00:29:21] that everybody I mean, it could be that all markups go up, but when you have international shocks, exchange rate shocks, or tariff shocks, it typically favors firms from one country and hurts firms from the other country. And so, you have some markups going up, some markups going down in a given market.  
[00:29:35] And the result in a large class of models, those effects wash out. And so, even though law of one price deviations is a very robust feature of the data, it's right there. If you have micro data on prices, you can see all sorts of law of one price deviations there. It just turns out they wash out to a large  
[00:29:50] extent. And so as a result in the end of the day, you just get that equation which links real exchange rate to wage-based real exchange rate. 1 - 2 gamma is the home bias, right? So home bias is important  
[00:30:04] and it's a home bias in tradeables as well as in non-tradeables is that typically our consumption baskets are biased towards the domestic value added whether through tradeables or through non-tradeables and it doesn't really matter.  
[00:30:15] But the point is that different things of with associated with law of one price deviations, they just drop out from this equation. You can crank up the amount of law of one price deviations and it will not affect that relationship basically.  
[00:30:27] And so basically what this suggests is that PPP approach to thinking about exchange rate kind of has a dead end. Even if you understand what happens to the real exchange rate, it kind of doesn't give you an entry into thinking about  
[00:30:41] general equilibrium properties of that variable. And so this is sort of like a starting point. This is a very partial equilibrium goods market approach. From here you really need to get into a general equilibrium modeling of exchange rates where it's a completely different set of equations that pins down the  
[00:30:55] properties of exchange rates. And this would this would end up being a side equation. It's useful as a side equation, but it doesn't pin down the general equilibrium properties of exchange rate. And so the way I don't know if Dima agrees with me, probably he does  
[00:31:10] but the way I think at least about the PPP puzzle at this point is that it's very simple. Something drives real exchange rates.  
[00:31:19] You need a model of volatile real exchange rate. Well, real or nominal, it's a little tricky, you'll see now.  
[00:31:24] But you are in a floating exchange rate regime where the central bank is good at stabilizing inflation. So it's not about price stickiness, it's not about variable markets, it's just a simple statement that central banks are pretty  
[00:31:36] good outside of 2021 2022 at stabilizing inflation. And if they're good at stabilizing inflation, whatever drives real exchange rate also drives nominal exchange rate, right? The link between them is monetary policy that stabilizes  
[00:31:49] inflation. And this is a trivial theory, right? It doesn't tell you what drives the exchange rate. It just tells you that if central banks are good at stabilizing inflation, you're not going to get a gap between real exchange rate and nominal exchange rate. This would be one kind of by default piece of the  
[00:32:04] theory, the fact that you know, in floating regimes rich country central banks know how to stabilize inflation, right?  
[00:32:10] But it doesn't help us make further progress. We need something else.  
[00:32:14] And so there is a paper by Eichenbaum, Johansson and Rebelo who makes a related point, right? That basically you can explain this co-movement between real and nominal exchange rates just if the central banks are good at stabilizing inflation. So PPP puzzle is kind of  
[00:32:27] super easy if you adopt that view. Okay, I love this one.  
[00:32:34] Is real exchange rate stationary, right? A huge question. I'm I'm not sure why. I think finance economists do, but at least macroeconomists are very committed to this idea that real exchange rates are stationary or mean reverting in the long run at least, right? Where does that  
[00:32:49] idea come from? Is there empirical evidence for that? Not really. Is there good theoretical reason for that? No.  
[00:32:55] There are two reasons. One, it helps us close the models without specifying the full model. You just say something is stationary and you throw out one equation it turns out, right?  
[00:33:04] Secondly, um there is the sense that exchange rates are mostly driven by monetary shocks.  
[00:33:10] Money is neutral in the long run, so it better be that real variables are mean reverting like real exchange rate. And so maybe that's kind of second reason.  
[00:33:17] But conceptually, how how do I think about exchange rate? Well, it's not mean reverting. There is no necessary financial equilibrium condition that says it needs to be mean reverting. The condition is a transversality condition on net foreign assets of the country.  
[00:33:31] You have a intertemporal budget constraint and it has to hold. So on one side it's a no Ponzi game condition from the financial market that's imposed on the country. On the other hand, it's a transversality condition of optimization  
[00:33:44] of the country, right? It means that you cannot have exploding net foreign assets neither in positive nor in negative part and that pins down the long run value of real exchange rate. It doesn't have to be stationary. If you accumulated a lot of net foreign assets, your exchange  
[00:33:58] rate will tend to appreciate in the long run, right? Of course we can write down models where there are forces through the financial equilibrium condition which make real exchange rate stationary, but this is an additional piece of theory, right? Like a priori there is no good reason to think of  
[00:34:11] exchange rate as stationary, but you really need to impose transversality condition on net foreign assets and that pins down where the exchange rate goes in the long run, but it means you really need a country budget constraint. And like who wants to write down a finance model and put a whole country budget  
[00:34:26] constraint into that model? That seems very onerous, right? So but but really what pins down the long run properties of the real exchange rate is is is the intertemporal budget constraint.  
[00:34:37] Okay. Oh, so next I can get to Backus-Smith. So this is kind of the fun stuff. Right? Okay, so how do people typically, well, starting from the Backus-Smith paper back in 1993, how do people typically think about the  
[00:34:51] relationship between consumption and real exchange rate, right? So I assume separable CRRA utility which is like a you know, simple starting point and then the stochastic discount factor is just  
[00:35:05] consumption growth. Times sigma times the relative risk aversion, right? And then in complete market models, nominal exchange rate changes in nominal exchange rate is just the M M minus M star relative SDFs, right? And SDFs are linked to  
[00:35:19] consumption. So consumption growth is linked to real exchange rate by the by the efficient risk sharing between countries. And so over there in the corner over there, so I wrote from the point of view of the planner, what is  
[00:35:33] the planner does? The planner says I have a marginal dollar left. Who should I give the marginal dollar to? To the home household or to the foreign household? And so the home household will convert it into a home consumption basket using the price level at home and that would give the home household  
[00:35:48] marginal utility. The foreign household needs to first convert the dollar into foreign currency, then use foreign currency to buy foreign consumption basket and then convert it into marginal utility. But the point is that the planner is trying to  
[00:36:01] equalize the cost of delivering marginal utility in the two countries. And of course if you write a complete markets model, Arrow-Debreu model, it decentralizes what the planner is trying to do. So it's not surprising if consumption is expensive, you want to  
[00:36:14] give less consumption to the country where consumption is expensive. Why?  
[00:36:18] Because you know, $1 will go a shorter way of delivering marginal utility, right? And so this is the idea that you give consumption more in a place where prices are lower. When where exchange rate is  
[00:36:31] depreciated. So high consumption, depreciated exchange rate. And in macro we measure depreciations going up, so it's like a positive correlation, right?  
[00:36:39] So depreciation, high consumption. In the data, the correlation is negative, right? Of course we don't even complete markets. Like I mean, who believe I maybe somebody does. But so it seemed that maybe Backus-Smith puzzle is like not such a difficult  
[00:36:54] puzzle to solve. Just let's give up complete markets.  
[00:36:57] Right? And so when you give up complete markets, financial equilibrium condition starts to look something like this. So instead of holding state by state, the relationship is between expected consumption growth and expected devaluation. This is how you will be  
[00:37:10] trading the bonds. But then you can figure out models which have the shocks here. So the shocks can reflect efficient risk premia, they can reflect financial frictions. They can reflect intermediation frictions and segmented  
[00:37:24] markets. And there is a a lot of models that would put that wedge in the equation. And so oh, you see the budget constraint was equation number one in the theory.  
[00:37:33] This risk sharing condition in this form will be equation number two in the theory. And it turns out in a large class of models that shock is also the UIP shock. It's something that you need to have a departure from you know, fama  
[00:37:47] fama regressions to go in the right way. And so in that sense solving this puzzle is quite related to to that puzzle as well.  
[00:37:54] Um Okay, what did I want to say here? Okay.  
[00:37:58] So so what's the problem? Well, the problem is that people in macro try to write down models of incomplete markets and the Backus-Smith puzzle remained a very persistent puzzle for the literature. So you write down incomplete markets with productivity shocks. You  
[00:38:12] write incomplete markets with no monetary shocks and sticky prices and still you the model produces a correlation between consumption and real exchange rate which is not plus one, but it's plus 0.97, right? You simulate these models and it's like departing  
[00:38:27] from plus one is very very difficult. And it turns out it's not about market completeness, it's about something else.  
[00:38:32] And so I guess where in a way where I think we made a big progress in in the papers with Dima is trying to fi- figuring out where that puzzle is coming from. And it turns out in order to solve the Backus-Smith puzzle, you need to go  
[00:38:46] to the goods market. It's kind of ironic, but you have to forget about financial market for a second even though Backus-Smith condition originated as a financial paper, right? But you have to go to the goods market. And in the goods market you have a goods market clearing condition. Goods market  
[00:39:01] clearing condition is present in every model that you write. Like markets need to clear. If you have goods trade of some sort, goods market need to clear.  
[00:39:08] And the point is the following. So this is already a you know, this is equation that you get from goods market clearing.  
[00:39:14] And so what's the intuition here? So remember this term right here is about home bias and we work with models where there is a lot of home bias like in the data. So you know, you divide by something, but that number is not very far from one, right? So you can ignore it.  
[00:39:28] But what does that equation says? Well, that equation says that there are two types of shocks. There are shocks that expand output available for consumption in each country. This is Y minus Y star.  
[00:39:40] And so if you expand output available for in a given country, that country will expand consumption. That's goods market clearing, right?  
[00:39:48] But then there is expenditure switching here. The blue term is the expenditure switching. It's a movement on the real exchange rate.  
[00:39:55] And basically you say that when the domestic goods are becoming cheaper, everybody in the world starts switching to consumption of the domestic good.  
[00:40:04] But the only way this can happen in equilibrium to clear the markets, if my relative consumption falls relative to the rest of the world relative consumption. Otherwise you're not going to have enough of the domestic goods.  
[00:40:13] I'm home biased towards my good. So it's only possible that everybody increases the consumption of domestic good when exchange rate depreciates if my consumption falls relative to the foreign person's consumption, right? And this is why the coefficient is negative.  
[00:40:26] It's just a requirement of goods market clearing that if domestic goods become cheaper in relative sense, domestic consumption should fall relative to foreign consumption unless there was an increase in output. So in models where there is no increase in output, just movement on exchange rate, you're going  
[00:40:41] to get that negative correlation. Uh let me it's a little tricky this logic a little tricky. Every time I explain it, it's even tricky for me. Uh but the logic going the other way is much easier. Think of now a financial shock.  
[00:40:55] I have some time still, right? Uh you you have a financial shock that makes you want to delay your consumption. So I'm Germany. I I don't know if it's a good example. Um Uh let's pick another country. Uh I'm  
[00:41:07] I'm I'm UK. There is turbulence in UK. I want to delay my consumption today.  
[00:41:13] Uh uh what happens then? Well, the goods are produced. It's just I want to delay my consumption, but the goods are already produced. I need to allocate them to clear the market. How can I allocate the goods that are produced?  
[00:41:24] Well, British goods need to become relatively cheaper.  
[00:41:28] And then everybody in the world will switch towards British goods. That's the expenditure switching, right? So even though the British people chose to do more savings today and less consumption today because of say financial crisis or you know some uncertainty increase or something like  
[00:41:42] that. But the exchange rate will have to allocate the British produced goods everywhere in the world and that requires a pound depreciation, right?  
[00:41:50] And so this is going from consumption to real exchange rate. But it really explains the same locus of equilibrium points on the goods market. When does the goods market clear? And you get that negative correlation. So basically if the shocks is something that comes  
[00:42:04] from outside of the goods market and either moves your savings consumption decision or moves your real exchange rate decision, but doesn't move the output, you're going to get that Backus-Smith negative correlation like in the data. The other cool thing is it's multiplied by gamma. Gamma is  
[00:42:17] openness of the economy. So this is a weak force. Expenditure switching is a weak force because you're trying to move the whole aggregate consumption with just the international dimension of things, right? And countries are not very open. So moving the whole aggregate  
[00:42:31] consumption is very difficult and it reflects in, you know, gamma being the openness parameter and this is quite small. And as a result, expenditure switching force always gives you the right sign of the correlation. It's negative like Backus-Smith have found in the data and it's a weak correlation  
[00:42:46] because expenditure switching is a weak effect. So now why macro models why macro models could not deliver the right things? Well, because macro models are macro models because they move output, right? So think about RBC model. RBC  
[00:42:59] model is a productivity shock moves output. So it's a shock to Y.  
[00:43:05] Uh the other type of model is a new Keynesian open economy model. It's a monetary extension and what does a monetary expansion do? Well, monetary expansion reduces markups, right? Like like money goes up, prices or wages do  
[00:43:18] not catch up with money, right? And you have a reduction in markups. Reduction in markups is like an increase in productivity, right? It it increases the amount of output in equilibrium. And so in that sense, both monetary and real  
[00:43:31] models, they both operated through increasing the amount of output that's available for consumption. But when you have more output, you have more consumption, but the goods become cheaper. In one case goods are cheaper because productivity shock happened. In other case goods are cheaper because  
[00:43:45] markup shock markup went down. And so as a result, you again get this counterfactual Backus-Smith correlation when consumption is high when prices are low. So it's a positive correlation between consumption and real exchange rate. So it was not about the equations  
[00:43:59] being wrong. Equations are the same. It's about the nature of shock. And so as a result, what this tells you is in order to get the Backus-Smith moment turns out to be the identifying moment.  
[00:44:09] If you look at the Backus-Smith correlation in the data and you're trying to get minus point two, you need to mix exactly the right amount of conventional macro shocks which gives you a correlation of plus one with the right amount of financial shocks over there which will move exchange rate in  
[00:44:24] equilibrium and give correlation according to the blue relationship, which is negative, right? And so because the blue relationship is weak, you need a lot more financial shocks as drivers of exchange rate than macro shocks. And that explains why the financial shock  
[00:44:39] has to account for like 80% of movement on exchange rate driving a mild negative correlation and the macro shocks drive 10-15% of exchange rate driving a very strong positive correlation. And then when you combine it together, you get  
[00:44:53] that unconditional weak mild negative correlation that Backus-Smith found in the data. So graphically, the way it looks, uh the blue curve uh is um the blue curve is the risk sharing  
[00:45:06] curve. So basically if I plotted this uh and this, you will get the financial market equilibrium is a upward sloping line and the goods market equilibrium is a downward sloping line. And so what you want is that you shift you you do  
[00:45:20] financial shocks, you shift around the financial market equilibrium curve, but you move along the market clearing condition on the goods market and that gives you the negative correlation, right? Instead of macro models that shifted around the goods market equilibrium condition and that you get  
[00:45:35] the correlation along the financial market curve, which has the wrong sign, right? So I mean I don't know if this helps, but uh you know, this is the same equations that here uh put on eh I'm sorry. Uh so basically  
[00:45:48] if if you can see it easier from equations, you know, this is this way.  
[00:45:52] If it's easier from the uh chart, it's this way.  
[00:45:55] Okay, so this is what we think explains the Backus-Smith moment uh in the data.  
[00:46:01] Okay. So now I'm I'm I'm I'm I'm I'm I'm I'm almost there. I got the unifying framework for thinking about exchange rate. So remember we have the uh equation number one was the budget constraint. So this is accumulation of net foreign assets and this is net  
[00:46:15] exports. And there is expenditure switching inside net exports. So when exchange rate depreciates, you can expect an increase in net exports at least eventually, right? Well, we're going to write it statically, but empirically at  
[00:46:28] least eventually there is a indeed a relationship like this here with a lambda being positive. Uh you know, this is just net exports written out. It takes a little work to get there, but this is without any cheating you can get to this equation. Um this is the  
[00:46:42] financial market equilibrium condition, equation number two.  
[00:46:45] Uh and you can see that it turns out that in a large class of models this risk sharing wedge psi is also the UIP wedge. And remember we wanted that a lot of the shocks come from that wedge, right? And finally goods market clearing that we just  
[00:47:00] discussed, equation number three. If you put those three equations together, uh you have a theory of exchange rate.  
[00:47:05] Um well, I guess I hear a question saying like, but wait, Alex, there is a real exchange rate there and it's a kind of nominal exchange rate in the financial market here. So like how do you go between nominal and real exchange rates? Because when we do carry trades,  
[00:47:19] we think about nominal exchange rates, not real exchange. Well, remember there was equation number four, which was monetary policy. Monetary policy is pretty good at stabilizing inflation rate. So real and nominal exchange rates are tied together by monetary policy.  
[00:47:32] That that would be equation number four. But I omitted it and I'm thinking about one concept of exchange rate.  
[00:47:37] In macro, we typically use these equations. That's more convenient. I guess in finance it's more convenient to think about risk sharing in in in in context of a UIP equation. But it turns out, you know, it's the same. It's it's the same risk sharing wedge between countries often times that that that in  
[00:47:52] both condition. Okay. And so the two propositions here, the first one is in order to get the right properties of exchange rates under the floating exchange rate regime, you need that this guy right here accounts for 80% roughly speaking of exchange rate volatility in  
[00:48:07] equilibrium. And we kind of walk through how you get different properties with correlations, very large very large volatility of exchange rate, right?  
[00:48:14] Because, you know, the spillover from exchange rate going back to this equation, uh I'm not seeing it very well, but I have one minus Okay. So what is one minus alpha? What is theta and what is gamma?  
[00:48:25] Gamma I told you it was the openness of the economy. One minus alpha is passed through from exchange rates into border prices, into import prices. And remember how terms of trade was much more stable than exchange rate. That suggests that  
[00:48:38] one minus alpha is a small elasticity pass through from exchange rates into border prices are not very high and this is where you need low fund price deviations, right? So this helps. And theta is the pass through from prices to quantities. And so if the product of  
[00:48:52] those blue terms is small, a lot of volatility in exchange rate, this 10-12% of volatility every year do not translate in a lot of macro volatility. You know, the the transmission of exchange rate into macro is muted by that blue term being small.  
[00:49:07] So this is why you can simultaneously match the disconnect in terms of correlations and in terms of volatility.  
[00:49:13] Okay, very good. But now the it turns out that once we reach this point, we realize that that that really doesn't tell you what is the nature of this shock, of the psi hat.  
[00:49:26] Um you you know, you have a variety of models.  
[00:49:29] Sometimes you can have convenience yield models where you put uh the bonds into the utility function.  
[00:49:35] Uh you can have uh financial friction models like Gabex majority. Uh, and many other papers in this literature. You can you can have risk-based models like intermediation base with segmented markets. This is the type of model that we adopted following  
[00:49:50] the older literature by uh, Schleifer and co-authors and Jean and Rose and then uh, it recently became also more popular the mark the segmentation with the uh, you know, portfolio choice of intermediaries. Um,  
[00:50:03] uh, then you can have models of heterogeneous beliefs or infrequent portfolio adjustment. All this model there is big literature on all of this topics and they all produce a PC shock in the end. They all produce the shock.  
[00:50:15] But without knowing what's the right model, you cannot go to the policy analysis because the shock is endogenous, right? You cannot really ask the question, is it optimal to peg or the float exchange rate without knowing the nature of this this shock? And so  
[00:50:28] the Musa puzzle gives you an entry way into thinking about it. It tells you that the only way and it's a whole paper and I will not be able to explain it very well, I think, in a couple of minutes. But basically what we prove in that paper is  
[00:50:41] that the only way you can conceptually um, reconcile those pictures about the Musa puzzle that exchange rate changed its property, but consumption didn't change its property is is is if this guy, the  
[00:50:55] UIP deviation, changes its properties between floating and fixed exchange rate regime. It becomes a lot more volatile under the floating exchange rate regime.  
[00:51:03] So opening up the possibility of exchange rate moving around creates bigger UIP deviations. Well, that's not very surprising perhaps like bigger bigger possible carry trade returns and also bigger possible carry trade losses, right? And so you need to write down a  
[00:51:16] model where this PC hat is endogenous to monetary regime.  
[00:51:20] A lot it excludes a lot of models, right? So if it's a model um, a purely real model where monetary regime does nothing and it doesn't work through risk, then that model will it would have a hard time generating a property of that UIP shock that changes  
[00:51:35] between float and a peg, right? Which is just a change in monetary policy. At the same time, if you write down models based on risk and market segmentation, it gives it very naturally to you. So what does a move from peg to float do?  
[00:51:48] It opens the door to exchange rate volatility. But exchange rate volatility is additional risk that intermediaries don't want to take. So they actually need to be more compensated for doing the same amount of intermediation. So it's either intermediation will go down or UIP deviations have to be bigger  
[00:52:03] under the float. And so as a result, you kind of in the reduced form get bigger UIP shocks in the model with segmented markets and um, uh, risk-based intermediation, risk constraints to intermediation. And so this is a different way of generating  
[00:52:17] monetary non-neutrality. Typically in macro, we get monetary non-neutrality through sticky prices. Here prices are not sticky. They're they could be sticky or flexible, it doesn't matter. Monetary non-neutrality comes through market segmentation because uh, the policy function of  
[00:52:31] intermediaries is different under the peg and under the float because the amount of exchange rate risk that they take on it's a fixed point, right? In equilibrium, how much there is exchange rate risk, but it changes between a peg and a float and as a result, it changes the policy function of intermediaries  
[00:52:46] and this is the source of uh, of non-neutrality, right? So why does a model without segmentation doesn't work?  
[00:52:54] Well, in a model without segmentation, the problem is that exchange rate risk will not be priced very much. Exchange rate risk is orthogonal to macro. So if you try to price it with a household SDF, you're not going to get much of a risk premium from it. It's just not  
[00:53:06] enough. But if you think about models of segmented markets and intermediation, exchange rate risk is held in a very concentrated way by a small subset of specialized currency traders and they value that risk, right? They value that  
[00:53:20] risk a lot a lot more than a typical household would, right? And so as a result, you get, you know, you can get a big kick here big enough that's could be consistent with this very low volatility of real exchange rate under  
[00:53:34] the peg and very high volatility under the float.  
[00:53:36] I actually know how to explain the Musa puzzle paper.  
[00:53:40] So let me go back. Marcus, can you give me another 10?  
[00:53:44] 10 minutes? Where's my picture?  
[00:53:50] Okay, so so this is interesting. So the question is which part of this picture is most puzzling?  
[00:53:56] Um, it's a question to the audience. So I hear the answer, it's the bottom left.  
[00:54:00] So why is it the bottom left picture that's most puzzling? And it's amazing because the literature went in somewhat the literature I I I was not there, but this is how I reconstruct how the literature went was like, okay, exchange rate became volatile.  
[00:54:15] But it should create this craziness in macro, right? How we as a central bankers will deal with all that exchange rate volatility? How will we stabilize inflation if exchange rate is so volatile? Wouldn't disrupt, you know, consumption and output and so on. And so  
[00:54:28] people really were worried about volatility here. They're like, if volatility opens up here, why doesn't it pass through into volatility of macro aggregates? And this turns out to be a very simple problem to solve. You just say countries are fairly closed. So if  
[00:54:41] you trade little, all that craziness in the exchange rate, well, it affects only a small part of the economy, the export and import sector, but you know, it's a small part of the economy. Then before exchange rate affects anything, it needs to affect prices and there is limited  
[00:54:56] pass-through into prices and then, you know, it needs to affect quantities a lot and there is limited kind of elasticity of quantities. And so as a result, it's a very partial equilibrium story how you go from a lot of volatility to not too much volatility here. So you solve it with that remember  
[00:55:10] that blue term being small is exactly why all this volatility doesn't translate here. So now the bigger question that was sort of omitted from the literature and this is sort of what we try the whole Musa puzzle paper is really about that is like why there was  
[00:55:23] not volatility here. And think about it, all this financial volatility, if we believe it's financial volatility, I tried to convince you that it is, right? That drives the exchange rate. Well, under the peg, the central bank must have offset the movements in  
[00:55:37] exchange rate with its policy rule. So it must have changed the interest rate rule to absorb all of the shocks here, right? The shocks they didn't disappear if the shocks didn't disappear, central bank is absorbing them and it's absorbing them by changing its monetary  
[00:55:51] policy. But if central bank is changing monetary policy, it needs to affect inflation or output gap or output eventually. Why don't we see any volatility that the central bank absorbs from here and pushes it into here?  
[00:56:03] Right? And so we don't. Why is that? Well, because UIP shocks sort of endogenously disappear. When you shift from float to the peg, UIP shock disappear because there is no risk anymore and intermediation is much more  
[00:56:17] frictionless now, right? The financial sector is willing to intermediate a lot more at lower UIP deviations and the fixed point of that is less exchange rate volatility without the need to change monetary policy rule very much,  
[00:56:30] right? And so one thing that we do in that paper, we also see how much volatility of interest rates changed.  
[00:56:35] And then says volatility of interest rates did not change all that much suggesting that it changed some, but suggesting that most of stabilization kind of happens endogenously almost through credibility of the peg. You need to change monetary rule very very  
[00:56:49] slightly. Uh, so this is what I wanted to tell you about the positive side of the models.  
[00:56:56] Then I have a couple of slides on the normative side.  
[00:57:00] I feel that if I show you equations, you will maybe hate me. Uh, so uh, I let me let me not show you equations and just show you pictures, right? At this point probably pictures are better.  
[00:57:12] Um, so basically once you wrote down this model and once you made once you made the PC shock endogenous uh, to the uh, uh, monetary policy regime, this gives you a very natural entry way into  
[00:57:27] thinking about optimal monetary policy because you know, the central bank now appreciates the fact that if it changes from peg to float, it changes something in the financial market and when you design the policy, you have to take it into account, right?  
[00:57:40] And so the way we're going to write down the model, I really I don't want to spend time here. This is really the expenditure switching condition, the goods market clearing expenditure switching condition that we sort of discussed, but we kind of know way to simplify it even further. Uh, it's kind  
[00:57:55] of here. But let me say two words about the financial market equilibrium condition right here. So this is the UIP deviation, expected UIP deviation in the numerator. This is the expected carry trade return.  
[00:58:06] Uh, this is the volatility of exchange rate. So this is the carry trade risk.  
[00:58:10] So sigma is volatility of exchange rate carry trade risk and omega is the risk aversion of uh, the currency trader. So this is how this um, segmented market uh, risk-based models look like. And  
[00:58:24] what's on the right hand side? Well, this is the position that the intermediary would like to have if this is the UIP deviation, but markets need to clear. So whatever the intermediary takes on, he needs to be on the other side of trade of the households or the  
[00:58:37] noise traders and of the central bank, right? That's the market clearing in the financial market. So when there is demand for dollars, uh, there is a size of the UIP deviation at which intermediaries would provide those dollars. The more there is  
[00:58:52] exchange rate risk, the bigger is the required UIP deviation to compel intermediaries to provide those dollars.  
[00:58:58] But if the central bank chooses to provide dollars, it would be the F here.  
[00:59:02] F would be the reserves of the central bank. Then intermediaries need to provide less dollars to the market, right? And the equilibrium UIP deviation can be smaller, right? And so this is the sense of this equilibrium condition in the financial market which we can rewrite it in uh, you know, risk sharing  
[00:59:16] terms like that. And so basically the size of the UIP deviation is the demand for currency that needs to be provided by intermediaries times the risk premium that is that that these intermediaries want to charge, which is omega risk  
[00:59:29] aversion times sigma squared, which is the amount of risk that they take on, right? And so this is a very finance approach to thinking about, you know, the nature of UAP shocks, right?  
[00:59:42] UAP shocks reflect the size of the position times the value of risk from the point of view of the intermediaries, right? And so now the central bank the central bank has two instruments. It can do monetary standard monetary policy,  
[00:59:56] so it can change the interest rate and affect output gap in the domestic economy or it can try to intervene in the in the in the exchange FX market and provide dollars in response to shocks  
[01:00:09] that require dollars. And so here is an illustration of how this model works.  
[01:00:15] Um so let's start from the conventional model. So conventional models I we can call them trilemma models or Mundell-Fleming uh models or new generations of Mundell-Fleming models going all the way to say Obstfeld-Rogoff and the uh  
[01:00:29] literature that was building on Obstfeld and Rogoff. This model say, well, I'm going to plot it in the space of output gap here, output gap and exchange rate volatility here.  
[01:00:41] And so basically the government can always choose to peg or to partially peg, reduce exchange rate volatility. It always comes at the cost of creating the output gap, right? You can choose to do something with your monetary policy  
[01:00:54] domestically, which will stabilize exchange rate. You will use your interest rate to offset exchange rate movements, but it's typically not going to be the efficient interest rate anymore and you will create an output gap. And so as a result, you are on this curve,  
[01:01:08] right? This is the set of choices that you have when you are in a trilemma style models in a Mundell-Fleming or Obstfeld-Rogoff or Dornbusch style models, right? And so then there is a very famous Friedman argument.  
[01:01:20] Uh we we call it a Friedman float. He says there is really no point in pegging the exchange rate. Pegging exchange rate is stupid. It only comes at the cost, right? Like we don't care about exchange rate per se. It's not part of our objective function. What we care about  
[01:01:33] is that there is no output gap. And floating exchange rate, you know, there is a particular notion of a float which guarantees that you don't get an output gap. You do optimal monetary policy from the point of view of inflation output gap trade-off  
[01:01:48] without thinking about exchange rate at all. And this would be the Friedman float rate here, right? And there is really no point to depreci- deviate from it. This is the efficient outcome and it's associated with some degree of exchange rate volatility, right? And  
[01:02:02] that's the equilibrium exchange rate volatility. That's good because we really don't care directly about the level or the volatility of exchange rate per se.  
[01:02:09] And of course, the policy maker can do the peg and go here and basically absorb all of this exchange rate volatility and put it into macro variables and end up in a bad in a bad point in a bad choice where you have a lot of distortion in  
[01:02:24] the goods market, but a fixed exchange rate. There is no use for from the fixed exchange rate. Of course, well, it turn I'm I'm I'm not sure of course or not, but this is a new result in our paper is that there is open economy divine coincidence. We typically talk about closed economy divine coincidence when  
[01:02:38] there is no trade-off between output gap and inflation. In an open economy there is an additional layer of divine coincidence and it's a point and the point is stated like that.  
[01:02:49] If the first best real exchange rate that supports efficient expenditure between tradeables and non-tradeables or between home and foreign good is stable, right? Imagine that the the first best real exchange rate does not need to  
[01:03:03] move. You can get no output gap, no distortion with a fixed real exchange rate.  
[01:03:09] Then there is no point in having floating nominal exchange rate. The only purpose for nominal exchange rate when it moves is to allow relative prices to change when prices are sticky. But the real first best real exchange rate doesn't need to move. It means that nominal exchange rate does not need to  
[01:03:22] move to accommodate that adjustment in relative prices. And as a result, the peg and the float are the same. You you you get to this origin point, right?  
[01:03:30] There is no trade-off between these two points. Whether you float or you peg, you're going to get the same outcome.  
[01:03:35] It's just like the divine coincidence in the closed economy. Whether you target output gap or you target inflation, you will get the same outcome under divine coin- And so it's a very difficult thing to check and it's also very implausible that the first best real exchange rate is stable. Why should the first best  
[01:03:49] thing I mean, there are examples of models where it can be the case as a knife-edge case, but in general we don't expect divine coincidence, the open economy version, to be really a relevant point of approximation maybe even. So we really think of most relevant as like  
[01:04:03] this trade-off in the trilemma models. But we don't like trilemma models because they result in exch- they they result in all these puzzles, right?  
[01:04:10] These are the models that result in puzzles. So how does in this space look a model that actually is consistent with the uh empirical properties of exchange?  
[01:04:18] Well, it turns out it looks like this. So check it out. So in in the model of exchange rates with, you know, segmented uh risk-based financial intermediation and a lot of, you know, noise trader shocks, it turns out if you do the peg, this point is still feasible. It's not  
[01:04:32] the point that you want, but it's still feasible. Why? Well, because that model is a little extreme. It basically says that if there is no exchange rate risk, intermediaries will do all the intermediation efficiently. It is it's it will be like a  
[01:04:45] uh a model with perfect supply of currencies, right? If you want dollars, you can get dollars. If you want Deutsche Mark, well, you can't anymore, but you can get euros. And basically as a result, this point is still feasible.  
[01:04:57] So you completely kill off all financial volatility if you promise a peg. And it's enough to just promise a peg and then financial market will do all that intermediation frictionlessly. It's a little unrealistic, but that could be an interesting point of approximation. Then  
[01:05:11] of course, you can say, well, there is a little bit of intermediation friction even if exchange rate is not moving around. And then you're going to get into a point which is somewhat to the right, right? But then what happens then? So this is a model of financial amplification. You depart a little bit  
[01:05:26] from a fixed exchange rate and suddenly endogenously the exchange rate volatility keeps going up. It's not the trade-off as in the Mandel model in the Mundell-Fleming model or Obstfeld-Rogoff. It's it's a trade-off that looks like this. The more you allow  
[01:05:39] exchange rate to move, the less intermediaries want to intermediate without further compensation for risk, which in itself creates a additional shock to um to exchange rate, right? The shocks get  
[01:05:52] amplified, right? So basically you want dollars, you try to come to intermediaries to get dollars and they say, well, you'll have to pay a bigger premium, which in itself is like a shock, right? In this model. So as a result, it acts like an amplification device. And so when you go to a complete  
[01:06:07] float where you close the output gap, well, it turns out you end up with this amount of exchange rate volatility.  
[01:06:13] We'll do all that financial volatility that you let in, right? Because it's it's a fixed point, right? Like if the in equilibrium there was less exchange rate volatility, right? Uh then the UAP deviation would have been smaller, but you find a fixed point where it's consistent with the behavior  
[01:06:27] of intermediaries and you kind of end up there, right? And so basically now the choice is not between this and this point. You still want to be where Friedm- where Friedman told us we want to be. We still want to get there. Only keep fundamental exchange rate volatility, but it's not feasible  
[01:06:41] anymore, right? You can be either here or you can be there.  
[01:06:45] And that point is not exactly feasible. Well, it turns out if the government can take on the role of intermediary, if the government becomes an agent that provides dollars when people want dollars, right? In in in in a country and provides euros when people want  
[01:07:00] euros, it's possible to reduce the red region. So doing open market operations allows you to reduce red region and get closer to the blue point, which is still the first best, right? That's still the optimal allocation. But in general,  
[01:07:13] without having enough reserves to fully offset demand for currencies of everybody, this is really would have implemented something like a Friedman rule in the closed economy an open economy version of it. It's still the case that the blue point is not feasible and you have to choose a point along  
[01:07:28] this red curve. And if you are very open economy, well, maybe you want to shift if you're very open economy, which way do you want to shift? Let me think about it. If you're very closed economy, uh you don't care about exchange rate from the point of view of the goods  
[01:07:43] market and that point is good for a very closed economy. The more open you are, the more exchange rate volatility is useful from the point of view of the goods market and you want to kind of shift in that direction. So regions that trade in the goods market less with each other,  
[01:07:56] um yeah, so so so there is a well, I don't want to get into that.  
[01:08:02] Yeah, let's yeah. So let me let me stop here. So basically you you want to choose a point here and like openness would be one of the parameters that makes you move along along that curve.  
[01:08:11] So now let me tell you what happens in models where financial shocks are indoge- exogenous. Imagine you have a model where financial shocks are present, but they're not endogenous to monetary policy regime. So then you're going to get this. You This models will  
[01:08:24] will will agree with our model in this point. They will be able to replicate all the moments in the floating regime. They will be consistent with exchange rate disconnect in the floating regime. But the problem is is that the central bank when it will try to peg  
[01:08:38] the this in endogenous financial amplification will not quite work. What will happen is that the central bank will need to absorb using the monetary policy rule all of that crazy financial volatility. And so that point is the Musa puzzle kind of point. It's the  
[01:08:50] point where under the peg, you get a lot of macro volatility, which we didn't see in the data, right? And so this is kind of closes the loop between uh you know, the policy analysis and the first Musa puzzle picture uh that I showed you. I  
[01:09:03] think I'm pretty much done. Uh yeah, so I have a a bunch of policy implications here. So we talked uh we talked about a few of them already and I will wrap up in a in a moment. So what this model really predicts  
[01:09:17] is that there is a very strong incentive for a lot of countries to do partial pegs, right? You can do you can offset some of the exchange rate volatility using open market operations. And a lot of countries do that. There is a question  
[01:09:30] why Europe doesn't do open market operations to stabilize the euro against the dollar and the argument there could be that the market is so deep that financial there are lots of financial intermediaries there, but also lots of noise traders. And you would  
[01:09:45] need a crazy size of the balance sheet of the central bank to do it. So it becomes basically easier to let the exchange rate float given fairly efficient amount of financial intermediation. But at the same time, if we look outside the developed countries,  
[01:09:59] pretty much most of the developing countries choose to peg or crawling peg or dirty peg or or dirty float or something like that.  
[01:10:08] So like most non non-developed countries actually do some form of a peg's exchange rate regime and it actually really does look like choosing a point on this red curve somewhere there, right? You don't go to the full float, but you also don't go to the full peg.  
[01:10:21] Of course, there are a bunch of countries that do a full peg, right? And so the answer here is the divine coincidence. If you're closer to divine coincidence where you don't want your real exchange rate with your trade partners to move very much in in in the first best, then there is not much of a  
[01:10:34] cost to a peg and there is a financial benefit to a peg. Okay, so I will stop here. I don't want to talk about that.  
[01:10:45] Oh, okay. Thank you. Yeah, so let me pick maybe one more policy to talk about. So it turns out Oh, let me talk about trilemma for a second. So the model that I showed you does not have a trilemma constraint. Why is that?  
[01:10:58] Well, whenever you have an additional friction, it gives you an additional instrument, right? So sticky prices gives you interest rate as a tool to deal with output gap, right? Without sticky prices, interest rate will not be affecting real variables, right? So  
[01:11:11] sticky prices is something that gives the policy maker an access to a new tool, the interest rate. Same thing here, financial friction gives the government access to a new tool, which is the effects interventions. The effects interventions in a conventional model do nothing, right? You basically  
[01:11:26] have a Modigliani-Miller result between households and the government. Whatever the whatever the government does with its portfolio is can be undone by the households. It's a very Modigliani-Miller style result. But with the segmented markets, Modigliani-Miller doesn't hold. As a result, the  
[01:11:40] government has access to a new policy tool, which is the effects intervention.  
[01:11:44] And this allows to relax the trilemma constraint on policy. So you can you don't have to compromise the domestic output gap and inflation stabilization if you want to partially peg the exchange rate using the effects intervention tool.  
[01:11:57] But effects interventions cannot fully resolve the problem. You cannot replicate any passive exchange exchange rate has to be passive exchange rate has to be consistent with the budget constraint of a country. And so the only thing that you can do with these effects interventions, you can kind of smooth out the passive exchanges. You cannot  
[01:12:12] permanently adopt a different level of exchange rate. That policy will result in you know, budget constraint being violated eventually. But what you can do is you can smooth the passive exchanges with effects interventions. And this eliminates part of the financial  
[01:12:26] volatility from exchange rate and this is really what the central bank wants to do. It wants to keep fundamental volatility in the exchange rate and get rid of to the extent possible of the you know, this noise non-fundamental volatility in the exchange rate, which will improve risk sharing, right? If  
[01:12:40] there is less less noise volatility. So this is one thing I was going to say.  
[01:12:45] Yeah, there is this very interesting result is Okay, so you are UK. You see So how do we interpret the UK back when when when was it? September when there was a big shock to the the pound. So how do we interpret it? Well, there is some  
[01:13:00] uncertainty that's created, right? Uncertainty about the future. People want to get get rid of pounds and buy foreign currency. It's a capital outflow shock. But the financial sector also see more uncertainty going forward. So it's the it's the policy function of the  
[01:13:13] financial sector changes as well. So people are trying to exchange pounds for foreign currency, move move assets from UK to other countries. They come to intermediaries, but intermediaries also pull out at the same time because there is more uncertainty about the future.  
[01:13:27] This is why pound depreciates so much, UAP deviations open up. Why is that?  
[01:13:32] Well, the pound needs to depreciate to a point where the expected UAP expected carry trade gains for the intermediaries compensate the increased risk, right?  
[01:13:42] Can the policy maker stop it? Can the policy maker raise the interest rate to stop the capital outflow? Well, our theory actually says that no. You will be you can effectively change the value of exchange rate, but over all of it will go into output gap. What happens  
[01:13:57] from the point of view of the carry traders, right? They really think about expected return and they think about the amount of risk. And doing a policy shift today does not to a first order approximation basically does not change  
[01:14:10] either the expected return or the amount of risk. It's a little I need to have better words to explain it. It's a result.  
[01:14:17] I think it's true, but I'm not giving you a great intuition right now. But the point is what the policy maker can do is really reduce amount of uncertainty tomorrow.  
[01:14:25] Not raise the interest rates today and tighten the output gap. That will help you with exchange rate. You would look at exchange rate, you see like look, I stopped depreciation, but it really will not help you with the capital outflow.  
[01:14:35] And the problem is the capital outflow. And how do you deal with the capital outflow? Well, you deal by stabilizing the future by either promising to have a partial peg in the future, not create a lot of exchange rate volatility in the future, or by reducing  
[01:14:48] uncertainty about policies, right? And in that sense, you don't want to raise the interest rate today, but you want to create conditions that kind of make intermediation less costly, right? Less risky, right? And this is one way of doing the policy. The other way is just  
[01:15:02] step in and provide those dollars, right? To the market if they want it and eliminate the volatility in that way. So it's either the effects tool or some tool about future stabilization of exchange rate using in principle monetary policy as well. So let me stop here. There are a bunch of  
[01:15:17] different things to discuss, but let me go to my conclusion here. So basically what we tried to do is you know, build a framework that's realistic in the sense that it's qualitatively and quantitatively consistent with properties of exchange rates. And at the same time, we want it to be tractable  
[01:15:32] enough so we can talk about policies. And as you saw, most of it was actually analytical. We didn't need to go to a quantitative model here very much. But then we also wanted to be practical in the sense that we can revisit a lot of this policy prescriptions that were you  
[01:15:46] know, kind of robust in the previous literature, but now you know, if you have this new financial shocks based model of exchange rate which is endogenous to monetary policy regimes, some of the answers actually change, right? And it makes it clear  
[01:16:01] you know, what are the tradeoffs of a peg a little more than before and sort of what are the costs and benefits of a currency union for example as well. And going forward, and this is my last bullet point here. So going forward, I think it just opens up the door to a lot  
[01:16:15] of research. Part of it is in macro, a large part of it is in finance. So in particular, we spend a lot of time in the goods market measuring markups, right? Markups has become a popular topic. But it arguably we spend not enough time measuring UAP deviations and  
[01:16:29] different components of UAP deviations. Because in order to really do the policy, you need to know not the value of exchange rates and interest rates. You really need to be able to have a good proxy for UAP deviations. CIP deviations are easy to  
[01:16:42] measure roughly speaking, but UAP deviations are hard to measure. And you want to know different components of UAP deviation. What part of it is markup, what part of it is a financial friction, what part of it is you know, segmented markets and  
[01:16:55] risk-based premia, right? And so can it There is already a lot of work, but doing even more will really help the policy makers to know how to do the optimal effects interventions, right? So far, we didn't get in this area as far as you know, in the goods market in  
[01:17:09] measuring you know, natural rate of interest, natural rate of unemployment, output gap. These have become a very conventional objects, but if you want to do effects intervention policies, you really need to do the same dissection to UAP, right? And this is an input into  
[01:17:21] the policy problem of the of the central bank.  
[01:17:26] Yeah, so this is what I want to Ah yeah, yeah.  
[01:17:29] So So there is a huge literature now measuring elasticities in the financial market, right? You know, the supply of different assets when the elasticity is high and low, what makes that elasticity change?  
[01:17:39] The big question is what is the nature of that financial shock? Where do these noise traders come from? We have no idea. I mean I know that in finance people probably tried very hard already in the past to figure out where those noise trader shocks come from, but this  
[01:17:53] is you know, for thinking about optimal policy, knowing where the noise trader shocks is coming from could be crucial.  
[01:17:58] Could it be that fundamental macroeconomic shocks generate financial shocks? What's the relationship between them and so on? And finally, obviously whatever is done for the open economy and exchange rates can be transplanted to close the economy and the stock  
[01:18:12] market, right? If if you want to stabilize the exchange rate, is there case to stabilize in the stock market in a in in a similar way or maybe stock market is different. And so there is interesting work in that area as well.  
[01:18:22] And so let me stop here. Well, thanks a lot Oleg. Let me just say a few words.  
[01:18:36] And you have time to go to the mic so there's Q&A. So if you want to ask a question, please go to the mic.  
[01:18:42] So thanks a lot for having this great lecture where you start out with the empirical puzzles and then provide an analytical framework how to think about it and came to a conclusion that actually what matters is really finance  
[01:18:55] for exchange rates. It's all about finance primarily and it's about segmented finance, you know, financial frictions.  
[01:19:03] And of course, we we are very pleased to hear that. And it's in a sense, it's it's a gift to us in the American Finance Association from Oleg, but I just was told by Philip Lane that actually we should give Oleg a gift today because today is his 40th  
[01:19:17] birthday. Is this correct? So let's give him a applause for his daughter as well.  
[01:19:25] Um so let's uh move on to the questions. So the microphones are here. Oleg, you have to come back.  
[01:19:32] Thank you. Um Oleg, I think I have a question for you to which I think I should know the answer, but I'll ask it anyway.  
[01:19:39] If I take you over to the United Kingdom before the September shock, interest rates there were lower than they are than they were here. And so interest rate parity, CIP, covered interest rate parity, seemed to be holding at the time where the FX rate there indicating a  
[01:19:53] strengthening of the sterling. This was at the same time that inflation futures in the UK were higher than here, which would indicate that we would look for weakness in the sterling. So is this a puzzle or is this simply the difference  
[01:20:06] between CIP and UIP? Yeah, yeah, yeah. I I don't have a great answer. I think it's at one of those questions that should be out there on the list, the relationship between UIP and CAP. Uh we we we have models which are models of  
[01:20:21] frictional intermediation that will give rise to CAP and UIP simultaneously, kind of in the same direction. It's like the balance sheet cost models. Uh we have models about risk which are really about UIP and not so much about CAP. Uh in  
[01:20:34] practice, I think it's an open question to what extent UIP and CAP comove.  
[01:20:39] Um uh is it are they of the same nature? We know that they don't comove perfectly. We know that you know, the reason why I focused on UIP much more than on the CAP is that UIP is a deviation as a persistent feature of the  
[01:20:52] data both uh during pegs, during floats, before 2008, after 2009. And there before 2008, there were two orders of magnitude bigger than CAP deviations.  
[01:21:03] Now they're about one order of magnitude bigger. So in terms of just, you know, magnitudes if we um measure them, they're about 200 basis points versus 20, 30, 40 basis points.  
[01:21:13] And so if if we wanted to build a model of just one object, we decided to go with UIP. Right? In fact, then it turns out this is for Musa puzzle, this was essential because risk played a role, right? But really making progress here, you want to write down a model which jo  
[01:21:27] can allows you to jointly think about UIP and CAP and to isolate a common component of the two which comes from the balance sheet cost, financial frictions, and so on. Uh and the component that's maybe peculiar to only CAP or to only UIP. And so, you know,  
[01:21:42] this is like a research agenda going forward. I really honestly don't have much to say uh uh about this. Yeah.  
[01:21:58] Uh I just have a question on your uh data.  
[01:22:01] Um I'm wondering where you get your foreign exchange rate data before the peg and after the peg.  
[01:22:09] Because it's it's is it something that was available in the market or is it's not like there's a central bank who um who announced a a an interest rate, the exchange rate  
[01:22:23] before the peg during the peg and after the peg. And so you mean Where did you get that information?  
[01:22:29] The question is about 1973 about the end of Bretton Woods. before or after the peg? During the peg or after the peg?  
[01:22:35] Where did you get the information on the foreign exchange rate?  
[01:22:39] Well, macroeconomic data exists. So we just use, you know, output, consumption, inflation, exchange rates. All of the data goes back to like '50s, '60s fairly easily. And so a seven It's uh published  
[01:22:52] by the federal by the government. Some of this data is, for example, agency.  
[01:22:58] Yeah, for example, OECD data, right? Uh would go for, you know, major developed countries, it will go back to like '60s basically. But in our Musa puzzle paper, we have a data appendix where we specify which data sources we use. It It's true,  
[01:23:12] it's a bit of a problem going that back and in order to study the pre-Bretton the Bretton Woods period, you really need data from before '73. It becomes trickier. And so a lot of the questions that have to do with the micro level data. If you want to know the balance  
[01:23:25] sheets of individual financial institutions, that data is not available. So a lot of theories the way they you want to test theories and look at the balance sheet objects of different intermediaries, right? And to do that, you really need a modern  
[01:23:39] period. Uh and you need to figure out episodes of the pegs like Switzerland gives a very interesting episode of a brief peg uh back in 2014, '15. And so we use a little bit that data if you want to look at, you know, the micro data. But macro aggregates are available  
[01:23:54] going back to '60s. Oleg, I think I'm trying to get to the your point that volatility in the foreign exchange market uh of is purely financial um  
[01:24:07] phenomenon. Because even before the peg or after the peg, it makes no difference what the foreign exchange rate, that really that kind of transaction uh takes place at  
[01:24:20] between the the good markets, the producers of goods and services. And and the the I think I think that's what you're trying to say. Yeah.  
[01:24:31] paper, there's some misallocation in the goods market. And and that is translated into the foreign exchange rate market. And um and that you're arguing that if the central bank  
[01:24:45] were to interme uh to to provide the uh the uh the intermediary function in the exchange rate market the same way that they do in the loans market, then it will smooth the  
[01:25:00] volatility in the exchange rate market. But well, my point is that they that um it's a purely financial uh phenomenon.  
[01:25:10] I I guess that's why I'm I'm asking where the data comes from because Yeah.  
[01:25:14] the data is is in the foreign exchange rate, it's like a an announcement. So I'm not sure if I interpret your question correctly, but let me try. This is uh So So the way we discipline our analysis was largely  
[01:25:27] common from simple unconditional macro moments during different subs subsamples during the peg and during the float. So we only look at macro comovement to discipline it largely. So now, if you really wanted to get at the  
[01:25:40] heart of the theory, you really want to go to micro data on the positions of intermediaries and see whether that So So for example, size of UIP deviations we can measure using micro data or average returns on CAD trades and so on.  
[01:25:53] But if you really want to ask the question, did intermediaries change their policy function? In the loans market, there's regulation that the rate the rate of Yeah, so borrowing is is very heavily regulated, but I don't  
[01:26:07] think it's the same in the foreign exchange rate market, and that's where you see a lot of the Yeah. So So looking at that data specifically, we didn't do, but this like looking at the micro level where the behavior is changing, I think this is really the direct test of the theory. It's It's much harder to do  
[01:26:22] than, you know, that type of empirical analysis. In fact, I travel myself, and I never, you know, whatever exchange rate people say it is, we never I never get it because they they do they will sell and buy currency however  
[01:26:36] whatever amount of money they want. It It's It's not It's not fixed.  
[01:26:42] I mean, it is kind of fixed, but it's not.  
[01:26:48] So uh thank you very much for your very interesting presentation. Uh I was not familiar with your work, but I I I looked a little bit into the exchange rates recently. And I wanted to ask you this question. If I understood  
[01:27:02] well your charts about the Musa puzzle, um so you don't find any relationship between increased volatility of the exchange rate and um  
[01:27:16] increased volatility there is no increase in volatility of consumption.  
[01:27:21] Uh my question is, but have you tested or do you think it would be interesting to test uh whether there is a relationship between the increase in volatility of the exchange rate  
[01:27:34] and the trends in real variables such as consumption because if you think of it uh historically, uh for many countries, the the the the  
[01:27:48] shift from fixed to floating exchange rate coincided with the change in the trend growth rate. Mhm. Mhm. So that is uh with my question, basically.  
[01:28:00] Yeah. So Baxter Stockman, uh so their original paper in 1989 really looked at the changes in trend shifts in trend growth rates between before '73 and after '73. We focused on second moments  
[01:28:14] because macro models, like we wanted the business cycle frequency comovement. And that's really not so much about the trends, but more about the comovement.  
[01:28:23] And our point is not that there was absolutely no change in '73. Our point was it's an order of magnitude difference. You look at exchange rate, and there was truly an order of magnitude increase in volatility. While you look at any macro variable, at most  
[01:28:36] the behavior there changes by 5%. So it's a change, you know, 10 times, the 1,000% versus a change of 5, 10, 15%.  
[01:28:45] And so this is really you don't really need a very precise statistical tools because of that because there is such a gap, right? And so identification comes from that. And that's why we chose to focus on, you know, second moments because it just gives you a very clear  
[01:28:59] identification. It We're not trying to suggest that there was no other change before '73 and after '73. Plenty of things changed. It's just this stark discontinuity and order of magnitude that gives us identification. Yeah.  
[01:29:12] Thank you. Hi Oleg, thanks very much for your presentation. It's really interesting and I'm also relatively new to the literature, so it's very insightful for  
[01:29:26] me to kind of see how all of these different puzzles link together and your kind of approach to trying to solve them. One of my takeaways I think from your presentation was that financial frictions probably play quite an important role in being able to help  
[01:29:40] explain the like a number of these puzzles simultaneously. I was wondering how much you've looked at kind of seeing how well your model kind of explains the data when looking at sort of normal times versus crisis episodes and in  
[01:29:54] particular I was thinking also about those macro charts that you were showing with the sort of exchange rate correlations and um do you find that actually these relationships come with a lot more let's say in episodes of stress, periods of  
[01:30:08] high uncertainty? Yeah, yeah, yeah. Now, I think I understood your question correctly, but let let this is my answer here. So, people often times think about peculiar events like 2008-2009 crisis  
[01:30:22] and like there is no disconnect there, right? There was a economic crisis associated with a steep appreciation of the dollar against many many currencies, right? So, it seems like a very much connected event. When we talk about disconnect, we really think about  
[01:30:36] unconditional moments over broad periods of time. And so, it turns out that outside those unique episodes where it's a very clear crisis, right? And we know the nature of the crisis.  
[01:30:47] Typically conditional on those events, there is no disconnect, right? But most of the time there is no crisis. Most of the time exchange rate still moves like a very volatile variable and most of the time there is no co-movement, right? And this is why there is a disconnect, right? So,  
[01:31:01] like you can find a lot of episodes when exchange rate moves by 3 4 5% without anything happening in the macroeconomy easily and this accumulates over the year to like this big volatility of the exchange rate. And so, in that sense, of  
[01:31:14] course, periods of sharp crisis they are driven by a common macro force. So, all stock all asset prices, exchange rate, a lot of macro variables move together, but conditional on some big macro event happening, right? Same thing happens  
[01:31:28] when the central bank raises interest rates in a major way, then it affects all asset prices, yield curve, it affects the exchange rate, it affects the macro aggregate. It just turns out if you do variance decomposition, those episodes do not account for the bulk of the volatility and as a result  
[01:31:43] unconditionally you get you get low correlations, right? But of course, thinking about what is the shock and transmission mechanism conditional on those episodes is very important as well. We're not making much, you know, much progress here, but you know, models with financial frictions would matter  
[01:31:57] there as well for sure. So, just one very quick follow up if I may. Would you then caution against trying to use your modeling framework to sort of try and explain sort of specific episodes that are happening, you know, in kind of like  
[01:32:11] like say the Brexit event for example that's been talked about or the most recent uncertainty in the UK.  
[01:32:17] Yeah, so I think this is very interesting and I'm sure that it will fit some of the episodes and maybe not fit others and then you need to think about additional ingredients. What was surprising to us that we really didn't need to change the basic transmission  
[01:32:30] mechanism for thinking about the bulk of the variation as long as you acknowledge the role of this financial financial shocks. But like for example, the couple examples I gave you the Brexit is very interesting to think through our policy framework. I think we  
[01:32:44] have something to say there and it kind of fits pretty well. Not the sorry, not Brexit, the the shock the pound shock in September. Also, the effect of sanctions on the ruble and Russian economy was another example that we did where, you  
[01:32:57] know, you you use a very particular episode of a set of shocks and the same model and it actually gives you something intelligible about what has happened and it can allow you to explain why there was a steep depreciation of  
[01:33:10] the ruble at first followed by a steep appreciation and then like it can predict what we expect to happen with the ruble going forward for example, right? So, these are examples, but yeah, no, I'm sure it's possible to find examples where it works well and we'll  
[01:33:24] focus on those and it's possible to find examples when it works poorly.  
[01:33:28] We have a few ideas and then somebody needs to write a better model. Yeah.  
[01:33:33] Thanks very much. Let me abuse my position as chair of the session and also ask Oleg a question.  
[01:33:44] So, is it correct that you said, you know, it's mostly driven by financial frictions, so the emphasis whether we have a price stickiness in consumer prices, producer prices or  
[01:33:57] you know, in dollar prices is not really the driving, so it's not so important or is this more important for other questions?  
[01:34:05] Not for the exchange rate volatility. And then I have another question, you know, what we had in the 1980s, we had this huge long-lasting dollar appreciation.  
[01:34:15] Is your framework able to say something about that too or is this something which is, you know, related more to bubbles and dollar bubbles essentially everybody talked about that time.  
[01:34:29] So, I don't know if everybody noticed, but the first one was a very charged question.  
[01:34:34] Um well, it turns out that we wanted to be provocative in this research and we wanted to kind of say you know, all these years of thinking about sticky prices, which way prices are sticky to the first order, it's not  
[01:34:47] that important for these big questions of PPP and so on. And so, it's a first order approximation. It turns out both me and Dima we have separate research agenda on thinking about sticky prices and how sticky prices in different currencies affect things. So,  
[01:35:01] it kind of maybe being very provocative here sort of might seem to devalue that agenda, but you know, it turns out that the way price stickiness and the way prices are sticky is very important for a lot of other questions. It's what we're saying here  
[01:35:15] to first order approximation, if you want to write a disconnect model, what happens at the border is not as important if countries are sufficiently close to trade. You can just ignore that. Of course, as countries become more open to trade the way prices are  
[01:35:30] sticky and in which currency the prices are sticky at the border becomes more and more important. So, if you're trying to understand, you know, UK or New Zealand, which are much more open countries than US, the Euro area or Japan, then these details are really useful and  
[01:35:44] without sticky prices you're going to have hard time matching some of the moments. So, sticky prices do help and the recent developments, you know, in thinking that prices are sticky actually in dollars or in euros, but not in producer or consumer currency become  
[01:35:57] very important then. And so, for a lot of smaller countries, which is the majority out there, right? You cannot really ignore it, but if you really wanted to write a kind of first approximation theory of US versus Euro Eurozone as a whole, then what happens  
[01:36:12] at the border is just not of, you know, of of that huge of importance and if even if prices were flexible, you'd be able to match most of the moments, right? Because, you know, both of these regions are fairly closed, right? Um the second question, I have not thought  
[01:36:26] about it's very interesting to think about, you know, the '80s um the depreciation that happened and like the Plaza Accord, what role the Plaza Accord played. We have not thought about it as much. Um  
[01:36:41] Yeah, no, it's an interesting question. So, one question that we had, I don't have an answer is like imagine there is a financially driven appreciation. So, say you're Switzerland and there is a huge demand for Swiss safe assets, which causes this huge appreciation. How much  
[01:36:54] does it affect your macroeconomy, your, you know, your output, employment, growth prospects and so on, right? Is there a connect? And so, what we emphasize is that at high frequency there could be a lot of disconnect. High frequency up to 5 years, right? But also  
[01:37:08] there is research that suggests that at at longer horizon, 5 years and onwards, there is a very strong relationship between exchange rates and trade balances, for example, right? At least for the United States. And so, it always happened that a dollar appreciation or  
[01:37:21] depreciation 5 years later followed with a big adjustment in in the trade balance. So, the question is is it also true for Switzerland, for example, the fact, you know, that there was this big appreciation, did it cause a persistent trade deficit that will last for many  
[01:37:36] years, right? Is Switzerland becoming more like US in terms of net foreign asset position and so on. So, we don't have answers here. I think this are fascinating questions, yeah. We just didn't have time to think about it, yeah.  
[01:37:47] I have one more. Finally, another provocative question.  
[01:37:52] Would you argue that the financial sector in the effects market is too small? We should actually from our social welfare perspective, we should actually, you know, subsidize arbitragers in the effects market in order to get the exchange rate  
[01:38:06] volatility down. Is did you do any welfare analysis on this?  
[01:38:12] Yes, now I I I see where Marcus is coming from with this one.  
[01:38:15] Uh do you need to recapitalize sort of not just banks, but intermediaries as well. This is the topic so, Pierre-Olivier in the IMF, right?  
[01:38:23] They're thinking hard about this. Uh it would require an extension to the model and I guess you can write the model in different ways, but the idea is if the central bank doesn't want to have a lot of reserves and do a lot of effects interventions, can it make sure that  
[01:38:36] financial sector is capitalized well enough so that intermediation is more smooth, right? And so, kind of the idea there is you want to leave some UIP deviations open, some carry trade profits there, so that would be a source  
[01:38:51] of profits for the financial sector. Out of these profits, the financial sector will build net worth and so, will be more efficient on intermediation. And so, if you completely eliminate all UIP deviations and carry trade return, then you will eliminate the financial sector  
[01:39:05] that's doing intermediation and that's probably not good. And so, the question is is there like a internal optimality condition that you want to leave some of this markup. Basically, do you want markup of zero in that sector or do you want a positive markup that results in  
[01:39:18] profits? And so, I mean, they're they're thinking very actively about this, but you would need like some extra some extra ingredients in the model.  
[01:39:27] Yeah, which we didn't have. Okay, thanks a lot again Oleks for fantastic lecture.  

## Q&A (01:39:38 – 01:39:58)

[01:39:38] And then thank you all so for for all the questions from the floor and also for participating and being part of it.  
[01:39:45] I hope to see you in the afternoon and tomorrow. Bye-bye.  
[01:39:50] Thanks so much, Oleks.