Transcript of the talk, from a Whisper transcription of the audio. Auto-generated: speaker names in particular are unreliable.
========================================================================

# A Macroeconomic Model of the Cross-Section of Currencies

Authors: Aleksei Oskolkov, Diego Perez  
Discussant: Charles Engel  
Video: https://www.youtube.com/watch?v=wEt4iKiNNZQ&t=4756s


## Talk (01:19:16 – 02:09:04)

[01:19:19] Yeah.  
[01:19:24] We're including us into the program.  
[01:20:01] This is a,  
[01:20:03] this clicker works has its own life, right?  
[01:20:06] So,  
[01:20:14] So,  
[01:20:15] academic model, the cross section of currencies joint work with Diego.  
[01:20:20] This was on the first slide. I apologize. I cannot go back to the first slide.  
[01:20:24] We're going to start with the second slide here.  
[01:20:27] We're asking this very old question, what is driving exchange rates.  
[01:20:30] We're going to focus on the cross section and ask the question what is driving the cross section of bilateral exchange rates.  
[01:20:36] There's been a lot of progress on this in the international finance literature. This literature documents that only two global components explain a great deal of variation in exchange rates if you allow for a country specific exposures to those common components.  
[01:20:52] So, you know, the dollar factor and the carry factor, you might have heard about them.  
[01:20:56] Adrian has done in particular a lot of work on this in the past.  
[01:21:00] Well, the other hand in the international macro literature, there's been a lot of recent advances into mapping fundamental macro shocks into exchange rates and explaining the currency dynamics with macro shocks.  
[01:21:12] We're going to merge these two approaches. So we're going to take the advances in the recent international macro literature to the cross section.  
[01:21:21] And we're going to achieve three goals. We're going to try to explain from the through the lens of a macro model from the macro shocks, what those common components and exchange rates are.  
[01:21:31] We're going to try to explain for every country in our data, what drives exchange rates from the macro shocks, and then we're going to try to bring this new cross sectional moments into macro and see what we can learn from the new cross sectional moments about the macro economic drivers of exchange rates.  
[01:22:13] So, this is going to have multiple countries because it's a, he's going to focus on a cross section.  
[01:22:20] The multiple countries will be heterogeneous in two dimensions.  
[01:22:24] This heterogeneity is permanent and exogenous.  
[01:22:27] This is going to be heterogeneity in the global assets of those countries that they have accumulated before and heterogeneity in their exposure to the commodity cycle.  
[01:22:38] So, the macro literature, as I said, specifically modeling financial markets as segmented and driven with service by global financial intermediaries in the middle of the financial system.  
[01:22:50] The macro shocks that I promised are going to map into currencies. I want you to focus on two dichotomies here, really.  
[01:22:58] One dichotomy is real versus financial shocks. So what is driving your currency?  
[01:23:03] To what extent is it driven by real shocks, shocks to output or financial shocks, something that comes from the financial markets.  
[01:23:10] The second dichotomy I want you to focus on is this local versus global shocks. To what extent is your currency driven by something that happens inside your country, or something that comes from abroad, something that is a global shock.  
[01:23:24] The exchange rates fall out into a two-factor structure. Those two common components from the previous slide are going to be here in the model. Those are going to be the dollar and the commodity factor.  
[01:23:37] For now, it's enough to know that the dollar is just how much the US dollar is moving against all other currencies on average.  
[01:23:44] And the commodity factor is how much the high-commodity countries' currencies move against currencies of countries with low-commodity exposure.  
[01:23:53] So the expectation of those factors, there's a mapping to them from the shocks. The dollar factor is going to be made of US output shocks, the US business cycle, and then the shocks to repricing the risk coming out of the US business cycle.  
[01:24:07] The commodity factor is going to be made of global output shocks or global business cycle, and then repricing of the risk coming out of the global business cycle.  
[01:24:16] So both of these factors have a real and a financial component, and I'm going to show you how they compose the factors.  
[01:24:26] It's going to help the people with exposure to them. Now, we take the model to the data where in the data, these two factors explain more than half of the variation in the panel of currencies.  
[01:24:36] This is also going to happen in the model, and then the loadings, the heterogeneous exposures to those factors in the data are correlated with some observables of the countries.  
[01:24:46] In particular, the loadings on the dollar factor are correlated with countries' dollar assets. The loadings on the commodity factor are correlated with countries' commodity exposure.  
[01:24:56] This is not a coincidence that we put them in the model. We put them in the model for that specific reason.  
[01:25:01] Those are the two sources of heterogeneity between countries that we fix in advance.  
[01:25:08] I'm going to tell you something about the principle. What drives the exchange rate of this country?  
[01:25:12] The main takeaway is that there's no single answer to the question of what drives your exchange rate. It depends on the country's fundamentals.  
[01:25:19] It's in principle a very large spectrum of things, but you go a long way if you know these two things about a country.  
[01:25:26] It's dollar assets and it's commodity exposure. Then you can be pretty confident in some various decompositions.  
[01:25:45] I'm going to talk about the way we point out where we contribute to specific gaps.  
[01:25:59] The model is a total economy model, multiple countries. In fact, I'm going to have a continuum of countries. They're indexed by I, and then there's the United States.  
[01:26:07] It's a special country. It's a standalone mass point in contrast to those points in the continuum.  
[01:26:13] It's going to be very similar in almost everything, and I'm going to point out the one small thing in which it's not exactly the same.  
[01:26:21] In every country, there are two agents. I'm going to refer to them as the household and the central bank.  
[01:26:26] Let's start with the household. It solves a standard model, utility, maximization.  
[01:26:31] The present discounted stream of utilities. The household consumes two goods.  
[01:26:35] One of them is a non-traded good that is local to country I. It doesn't leave the country I. It's consumed inside.  
[01:26:42] The other one is a global tradeable good. There's only one global tradeable good, so it's an extremely simple model.  
[01:26:48] The global tradeable good is a numerair. The price is normalized to one. I'm going to quote old prices in terms of this numerair.  
[01:26:57] The right side is the household. Where can the endowments go?  
[01:27:04] And the traded good. Then comes the borrowing.  
[01:27:08] So this household can borrow at the interest rate RT from what I'm going to call a global financial intermediaries.  
[01:27:16] This borrowing is denominated in domestic currency bonds.  
[01:27:20] There's no currency here. By currency, I'm just going to refer to the domestic consumption baskets.  
[01:27:26] The domestic consumption basket is Q.  
[01:27:38] A different elasticity of intertemporal substitution.  
[01:27:42] I'm going to have a different elasticity of intertemporal substitution. That's not going to be one within period is going to be one.  
[01:27:54] So this is where a lot of nice things with a lot of things.  
[01:28:04] With collapse in here.  
[01:28:27] I'm going to be very important here. And what assets we can trade is important. There's another layer to this. There's an asset that this household choose. There's going to be also an asset that they don't choose coming up the next slide.  
[01:28:39] But this one is denominated in domestic currency and we assume they have access to domestic basket paying bonds.  
[01:28:47] The rest of their budget constraint is this fiscal rebate from the central bank coming up and the profit of the global financial intermediaries that is allocated to every country.  
[01:29:38] The non traded endowment is hit by this local idiosyncratic shocks. This is going to be independent across countries.  
[01:29:45] The traded endowment in contrast is hit by one global shock.  
[01:29:53] There's going to be an exogenous shock. Everything in red is going to be country specific.  
[01:29:58] The global shock to tradeable endowments lifts all the boats or sinks all the boats, but to a different extent.  
[01:30:06] Think about a country that is exposed to commodities. This is a country whose income goes up a lot when global income goes up.  
[01:30:15] So, I think that's a good example of a country whose income goes up a lot when global income goes up.  
[01:30:26] Also high, importantly covariance with global output and global consumption.  
[01:30:53] I think that's a good question.  
[01:31:19] I did not restrict the sign of the borrowing or saving in domestic currency bonds. This is separate.  
[01:31:25] So this agent is managing, if you will, the country's dollar reserves, but we would actually like you to think about more broadly agents that have mandates to manage countries accumulated savings and invest them abroad.  
[01:31:37] Think about tension funds and insurance companies that also source dollar investments or safe investments abroad.  
[01:31:45] This agent is very simple.  
[01:31:55] So, looking at one thing, the dollar exchange rate.  
[01:32:00] The households is just the profits from managing those dollar assets. The way they manage these dollar assets is when the dollar is expensive, they realize a bit of capital gains that they got and rebate it to households.  
[01:32:13] When the dollar is cheap, they buy more dollar assets. So they try to counterbalance what the dollar is doing.  
[01:32:20] So this similar extent in terms of elasticity is here the towel in this policy is the elasticity of what they do with respect to the dollar exchange rate movement.  
[01:32:29] The thing that they have different across countries is the target level of reserves.  
[01:32:34] That's green here, I don't know how much you can see green but this is different across countries, am I and this is the second and last source of heterogeneity across countries.  
[01:32:43] So this is the same level of target reserves. So the same appreciation of the dollar makes them realize the same portion of capital gains, but in absolute terms, this is going to be different across countries, depending on my.  
[01:32:55] So countries households will have different rebates in absolute size will have different increases in consumption and this will feed into their exchange rate moving to a different extent.  
[01:33:06] That's how the currency of country I stabilized against the dollar by this agent.  
[01:33:19] The average compounded basket price of the world.  
[01:33:24] And this qt divided by q ut is the average dollar exchange rate, the broad dollar DXY.  
[01:33:38] Sorry, this is the question. Yes, sorry, that was the sorry, sorry.  
[01:33:46] Correct.  
[01:34:22] I think that they could use to do that. Yes, I could specify.  
[01:34:27] No, no, no, no, no, I simplify this a lot. I just give them this policy that is a stand in for another policy, you could imagine.  
[01:34:36] Exactly, yes, and then you could give them a problem that they would solve and then this would be to something very similar with maybe an indulgence now an indulgence. Am I right.  
[01:34:45] I'm just going to simplify it here and give you implications of this specific model.  
[01:34:56] The US central bank.  
[01:35:08] That it's a Ricardian model so the households will just undo whatever the US central bank is doing. In all other respects the US is the same except this large.  
[01:35:17] Now, the last agent in this model is this global financial intermediary. After that we clear the markets and analyze.  
[01:35:28] This facilitates investment across countries.  
[01:35:32] Specifically, they're on the other side of the transactions and local currency bonds with households around the world. They also participate in trading the dollar bonds as well.  
[01:35:43] So what they're going to do is they're going to live for one period. It's a simple static agent. They have no capital of their own. So whatever they do has a zero capital constraint.  
[01:35:53] They participate in markets for all bonds. This constraint here says they have no capital so whatever they save. They're also must borrow at the same time.  
[01:36:02] They have access to all currencies their problem is a simple problem trying to maximize expected discounted returns on their portfolio.  
[01:36:11] This can be written in terms of excess returns versus the USD, just because of the zero capital constraint with their position some to zero.  
[01:36:20] The important thing here is going to be this stochastic discount factor. So how they evaluate the riskiness of their portfolio.  
[01:36:27] This tells them state by state how much they value an additional dollar of payoffs.  
[01:36:32] This part is pretty standard. I'm going to show you the specific form of the SDF. Next slide.  
[01:36:37] The part that is slightly non-standard is that they also face this portfolio management cost.  
[01:36:41] We need this to stabilize the solution specifically for the steady state. It's more of a technical reason here.  
[01:36:48] What this is doing here is it penalizes deviations from a simple uniform portfolio.  
[01:36:55] So think about a problem of this intermediary who finds it hard to compute the optimal portfolios has to invest something to deviate from the simple thing.  
[01:37:04] A simple uniform portfolio is evaluated at some benchmark level that benchmark level is also going to move over time.  
[01:37:11] For simplicity, we're going to set it to the global demand for reserves so that if we're in this benchmark, basically the only capital flows on that that happened is the intermediary satisfying the demand for reserves.  
[01:37:23] And that's it.  
[01:37:24] The optimal allocation of capital coming out of this problem is going to be to country I is going to be a combination of the benchmark, the simple thing, plus the fundamental part, the returns that they care about.  
[01:37:37] I'm calling acts the excess returns.  
[01:37:40] The expected excess returns matter. The covariance of the excess returns with the SDF also matters.  
[01:38:02] We think it just combines an average SDF of all households in this economy, all households in the global economy.  
[01:38:10] Here specifically I'm computing this average of the households as the F's across countries, countries I am the US.  
[01:38:18] And then I'm doing one more thing to it I'm giving them an increased curvature, sometimes.  
[01:38:25] So sometimes you would take this SDF, and then you would just increase the effective risk aversion by putting more weight on the part that has the curvature and less weight on the flat part.  
[01:38:39] This would mean the SDF of the intermediary amplifies the state dependence of the SDF of all households.  
[01:38:46] If gamma goes down the SDF of the intermediary becomes something that is close to a flat function doesn't care about the allocation of payoffs across states.  
[01:38:54] So gamma goes up is effectively risk aversion going up.  
[01:39:17] So then,  
[01:39:25] US basket that's a combination of the tradable good and US non tradable good.  
[01:39:33] Yeah, in the end it's not going to leave the country yes.  
[01:39:46] All goods across countries in the within period. Yeah.  
[01:39:49] Correct.  
[01:40:21] The long.  
[01:40:52] So we're going to be this way of setting up the SDF is going to make the intermediaries risk preferences aligned with global shocks, global consumption.  
[01:41:04] That's, that is what is something that's going to give us the correct profile of riskiness of countries.  
[01:41:12] So that's a problem.  
[01:41:19] I would think that the intermediary issue, equity and try to sell it.  
[01:41:35] And if you think that they sell it that shares the free we traded across the world, then there's going to be a price that maximizes the.  
[01:41:48] So the set of strategies that maximizes the revenue from selling those shares that is going to try to satisfy something like the average consumer.  
[01:42:13] I'm sorry, it is.  
[01:42:16] Marks.  
[01:42:20] I'm going to hold this has to be everywhere in the large she has to be everywhere. It's the exact same thing.  
[01:42:41] It's how you evaluate the tradeable consumption across periods. So this this large T is also here and here and here and here.  
[01:42:49] Sorry about that.  
[01:42:52] I didn't tell you here on the slide. You also redistribute profits and full disclosure. It's not exactly equal to the equal shares.  
[01:43:01] So control rights are not exactly equal to the profit shares. They're different in the second order.  
[01:43:07] The way the reason I make them slightly different from averages is that I need to economize on state variables later when I solve the model.  
[01:43:15] I need to find a profit rebate rule that is not too different from the average, but also kills the wealth distribution as a state variable.  
[01:43:53] They're on consumption they consume some goods.  
[01:43:57] Then, would there as the FB as tightly related to global shocks as it is here. I need to think about it more but it's certainly some something else that it's on the table. Yes, and I know you've done this before.  
[01:44:09] I'm going to go. Okay.  
[01:44:11] So I have very good questions. I need to start solving the model now. So the markets that are clearing in this in this model are markets for the non tradable goods country by country they never leave the country.  
[01:44:21] The tradeable goods market is global. Everything that's produced in the economy is consumed in the economy.  
[01:44:27] The market for currency I bonds also clear bond by bond, and then the US bond market is the same except it has a slightly more participants it has also all all the central banks that demand those reserves.  
[01:44:40] What are you doing here. Yes, you cannot see the C I bonds across the entire world, a segmented markets. I'm going to have this shocks just to reiterate the two.  
[01:44:56] So this global shock Z is something that lifts the global output here and gives us the commodity differential commodity exposure.  
[01:45:03] There's a US non tradable shock. It's one of those ideas in credit shocks but the US is big so it has global importance here.  
[01:45:10] So the gamma shock is the shock to the risk aversion of the intermediary repricing the risk, and then set of ideas in chronic shocks country by country. I'm going to do them here.  
[01:45:21] Well, I saw the model was this solution concept that I'm doing I'm going to do a first order solution around something that is a version of the risky steady state.  
[01:45:30] The risky steady state is going to be a set of allocations and levels that is determined by the riskiness properties of the first order solution around the steady state.  
[01:45:41] So the literature that takes the size of the shocks to zero but also inflates the risk aversion of the intermediary at the same time to keep risk premium in the first order.  
[01:45:54] This is the important part. How does the intermediary assess the riskiness of different currencies. This depends on what shocks enter the SDF of the intermediary and what shocks will therefore command the price of risk.  
[01:46:08] So in the intermediary, it has two shocks inside of it. One of them is this global shock. This is very much expected. Whenever global consumption is high the marginal utility of global consumption is low in the intermediary's problem.  
[01:46:30] So the US level of stock.  
[01:46:35] The city of inter temporal substitution. When the US gets a lot of consumption in non tradable sectors.  
[01:46:41] The marginal utility of tradable consumption also changes, whether it goes up or down depends on how the LST city of inter temporal substitution compares to the LST city of substitution within period.  
[01:46:52] So this is missing if we have co-op said unit LST series everywhere. This part, if we have the standard calibration high.  
[01:47:08] This means for the intermediary.  
[01:47:12] The notion of good times is when the dollar is cheap, because there's a lot of US consumption around the notion of bad times is when the dollar is expensive because there's very little US consumption.  
[01:47:23] So these are going to be a good hatch from the intermediary's perspective. If they don't depreciate too much when the dollar goes up.  
[01:47:44] I'll enter here gamma shock is a distributive shock changes risk aversion but does not change anyone's sorry aggregate consumption. So it will cause a reshuffling of capital flows.  
[01:47:54] It will not cause a change in global consumption therefore is not going to command the price of risk associated to it.  
[01:48:04] So here, let's look at the price of the local consumption basket. It has two parts to it the tradable consumption and the non tradable consumption. This has a very easy functional form because of the coped Douglas assumption.  
[01:48:16] The tradable consumption can go up whenever you get this windfall from commodities for example, global boom translates into a windfall income when you have a high EI, maybe a slightly lower windfall if you have a low EI if you're Japan.  
[01:48:30] If you're global booms you're still consuming more than an average. That's a positive EI, but maybe because it's lower than the average you're lower than average.  
[01:48:39] You get inflows. So this is what I denote the financial inflows from the intermediary by sometimes the intermediary is going to decide that they're going to invest more in you or less of you.  
[01:48:49] And then this reserve management channel also affects tradable consumption. This is the channel through which the reserve or foreign asset policy stabilizes your exchange rate by moving your consumption from those realized capital gains.  
[01:49:03] The non tradable consumption part is easy, more non tradable goods, they're more abundant, cheap exchange rate falls.  
[01:49:09] Now why would capital inflows happen? Three reasons.  
[01:49:13] Before we go to the exogenous objects, let's look at the intelligence objects.  
[01:49:17] Reason number one, the expected excess returns on your country go up. This is the UIP deviation that your country offers global markets. This can sometimes be positive.  
[01:49:27] This is going to cause intermediaries to get more interested in your assets and invest in you.  
[01:49:32] The second thing that's going to happen is the shock to the risk aversion that might affect you in different ways depending on whether you are a risky currency or a safe currency and it's captured here by this covariance term C I.  
[01:49:45] And finally, there's an overall expansion of the intermediaries balance sheet that happens sometimes when people demand more dollar reserves, for example, the dollar is cheap.  
[01:49:54] This has to be intermediated through the intermediaries balance sheet and everybody gets inflows because now the balance sheet is larger.  
[01:50:01] And the dollar means a lot of investing around the world as well. And what is this covariance here? This is exactly what I showed you on the last slide, the SDF, how the SDF covariance conditional on T with the properties of the country's currency.  
[01:50:20] Remember that it has this DZ shock and the US shock.  
[01:50:24] This appreciation of the exchange rate of country I against the dollar also has the DZ shock and the US shock. Let's look at this.  
[01:50:32] This depreciation of the country I currency.  
[01:50:38] I'm going to just walk you through the four exogenous shocks. But first of all, this is a bilateral depreciation now against the US. So the first thing that matters is when there's a global boom, how much more are you exposed to the global windfall of income compared to the United States.  
[01:50:53] That's the easy part.  
[01:50:57] The US shock. How does it affect the bilateral exchange rate? There's a direct channel here. The one comes from the fact that more US consumption means it's abundant. The US dollar is cheap. So the US dollar falls.  
[01:51:09] There's the second part here. This is coming from the reserve management or the dollar asset management because that policy responds to the exchange rate of the US changing.  
[01:51:20] If you have a lot of dollar assets here, you're going to respond a lot. So when the dollar is expensive, for example, your country is going to realize a lot of capital gains in levels. If your MI is high and this is going to lift your currency and bring it closer to the dollar.  
[01:51:37] It's not about because it also affects the intermediary's balance sheet size. Remember, and all the capital flows from this management are going to be recycled back.  
[01:51:49] So what matters here is your level of dollar reserves relative to the average.  
[01:51:57] Riskiness. It's going to be the covariance of the intermediary's SDF with these two things with the DC shock that appears in the bilateral exchange rate and the US shock.  
[01:52:07] And the final part is easy. This is the local non-tradable shock. Now we have this form for the covariance term just because we know everything now. So we know how the Z and US shock entered the SDF.  
[01:52:19] We now know also how they enter the depreciation of currency I and it's easy to compute this covariance.  
[01:52:27] The nice part is that these things that differ country by country, the exposures to DC and the US shock are inherited here by the covariance.  
[01:52:37] So the covariance term is going to inherit these exposures.  
[01:52:41] If you're a country that is more exposed to global output, you're also going to be more risky because of that from the intermediary's perspective.  
[01:52:48] And when the intermediary's risk aversion changes, it's going to be reflected in how much your currency is affected in your capital outflows.  
[01:53:01] Now, so this is bilateral depreciation of country I against the dollar in terms of all the shocks. I group the terms by what here depends on the exposure of country I to the commodity cycle and what depends on the exposure of country I to the dollar assets.  
[01:53:19] So I'm not seeing this cross section. This is a whole cross section. Sorry, it's a whole panel of bilateral exchange rate depreciations, but it has two components and I can slice it along two dimensions here.  
[01:53:31] One dimension is, I'm going to try to integrate out the dollar assets.  
[01:53:39] I compute the average depreciation of the currency I against the dollar. That's going to close the dollar asset part.  
[01:53:47] I'm going to integrate out the commodity exposure. So I'm going to take high commodity countries versus low commodity countries. That's going to isolate this commodity part here.  
[01:53:57] Once I integrate all these depreciations. I have a very nice and simple formula for the exchange rate in terms of two global factors. Thanks.  
[01:54:07] Here in brackets, and I have exposures to them.  
[01:54:11] The commodity factor is this appreciation of high commodity countries versus low commodity countries.  
[01:54:17] How you're exposed to that depends on your exposure to commodities. The dollar factor is the appreciation of the dollar against everyone else.  
[01:54:23] Your exposure to that depends on your dollar assets, but notice that both of these factors have both the real part that is doing the actual work in this place of goods.  
[01:54:35] And then the financial part that is the repricing of the risk that is coming out of the exposure to the real part.  
[01:54:43] Both parts are conceptually equally important. And then our job is going to be to tell you how important they are relative to each other within those factors and then eventually for the currency itself.  
[01:54:57] That's pretty good. I'm going to go to the data now.  
[01:55:02] I'm going to run this regression in the data and indeed many people before us have run this in the data or a version of this here. I'm going to show you the fit country by country just just to have an idea of how much variation in currencies.  
[01:55:16] The simple two factor regression can explain.  
[01:55:19] In the data I'm computing the exact same objects as in the model for the factors, and I'm plotting here on the right for each country in my panel, the R square of this regression.  
[01:55:33] And two global components explain currency by currency on the Y axis here I have a theoretical upper bound. What you could even expect out of regression with two global components, just do PCA, right, and plot the R square from the PCA.  
[01:55:49] And it turns out that this regression with two factors comes very close to the PCA, and we even flatter the PCA a bit because when we construct these factors we leave out the currency itself, just to avoid the mechanical high R square.  
[01:56:04] On average, this R square is pretty high.  
[01:56:07] The second thing that's going to be important is, there's a lot of heterogeneity in this regression.  
[01:56:13] In some countries, global components explain a lot, almost everything. For some countries they explain very little, and our model can tell you exactly why that is, depending on the country's fundamentals.  
[01:56:27] And those loadings on the factors they depend on high, and we found correlates of those loadings. The thing on the left has been known in the literature.  
[01:56:36] The commodity factor loading is correlated with the commodity exposure of the country here computed as the share of commodities in your exports.  
[01:56:44] The loading on the dollar factor, and this is a thing that's relatively new to us, to our paper, the dollar factor loading is correlated with the USD asset holding so they can share here.  
[01:56:58] This is the model. This is also why we started to model those things in this way. Yeah.  
[01:57:08] You know the other, the other assets.  
[01:57:12] Without the FDA.  
[01:57:19] Works.  
[01:57:21] Ategories separately, but slightly less tight.  
[01:57:29] The compositions in the quantity data analysis.  
[01:57:32] We're going to use the cross sectional moments to discipline the sizes of the shocks and the parameters.  
[01:57:38] The new moments that we bring on the table are these cross sectional moments, the moments of the global components.  
[01:57:43] These global components combine cross section the cross section in some ways, they slice it in two ways.  
[01:57:49] So we take our model to match the standard deviation of the two factors and their correlation as well.  
[01:57:55] And this mapping we have it in analytical form is just slightly large so it's not on the slide here, but identification there is exact.  
[01:58:03] We also match the standard deviation of the residual something that's not explained by the global, global components and in the exchange rates, and we match the standard deviation of currencies on average right so this is not currency by currency we just match the averages.  
[01:58:17] But then in the cross section turns out we have pretty good fit from our model. We also match the average explanatory power of those factors.  
[01:58:28] At the moment here I'm going to show you how the R squares country by country line up from the model and the data on this picture we match the average exactly but then in the data countries that have high R squares of global components also have that in the model.  
[01:58:44] So our mechanism is partly validated by this and I'm going to tell you why exactly on the last slide I have stories about countries here country here and the country here.  
[01:59:05] So that is going to be high, but then also the overall volatility of the exchange rate could be higher low, you cannot tell from just from this graph.  
[01:59:25] Yeah, yes.  
[01:59:29] Ascribe all of that to the non tradable local shock or the local business cycle if you will.  
[01:59:35] Okay, so my there is the composition for important global series. Sorry.  
[01:59:51] Right here just assumes the equal share of the US and the rest of the world.  
[01:59:58] You can actually do different shares of the US and the rest of the world and this is going to rescale some of the parameters but keep all the results exactly the same.  
[02:00:06] So we have this equivalence, but I haven't computed exactly every decomposition in that world where we match the US size by GDP.  
[02:00:18] I can tell you that the US is not a half even in our data is something more of like a third budget EP in our data. So this is potentially important. Yeah.  
[02:00:27] Okay, so decomposing the important time series here.  
[02:00:32] What is the change in the US local price basket.  
[02:00:35] What is it made of mostly we find that it's made of the US local shocks the US business cycle.  
[02:00:41] But then when we take the relative price of the US and the rest of the world which is the dollar factor in our model turns out that the financial shock gamma contributes much more to it than to the price alone.  
[02:00:52] Why is this well because gamma is moving capital flows.  
[02:00:55] When gamma is high risk a version of the intermediary is high is moving capital flows into the US from all other countries the US is ultimately the country that is safe from dollar risk.  
[02:01:05] And every other country has dollar risk. So gamma is moving exchange rates in the US and the rest of the world in different directions always. That's why it has a high importance here for the relative price.  
[02:01:22] And the 40% is the reshuffling of capital flows from low commodity countries to high commodity countries again another version of flight to safety.  
[02:01:31] So the risk of the exchange rate goes up and the intermediary freaks out about the commodity risk now. So whenever you see the Australian dollar moving against the Japanese yen.  
[02:01:40] Part of it is because of the metals demand for commodities coming out from Australia, but part of it is because the intermediaries know about this risk, and they now are afraid of it more.  
[02:01:57] So I have two fundamentals here. So on the left I'm keeping the EI the commodity exposure at the average and varying the dollar assets on the right I'm going to do the opposite here.  
[02:02:08] When you keep the commodity exposure at the average, the Z shock the global output does not show up because you're exactly at the average.  
[02:02:16] So when you increase the dollar assets, you isolate your country more and more from dollar shocks. This means the contribution of the US real shot the US business cycle goes down over time, but also this makes you less risky.  
[02:02:30] So the intermediary doesn't care about you too much when the risk aversion goes up, and the contribution of the financial shop goes down.  
[02:02:36] So when you get to the market, you get a more risky outcome. You become driven by your local shops you're I insulated from the rest of the world. So most of it is just your local non tradable.  
[02:02:49] Since at the average and you increase your commodity exposure, the global output shocks he becomes more important. This makes you more risky so the financial shop also becomes more important.  
[02:02:58] This is more obvious.  
[02:03:05] So these are three different stories just to navigate this two dimensional cross section of commodity exposure and dollar assets.  
[02:03:14] We're going to start with Thailand. So Thailand here is a country that has very low commodity exposure. It doesn't export a lot of primary inputs. It also has pretty high dollar assets in our sample will compare to its size.  
[02:03:26] In the model and in the data.  
[02:03:29] Thailand has low volatility currency. It doesn't have a lot of variants.  
[02:03:35] And also the factors the global factors are not very important to it. So this is an example of a country that is insulated from us shocks because it has relatively large dollar assets.  
[02:03:44] So most of what happens to tie that is driven by Thailand, local shocks here, according to our model.  
[02:03:51] So the example here is Hungary. So Hungary also does not export a lot of primary inputs very long commodities. The big difference is that Hungary doesn't have a lot of dollar assets.  
[02:04:01] Hungary is in Europe. It has its own currency but it mostly trades with countries that use the euro. We think that's why maybe they're foreign assets and mostly denominated in euro.  
[02:04:12] So this is a little to them perhaps but this renders them very exposed to us local shocks. So whatever happens to the US business cycle moves the dollar against the Hungarian foreign a lot.  
[02:04:24] This also makes them risky from the intermediaries point of view. So financial shocks are more important for them. Global factors are more important for them.  
[02:04:31] Global factors explain 70% of the variation. Yeah, feasible in principle and I would love to do that.  
[02:05:05] The moment I start estimating Tau's country by country I'm left with very, very few data points. So the bands, the confidence bands are going to be all over the place.  
[02:05:14] Unfortunately, that's why I'm kind of pulling the data from multiple countries to do the towel. But MIs the reserves or sorry the foreign assets I just take them from the data country by country.  
[02:05:48] So that is exposed to commodities relatively low dollar assets very volatile exchange rate.  
[02:05:53] Global components are very important but because it's also doubly exposed to risk to its exposed to two kinds of risk.  
[02:05:59] This is the country where financial shocks are a very important driver of their currency, almost half of what is explained by global components.  
[02:06:06] Okay, so this is a global economy model with two types of heterogeneity between countries, where we map macro shocks into global factors that explain depending on countries heterogeneity.  
[02:06:17] And a lot of variation in the exchange rates.  
[02:06:21] We exactly replicate those factors in the model and the loadings on those factors replicate pretty well. And then we decompose the variation in exchange rate country by country.  
[02:06:31] The lesson is there is no single answer for that. It's very hard to account for the entire cross section but if you know these two things about the country, it's commodity exposure and dollar assets.  
[02:06:42] And it's a long way to explaining it's the variation in its currency and how much is of it is global and how much if it is local. Thank you.  
[02:08:18] Hey, there are several things I like about it I mean one is that the, you know, that the modeling is very clever I mean it's difficult to build a general equilibrium model with a continuum of countries and have this financial intermediaries,  
[02:08:32] and they're all hanging together so that's a that's a really neat thing and then that the empirical work is very thorough and and to me in a way very convincing although, you know, a lot of my comments are going to be, I'm not sure really.  
[02:08:44] I believe that the model is what is the right explanation of it but but but but then you know it's important the paper is important for, you know, kind of building a macro model that is consistent with this.  
[02:08:59] These studies of the factor structure of exchange rate movement so I think that's a big step.  

## Discussion (02:09:04 – 02:23:40)

[02:09:06] I'm going to start by just talking about the empirical evidence and because that's a lot about what my discussion is about so. So the idea is that first of all you look at the movement of the exchange rates relative to the dollar for 32 countries and, you know, at  
[02:09:22] these 32 countries maybe a close to half of them are kind of high income, low inflation countries.  
[02:09:28] The other half are kind of emerging markets some of which are big commodity exporters. And so, so for one thing there's a dollar factor which means that basically, what's the dollar factors the change in the dollar exchange rate against the average and so, to that extent,  
[02:09:44] the other, you know, the 32 countries are exchange rates move up and down with the average dollar exchange rate. And then the other factor is well okay well what else is there so we'll look at the difference in the movements of  
[02:09:59] the two factors that are big commodity exporters versus those that are not. And that explains some of the variation some of the, you know, some of the stuff that's not explained by just the correlation of the dollar exchange rate with the average so.  
[02:10:16] So the first thing you do is say well, our model says that the commodity factor should be correlated with how how much you exported the commodities and also with the risk factor and I'm not going to talk about the risk factor because I'm going to get confused if I tried to talk about that.  
[02:10:37] And then the the dollar factor is correlated with with the service employment in employment and services in the US but the idea here is that if the non traded sector in the US contracts that's going to cause a dollar appreciation and that that would be against all countries so  
[02:11:04] Okay, so then you do this so then you say okay well these two factors that we got they look pretty consistent with our model, but let's also see what the loadings are on these two factors so the cross sectional loadings and that works out pretty well too because the countries that low most on the commodity factors are the ones that are commodity  
[02:11:31] are the ones that are commodity exporters and and then this is a tricky one. The ones that load on the dollar are the ones. The main thing is that they're the ones that have a lot of dollar assets, so dollar assets dollar portfolio assets relative to their total portfolio assets.  
[02:11:53] And then they do a whole bunch of you know robustness stuff controlling for a bunch of stuff.  
[02:12:06] So, I've already kind of indicated a little bit of the model mechanism but you know the way these these exchange rate things all work through kind of demand and how they affect the relative price of non traded goods so if you're big commodity exporter and there's a commodity boom, then your income goes up and that  
[02:12:29] you know relative to other countries and that raises the price of your non traded goods so your country your currency appreciates more than the average currency.  
[02:12:37] And then, you know the dollar asset channel the way this works is, you know again, you know there's this contraction in the US economy that's  
[02:12:46] the non traded sector and that's driving up the, or is appreciating the dollar. Why do some countries appreciate or depreciate less than others well, it's because they hold dollar assets and so an appreciation of the dollar.  
[02:13:03] Well, okay, that that's not so bad for us because we're holding some dollar assets that raises our wealth and raises our consumption and so, you know it mitigates the effect on the demand for non traded goods.  
[02:13:19] Okay, so that's my understanding that maybe maybe is wrong. But anyway, and then I'm going to skip this stuff about the financial intermediary.  
[02:13:32] So, so one of my topics of discussion well, as I say, as already indicated I'm not really convinced about the mechanism the model.  
[02:13:41] And I there are other things I would want to bring in, but I don't want to be one of these old guys who say, I am. I don't want to be one of the ones that say is wrong.  
[02:13:51] You know, because I do think that there's real insights in this paper and that's really how I judge the paper but there are.  
[02:13:59] First of all, I'm kind of other things that might be driving exchange rates and I don't really have I didn't like to try to write my own model down and solve it but the things I would look into so there's suggestions for future research.  
[02:14:18] And the things I'm going to talk about is first of all, you know, our real exchange rates really determined by the relative price and non traded goods.  
[02:14:27] Second thing is I think the stance of monetary policy, you know how, how much you're targeting inflation how strongly you're targeting inflation I think that matters for exchange rates.  
[02:14:39] And the short runner driving are at least correlated with the real exchange rate. And not only that but a lot of these emerging markets, you know, they've taken up inflation targeting in the last  
[02:15:25] 15 years and they're looking more and more like low and low inflation.  
[02:15:42] I think it's worth looking into and measuring the relative price and non traded trade because remember everything in here is coming through the consumption.  
[02:15:50] CPI is not the terms of trade per se except how the terms of trade effect income which affects the demand for consumption goes.  
[02:16:01] Now, measuring the consumption price and non traded traded goods is controversial but one thing is that if you write down a really simple loss of loss of Samuelson model and do a kind of a reasonable calibration.  
[02:16:17] Well, like there is two thirds non traded one third traded and then I, I made the additional assumption that they're non traded and trade or productivity shocks that are perfectly correlated but the non traded moves one third as much as it created then actually the consumption real  
[02:16:34] wage should equal exactly the real exchange rate but my point here is that, you know, in the data.  
[02:16:47] Little as a real exchange rate so that so if you look at this, I mean I just did it for their three favorite countries I'm going to run out of time.  
[02:16:55] So it doesn't look too bad for Thailand this is detrended. So the blue is the real exchange rate the red is the real wages they're kind of correlated.  
[02:17:04] The real wages. This is all relative to the US is less volatile but it's not too bad and then you look at Hungary and done look that good at all I mean the relative real wages a lot less volatile than the real exchange rate.  
[02:17:20] And then you look at Australia. And I mean this is kind of typical of the high income low inflation countries that the relative real wage is like way less volatile than the real exchange rates.  
[02:17:32] So, you know, that's the first point. Second point is I think that monetary policy stance matters and here I just want to, I guess I'm really maybe plugging my own paper although I do think it's an important point that, you know, we have this view that there's  
[02:17:48] disconnect I don't know these guys it's Cokey Mook and make you think there was and that used to be true but I don't think it's true anymore.  
[02:17:55] And so I have this paper with Steve Wu, and you know we look at kind of the Macy wrote off regression but add in the financial variables.  
[02:18:04] And it actually fits pretty well these days and I might my hunch is that it fits pretty well for emerging markets as well as fitting better and better and actually john was shun and student is working on this.  
[02:18:17] But here, you know, here is the, you know, the R squared of the of the monetary model with financial variables thrown in with rolling samples that start in in this year.  
[02:18:35] You see that these days the R squared of the model is really high. But that's being explained both by kind of the financial variables that the spreads and that the traditional variables and Macy wrote off.  
[02:18:48] And you look at this and say oh no no but everything was 2000. So this is the rolling sample 20 years sample starting in 1988 so everything was so this 2008.  
[02:18:59] Everything was explained by financial variables, but the red and the green are the shaftly value.  
[02:19:06] How much each of the two sets of variables are contributing to the exchange rate and even in 2008.  
[02:19:12] The traditional variables explain more than the financial variables. And so my point is that, you know, contrary to what we learned when we were young, actually there's not disconnect.  
[02:19:25] And so I do think that monetary policy matters and in fact if you look at the whole sample since 2000, or if you just look at 2008 2009, or you look at the whole sample not including 2008 2009, the traditional variables account for around half of the movement of the exchange rate.  
[02:19:44] So the, you know, the star square doesn't look that great point three but actually if you look at quarterly data. It's around point five so there's a lot less disconnect and I think that monetary, the stance of monetary policy matters.  
[02:19:58] And as I said, a lot of these countries now are inflation target or so all the yeses are all the countries that have adopted inflation targeting. And I think they're starting to look a lot more like the advanced countries.  
[02:20:12] Let me just make two more points.  
[02:20:15] You know, in the model, it's gross dollar asset holdings that matter, but it's sort of because of this unusual structure of the model that households can only trade assets in their own currency with the intermediary.  
[02:20:30] But the central bank can acquire dollar assets. But it seems like it ought to be net dollar asset holdings that matter so I did this. I mean this lazy man's calculation, because I'm a lazy man, and I said, Well, what's the, what is net dollar holdings look like and by the way this would drive like  
[02:20:48] shut them crazy because I took net dollar holdings from one data source, which was CPIS and then I divided that's the numerator and I took the same denominator. They have here, which is total asset holdings from the tick data.  
[02:21:05] I did it anyway because you know how much are they paying me to do this.  
[02:21:11] Anyway, but no, but the point is that somebody should do this, not me. And the point was that when I did this and I cut out the outliers that were, you know, bigger or less than 100%.  
[02:21:24] The correlation was like zero I was point zero three six or something so somehow it seems like it's net dollar assets that ought to matter. And then the last point I want to make is,  
[02:21:35] so actually the paper starts by talking about how much each of these currencies appreciated are depreciated against the dollar during the crisis, but a lot that that happened during that period, dependent on how much intervention the central bank did.  
[02:21:51] And I think that that mattered in the case of Thailand they they came in and supported the bot strongly hungry didn't do anything. It's kind of a funny case because they, they were swearing up and down we'll never intervene in the exchange rate market and then  
[02:22:04] their currency depreciated massively in August 2008 and then they said, well, okay, we're going to do some intervention. So I think that intervention does matter and and that should be something that we look at more carefully.  
[02:22:17] I mean in the end.  
[02:22:19] Oh, well let me just skip that.  
[02:22:23] In the end, you know, I don't have an alternative model I mean it is certainly the case that they find that the dollar factor is correlated with this dollar asset holdings and you know I don't know.  
[02:22:36] It's something that's.  
[02:22:40] That's my.  
[02:22:43] I think.  
[02:22:45] So.  
[02:22:52] You know, I could hear.  
[02:22:55] But on the other hand, I don't kind of believe it because the model, because it's all coming through this wealth effect. You know, like the dollar appreciates we're holding dollar assets which again I feel like should be net dollar assets.  
[02:23:06] But but even if it was dollar assets I mean the consumption literature doesn't find big wealth effects.  
[02:23:12] And it seems like this should should be smaller so.  
[02:23:16] Anyway, but like I said at the outset I'm going to do really like the paper I mean again I judge a paper by, you know, do we learn from it?  
[02:23:24] Is it an advance in the way we should be thinking about things. And I like the I like the structural framework I like the empirical discipline and I like that it's bringing this to this cross section exchange rates.  
[02:23:37] And so I do think it's going to serve as a template for future research in this area.  

## Q&A (02:23:40 – 02:26:00)

[02:24:53] It's been the outside of an area.  
[02:24:56] No, go away.  
[02:24:57] You're first.  
[02:24:58] Yeah, no.  
[02:25:01] Yeah.  
[02:25:02] Yeah.  
[02:25:03] Yeah.  
[02:25:04] Yeah.  
[02:25:05] Well,  
[02:25:06] Well,  
[02:25:07] Well,  
[02:25:08] Well,  
[02:25:09] Well,  
[02:25:10] Well,  
[02:25:11] Well,  
[02:25:12] Well,