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.
=

# The US-China Trade War and Global Reallocations

Authors:   
Discussant: None  
Video: https://www.youtube.com/watch?v=_fKc5wtqjqM&t=262s


## Talk (00:04:22 – 00:55:15)

[00:04:26] allotted time. So, let's not get into discussions; we'll leave those discussions for the end. But in terms of being able to follow what he's presenting,  
[00:04:38] questions are welcome. Pablo, thank you very much. Everything is great. Thank you, Juan Carlos, for the invitation, and thank you to [name omitted] for reading the paper, which is very preliminary. I would have liked to  
[00:04:51] be in your place and have a preliminary [document/paper] like this. Thank you for that.  
[00:04:56] And as Juan Carlos said, ask me all the questions you want as we go along; it will make it more fun.  
[00:05:04] This work is in English. The cell says it's not a problem; it's [name omitted].  
[00:05:09] Patrick [name omitted].  It would give me, and it starts a bit with him, the motivation is surely with you all, you are all aware of  
[00:05:24] the US-China trade war that officially began in February 2018.  
[00:05:30] To March 2018, the US imposed a series of tariffs, first non- discriminatory, on a group of products, mostly  
[00:05:43] metals, and then it became clear that the trade war was going to be a trade war specifically aimed at China. So, in a series of  
[00:05:55] different phases, the US imposed tariffs on China throughout 2018 and 2019, and China responded to these waves of  
[00:06:07] tariffs so that by the end of 2019 these tariffs had accumulated, they kept increasing and never fell except a little. A significant amount of trade between both countries  
[00:06:22] was subject to tariffs. There is a significant number of papers that try to analyze the effects that this trade war had or is having on the US, on China, and on a series  
[00:06:39] of [Music] variables, mainly on trade itself, because well, that's what one can most directly analyze. The data is available and so on, but  Also, regarding other internal variables like production,  
[00:06:52] employment, and so on, and I have to say, like a football match, what a pleasure it is to present in Spain, it truly is a pleasure.  
[00:07:01] So what we're going to do in this paper is shift the focus of this literature, which is centered on the US and China, and ask ourselves  
[00:07:15] what the effects of this trade war are on most countries in the world, focusing particularly on US exports, how the exports of other countries, not just China, to the US  
[00:07:26] responded to the trade war, and the motivation is not only to learn exactly what happened, but to use this episode not only to understand what happened here, but also to try to learn a little about some  
[00:07:40] mechanisms that are generally important when countries experience trade shocks, and, let's say, the mechanisms that can make them adjust, and how these mechanisms can be different between different countries. I know that this seminar is  
[00:07:54] obviously being held in Argentina, and it's not early, it's Argentina, but Argentina is in the sample, so when we're going to show you the results, we can do your mixed [unclear - possibly "on" or "on"]... These are just for motivational purposes; this is  
[00:08:08] a graph with the old raw data of how much they changed  The exports of different countries—these are the countries we have in the top 150 global exporters— and what this simply shows is how much  
[00:08:22] each country's exports changed to the US and China, plus the rest of the world, that is, to everyone except the US, to China and the world. And obviously, many things happen; countries  
[00:08:35] have very heterogeneous changes in exports. But in principle, one might think that the trade war had something to do with how these countries' exports changed. And when one asks how, certain mechanisms come to  
[00:08:50] mind very quickly. And the first question is, what are these mechanisms? And the ones we'll have in mind when we look at the regressions will be the following: Let's take a country, I'll use Malaysia as an example. Malaysia is here,  
[00:09:04] and in our estimates, it's a country that benefited quite a bit from the trade war. And what might be happening is that perhaps a country like Malaysia specializes in some products that receive higher taxes, and the US  
[00:09:19] imposes taxes on heavy machinery. Malaysia is a producer of this.  The type of goods then benefits because perhaps it can access the American market by replacing Chinese goods. So that's one mechanism, but there are other mechanisms. And we can say that even if  
[00:09:33] we condition on the specialization pattern, the fact that Malaysia produces certain goods, or Israel, for example, which turns out to be a loser and its gross domestic product has fallen relatively, Latin American exports  
[00:09:46] to the rest of the world produce other goods. The type of shift that's happening is that the type of goods still conditions the type of goods. These goods complement or substitute each other differently with China and the US. Perhaps there are countries like Malaysia that produce goods that substitute for Chinese goods, which allows them to  
[00:10:01] enter the US market, but other goods complement Chinese goods. So in this case, when the US imposes a tariff on Chinese goods, what happens is that instead of being a boon for you, you can penetrate the American market. It's negative because your goods no longer complement Chinese goods. So when demand in the  
[00:10:15] US falls due to China, your export capacity also falls, the demand for your goods falls, and then, let's say, we can call it a mechanism on the supply side.  
[00:10:26] We have mechanisms on the supply side, so We can think that there might be something related to supply elasticities. Some countries may have very flexible internal factor markets and allow for fairly rapid factor reallocation. There may also  
[00:10:39] obviously be an input-output history of imputed Wood Link gadgets that may be playing some kind of role. And we'll see that some results suggest there may be goals.  
[00:10:53] If you have any questions, please let me know. I'll make a little model to somewhat discipline the way we think about this as a crime that captures these forces, and basically it's an empirical paper.  
[00:11:07] So the experiment we're doing is this: we're going to estimate the effect that each tariff in the US on China and the tariff from China on the US had on the exports of each  
[00:11:21] country to each destination: to the US, to China, and to the rest of the world.  
[00:11:28] And the variation will come from the variation between products. So, somewhat intuitively, what we're going to be doing is looking at a country and seeing how that country's exports, when comparing different products  
[00:11:43] to the United States, to China, or to the rest of the world, changed based on the tariffs that the US imposes on these products, or China imposes on the US. So, the heart of the matter...  
[00:11:56] But let's create a table that has the entire coefficient matrix, all the combinations of how each tax affects each response.  
[00:12:05] Then, using this model, we'll try to tell a story about what's happening, like the outcome. What we'll  
[00:12:21] see is, first, that the United States and China reduce exports to each other, which is expected and consistent with previous results.  
[00:12:29] But we also have some new results here, specifically that when the US imposes tariffs on Chile, and when China imposes tariffs on China, not only does the US import less from China and export less to the US, but the US also  
[00:12:44] exports less to China. With tariffs on a particular product, the US exports less of that product to China.  
[00:12:52] That's at the bilateral level. But when we look at other countries, what we'll see is that, on average, there's a lot of heterogeneity, which we'll also explore. We'll allow for results to vary by country, but on average, the  
[00:13:05] typical country penetrates the American market. Whether the US imposes tariffs on China or not, the average country enters the US market, but not as much.  In China, when China imposes  
[00:13:18] tariffs on the US, another thing happens—and this seems to me to be a bit of an effect of thinking about the impact of the trade war on the development of different countries—one thing that happens on average, which I did n't think was so obvious,  
[00:13:31] is that the average country either increases its exports when the US imposes Chinese tariffs, or when China imposes tariffs, the US also increases its exports to the  
[00:13:43] rest of the world. I mean, exports increase to countries other than the US or China, but to other third countries. This is like trade increases within the rest of the  
[00:13:55] world as a response to the restructuring of the US and China. And then we do a very simple aggregation exercise— not through the model, we're not going to do a quantitative exercise or anything like that, but we're going to  
[00:14:10] take the results of the regressions, keeping the fixed effects in mind.  
[00:14:14] We're simply going to aggregate using weights that represent the importance of each good in the export basket of each country and each destination, and we're going to make a prediction of how this affected aggregate trade, conditioning on these fixed effects.  
[00:14:31] Surprisingly, when you do this calculation, the increase in exports that the typical country has to the rest of the world is so strong that these results imply a global increase in trade.  
[00:14:45] Then, in light of the model, and as you might imagine when we look at these mechanisms, when we look at what mechanism would explain these responses, well, what we would end up interpreting is that the typical country is a country that substitutes for China  
[00:15:00] and not so much for the US in terms of the type of goods they produce. And typically, we'll say that it's possible that they operate under supply curves that have a negative slope, with some kind of returns to scale. If there is some kind  
[00:15:14] of complementarity here, where if you grow, the US grows, depending on how the US grows, and China grows, the US grows to the rest of the  
[00:15:26] world. We're also going to look at some effects that appear, which are consistent with a significant drop in Chinese demand, which are suggestive of  
[00:15:39] production issues. And then we're going to look at the heterogeneity, the winners and losers—in quotes.  
[00:15:45] When I say winners and losers, I mean in the sense of increasing exports.  Or no, not in the sense of well-being, meaning that increasing exports makes a country a winner. It's a transfer of power.  
[00:15:57] In some audiences, I want to attack you. I think I can say that increasing exports makes a country a winner.  
[00:16:06] So, I came to a question. Okay, I'm not going to spend time talking about the literature. Obviously, if  
[00:16:21] you want, we can discuss this later. If you don't have any questions, if there's a specific interest, we can talk about it. I'm just going to point out that there are a number of papers, as I was saying, that specifically look at the effect of trade wars between the US and China, and how these  
[00:16:34] papers document quite well the bilateral decline where, with the US imposing tariffs on Chinese goods, we import less. There's one thing, like a detail, which isn't so much a detail, which is that the identification in these papers  
[00:16:49] typically compares how US imports change when you compare different origins of the same product. So what you'll see there is that when the US imposes tariffs on China, the US imports less from Chile and more from other countries. This paper  
[00:17:03] is different in a rather fundamental sense, which is that we're comparing exports between products. So, let's say when...  The US is when China imposes tariffs, united as it changes the exports of different products from the  
[00:17:18] US to China and the rest of the world, and the same for other countries.  
[00:17:26] Okay, the terminology of the data, I prefer to use a bit to give you a kind of picture of how the data is organized.  
[00:17:41] So we're going to take global trade data throughout this entire period for these countries.  
[00:17:48] So we're going to think of the data as having all the exporters, and for each exporter we see exports to three destinations: the US, China, and the rest of the world as an aggregate. Well, I'm going to be looking at  
[00:18:01] how much Korea exports to Vietnam or Korea to Brazil. I'm looking at how much Korea exports to the whole world excluding the US and China as a destination. This is when I say " rest of the world," it's the rest of the world  
[00:18:15] as a large country. And then we're going to be looking at these tariffs, which are what the annotation shows. It's when it appears like this, meaning the tariff that the US  
[00:18:28] imposes on China. The supra-index is the destination and the sub-index is the origin.  
[00:18:37] So we're going to be looking at the tariffs that the US imposed on China in each  Mega product, those that the US imposed on other countries because, as I was saying, the first tariffs of the trade war were not discriminatory, they weren't just  
[00:18:51] against China, but they were for all countries.  
[00:18:55] Then we have the tariffs that China imposed on the United States and those that China imposed on the whole world. This is typically tariffs that China did; what it did was not only impose taxes on the US, but it lowered the  
[00:19:07] most- favored-nation rates for all countries at the same time as it raised tariffs on China, making it even more difficult than the US.  
[00:19:20] And then, in terms of the time horizon, we are going to be looking at changes over two years.  
[00:19:27] So we are going to look at how exports change over the cumulative period between 2000 and the two-year period 2016-17 and 2018-19.  
[00:19:41] So these are going to be two-year changes, and we are going to compare this. We are going to control for a hybrid PRP that will be the changes in the previous two years. We are looking at 16, 17, 18, and  
[00:19:55] 19. The 'p' is the period of the trade war.  We're looking at how much tariffs and exports changed when comparing the 1819 average to the 67 average, and that's compared  
[00:20:09] against the change two years prior. Okay, so the time horizon is two years.  
[00:20:19] One question, okay, these are simply some data on world trade in 2017, like,  
[00:20:37] before the trade war started. The composition of the sectors— this is still the fraction of each sector in international trade. Machinery,  
[00:20:51] transport, materials, chemicals, you know, are the most important. The number of products—yes, this is the variation we're going to look at. That is, this is the number of products in each sector and the proportion  
[00:21:05] of products that corresponds to each sector.  
[00:21:09] Some examples, for instance, the level of disaggregation will be: soybeans.  
[00:21:13] So, soybeans will be one product, let's say, in the data.  
[00:21:18] Then aluminum, steel, iron, and so on are other products.  
[00:21:28] And now, these are the tariffs in each of these nine sectors in which the data is organized. What we have is a tariff variation.  
[00:21:36] These box blocks show what the tariffs were that the US set.  
[00:21:42] These are the changes in the Relative tariffs on what the US already had to generate against China are around 2% on average.  
[00:21:52] For example, in the agricultural sector, the US imposed tariffs on China, increasing them by an average of 10 %, plus machinery, where the tariff increase was particularly pronounced,  
[00:22:06] averaging around 12%, with many variations. Some machinery products had tariffs of up to 20%. Then we have the non-discriminatory tariffs, which tend to be a bit lower, focused on  
[00:22:19] machinery and metal. We have something similar for China. In China's case, we see a more pronounced and larger increase in tariffs imposed on the US, particularly  
[00:22:33] in the agricultural sector. Soybeans are obviously an important product that the US exports to China, which received considerable protection.  
[00:22:44] In China, the barriers increased to 30% from a higher base. China already had many products with tariffs of 7-10%, which were higher than those the US had on China. Sometimes that was one of the reasons cited by...  The TRAM administration said something had to be done  
[00:22:59] because China already had higher tariffs than the US, and what we see here are these most favored nation tariffs that China, at the same time, lowered.  And it made it easier for the rest of the world to enter, so it went up for the US and down for other  
[00:23:13] countries. It ends up going back to the motivation of what factors or forces we have in mind can  
[00:23:28] explain the results when we look at changes in exports.  
[00:23:31] Obviously, what countries export is very heterogeneous. So, suddenly, going back to the example I mentioned before, Malaysia has a very concentrated export basket, roughly 50 percent of its exports are in  
[00:23:45] machinery, which, as we saw before, is a sector in which the United States imposed very high tariffs.  
[00:23:51] Meanwhile, Argentina obviously has 60 percent due to new exports in the agricultural sector, and then it affects the other way around.  
[00:24:06] So we move very quickly. I'll tell you about a model from which I'm going to derive an equation to think about the results and the model. How are we going to  
[00:24:19] think about it in a very simple way? Let's imagine that each country has a supply curve for a product, omega. This is the inverted supply curve.  
[00:24:28] So this is the price p as a function of sales x. This supply curve has a slope, and we think of it as being specific to the country. It could be specific to the product and  
[00:24:42] the country, but let's think of it as a country component and if it is, let's say, and if this supply curve, if this lower number means that the supply curve is more elastic, so it's like the country responds more easily.  
[00:24:54] Potentially, if there were economies of scale in production, this number would be negative.  
[00:25:00] And then there's a component here that would have to do with the supply chains, the input-output matrices, the production linkages.  
[00:25:09] The supply curve has a trick that comes from the cost of using inputs in that product.  
[00:25:20] So, today we are an aggregate of different varieties or different products, and given how we run the regressions, in fact, the data will suggest that it seems important that the products  
[00:25:33] use inputs that the same product uses. That is, we're going to look at them. For me, the data are somewhat suggestive that a product, or a product within machinery, will use other products as inputs. They specifically have  
[00:25:48] the product itself. So, this 6 omega will be the cost of producing, or 'friend, the cost of getting a unit of omega in the country. And to produce omega, close  
[00:26:02] to a supply to demand, we have a trans demand, which this...  This means the following: this is the fraction of spending that each  
[00:26:15] country dedicates to each small product originating from a certain country.  
[00:26:24] So, what we're going to have is the reverse: this is what the country with destination n, destined for, let's say, the US, dedicates to importing a certain product omega from  
[00:26:38] a certain origin, that is, the exporter. And what matters is that how much the US spends on imports of a certain good will depend on the price of all goods with elasticities of substitution,  
[00:26:51] which are a function of the other goods of the particular good I 'm looking at. So what I'm saying here is that the elasticity of substitution between Chinese and  
[00:27:05] Malaysian goods can be different from the elasticity of substitution between Chinese and Argentinian goods, and there's a matrix that captures that substitution. These sigmas capture that substitution, which is bilateral. So these are the forces we have, that is, basically, the  
[00:27:19] deltas, the sigmas, to guide the analysis.  
[00:27:25] Now, raise your hand, I hadn't seen that.  
[00:27:46] There, thanks. You have that the sigmas, which are what ultimately determine the substitution patterns, don't vary at the product level, they vary at the  
[00:27:59] origin-destination level, and at the level of... No, product, and that's kind of the basis of the whole empirical strategy, right? And what I don't understand then are the graphs you just showed, as if to motivate, and where you say, well, the  
[00:28:12] composition of exports is different, so there isn't that.  
[00:28:17] Argentina and Malaysia export different things, so they're going to be affected differently, but it seems to me that your empirical strategy is n't based on that, actually, because it doesn't affect the composition of  
[00:28:32] exports. I agree, this comment is to motivate rather the aggregation, because when we aggregate, totally agree, when we aggregate, what we do is aggregate the responses that are at the product level, at the  
[00:28:46] country level, you see? So they were saying there, when we aggregate, there are two forces: the signs can be specific to the country, yes, but also, countries have  
[00:29:00] different weights between sectors or between products, rather. So it's simply to motivate the aggregation, not the estimation, right, right, it 's the aggregation, so, that is, the fact that they are either more substitutes or more complementary at the  
[00:29:14] product level, or it's due to some other characteristic of the country that is n't explained here, corrective later when aggregating, in whatever you have, right, right, ready then.  We take this model—I don't want to go into too much detail because I don't  
[00:29:28] have much time— but from this model, if we do n't take a first- order approximation, when tariffs change, we get an equation that guides the estimation where each country's exports to each destination (going to Canada) for  
[00:29:43] each product will depend on the tariffs with a coefficient that will be specific to the country of origin and destination, controlling for fixed effects at the sector level of origin and destination. So the  
[00:29:56] variation will be across products within the sector of origin and destination, and this coefficient has a very simple interpretation, which seems relatively simple to me, based on the coefficients I  
[00:30:10] showed you before: the elasticities of substitution and the elasticity of supply and demand of the good itself. I'm not going to dwell on explaining this  
[00:30:23] coefficient because I don't have time, but let me show you intuitively what it implies. So what we have to do is a thought experiment: imagine I tell you what the sign of this coefficient is.  
[00:30:34] Run a regression of how much each country's exports to the US and to the rest of the world increase based on the tariff that the US imposes on China. As an example, let's...  Doing it  
[00:30:48] with the others, and then, depending on the sign of the stop of these veins, we are going to identify the sign of the coefficients.  
[00:30:57] So, let's imagine that when this regression runs, what it finds is that a country, or the typical country, increases its exports to the US as a function of the tariff that the US imposes on China. That will suggest to me that this country is a  
[00:31:11] substitute for a Chinese good, which it puts in this column.  
[00:31:16] Now, let's imagine that, in another regression, it also identifies that this country increases its exports to the rest of the world.  
[00:31:24] Conditional on having identified that this country as a substitute, I am going to identify that because it also increases its exports to the rest of the world, it has to be operating with a negative supply curve.  
[00:31:38] And conditional on this response, if, on the contrary, I had identified that this country decreases its exports to the rest of the world, that would have told me that this country has a supply curve that is growing, while in this other  
[00:31:50] column, what changes is the pattern of substitute, ability, complementarity. If I had identified that this country decreases its exports to the US as a response to the analysis of the US imposing tariffs on China, that would mean that this country  
[00:32:04] complements China, and then, depending on what happens with exports to the rest...  The world can have a positive or negative supply curve. In conclusion, in this column, I'm supplementing with China.  
[00:32:15] You see the little hand, the little hand of the...  
[00:32:22] now in the right-hand column, replaced with China in this diagonal, in the diagonal below left and above right,  
[00:32:35] economies of scale, and in this other diagonal of economies of scale. So, the exercise of the paper is going to be to put the countries in these buckets, let's say, and try to learn what happened. And obviously, this is for a tariff. The  
[00:32:48] same applies on the Chinese side. If we bring in the production linkages, more complicated things start to happen, and in particular, what will happen is that if China turns out to be a country that uses inputs very heavily  
[00:33:03] in the production of goods, if it turns out that to produce machinery, China uses machinery within the same product, what will happen is that, since the US puts a tariff on China, China will decrease its exports to the US, but also decrease its demand for that same  
[00:33:18] product from the rest of the world from the US, and that will cause the rally that it may have to fall, as Chinese exports to Asia from...  
[00:33:29] The rest of the world from the US, additionally these forces, the people, I have a question: is that equation that you showed me before the correct one, that the level of aggregation made ticks to  
[00:33:42] six digits, that is, for each product you have a beta coefficient, not equipped, it's not for each origin and destination, I'm going to run this review between  
[00:33:57] products, it doesn't vary by product, and because the identification is between products, now of that regression, what we are  
[00:34:13] going to do now that shows it more clearly, before going to the regression, I tell you what I showed you in the cinema was with a single coefficient, but obviously from the model we have all the coefficients, so all the tariffs matter to  
[00:34:26] explain the exports of each country to the rest of the world, to China or the US, all the tariffs matter, and in particular those that the US is putting on the rest of the world, those that China puts on the US, and for each of these betas we have  
[00:34:41] an interpretation, we are going to run that, okay, and now we are going to the regression and be clearer about what Juan Carlos was asking before, before returning, I show you these, these Vince Carter,  
[00:34:54] okay, Skater, they are like the, like they are the  The data, taken directly without any control or transformation, is truly suggestive of the main results.  
[00:35:03] Here, we have China's exports to the US as a function of US tariffs.  
[00:35:11] What we see is that in the first period, this relationship is flat, but during the trade war, we see that tariffs correlate negatively with China's exports to the US. There are  
[00:35:25] products where the US, due to higher tariffs, exports less, and comparatively even less relative to GDP.  
[00:35:36] Here, we have the same pattern for China's exports and for the exports of all other countries.  
[00:35:41] What we see is that other countries increase their exports to the US where the US imposes higher tariffs, and this wasn't happening before. This is like looking at the exports from the rest of the world to the US  
[00:35:55] as a function of tariffs in the subsequent period.  
[00:35:59] Here, we have exports to the rest of the world, and we see that they increase while the first period is advanced. Now, we move to the same results for Chinese tariffs. What we can extract is  
[00:36:14] the China-US tariff; we clearly see  That exports from China to the United States fell where there were higher tariffs, and we didn't see much action here. We didn't see exporters from the rest of the world  
[00:36:28] changing their exports to China much as a function of Chinese tariffs. Here, we think there are these opposing effects that, in principle, could be increasing, but what's happening is that there's a story about the PSUV (United Socialist Party of Venezuela) in La Vega, while  
[00:36:42] productive sectors, where China demands less from the rest of the world, and finally, we see that the rest of the world increased its exports to China as a function of China's size, although comparatively, it doesn't seem to have done much more than in the previous period.  
[00:36:57] We're going to control for these trends in the regressions, and now, as I was promising, this is the main regression, motivated by what I showed you before: how exports from each  
[00:37:10] origin and to each destination change for each product as a function of the tariffs from China and the US.  
[00:37:16] For this product, with a coefficient that varies by destination, but we're going to run the relationship separately for each origin: US-China, the rest of the world. So these coefficients will vary  
[00:37:31] when I present it.  origin and we also have these fixed effects by origin of Sting sector ok and I'll leave it for a minute to  
[00:37:44] process it because I think the key is you can't repeat again exactly, I mean, not far off, three regressions, let's say one for each origin, exactly here, three, then when Vettel or great 50 but here 3,  
[00:37:58] so think of the rest of the world as one country and we are and we are going to finish a review of how much the rest of the world exports and we fix a destination, the US, through different Omega products, look, look, so  
[00:38:14] the only variation you have between Omega products, so I'm looking for the origin, the rest of the world, and the destination, the US, different products, and for those products I'm looking at how  
[00:38:27] exports changed based on the tariff that the US puts on China, well, the variations are always between products, because tweet, which is what is restricting, because they are not  
[00:38:39] separate regressions for each pair, the only reason here is that it would be very similar to the only reason that Kaká, the Spirit Friends, do not vary by origin,  
[00:38:52] destination, they vary by destination, nothing more, so there is the regression, it's all right, if this except for  The coefficients and trends that are common to all origins—this is a  
[00:39:04] review. We're pooling all the data, and it's the universe, but if not, this will vary, which would be the same as running a separate regression for each origin and each destination.  
[00:39:16] And in fact, when we run it with all 50 countries, because it would be very cumbersome to do it all, we run it separately.  
[00:39:27] And well, there's the main table. There are many coefficients here, so let me skip ahead for a second and show you one that's more restricted, and I'm going to skip almost the rest of what  
[00:39:43] remains. Then I jump very quickly to heterogeneity, and that's it.  
[00:39:48] Here we have five columns that show me how exports from China to the US change.  
[00:40:00] Let me start again. I look first at the first two columns, and this is how exports from China to the US and from the US to China change, and their faces.  
[00:40:09] We focus on the first row; each one is a different regression.  
[00:40:15] But the way I'm going to tell the story is to focus on the row that is the tariff that the US puts on China. I'm going to ask myself how the tariff that the US puts on China...  The US imposes tariffs on China, affecting each of the responses. It's easier to tell the story this way because there are interactions  
[00:40:28] between responses. So, when the US imposes tariffs on China, Chinese exports to the US fall sharply  
[00:40:41] because the US reallocates from China, which is expected.  
[00:40:45] Second, when the US imposes tariffs on China, the US also exports less to China.  
[00:40:51] This is one of the different forces in the model. One that one might consider is a Lerner symmetry: the US imposes a tariff on China, an import tax, which is symmetrical to an  
[00:41:05] export tax. There is a reallocation towards the domestic market; the United States exports less to China. This suggests that there are supply curves with a normal slope; otherwise, you wouldn't have this area effect. It could also be that there is some kind of  
[00:41:20] production bottleneck where, because China exports less to the US, it is demanding fewer inputs from the US.  
[00:41:28] We move to the next column, which is how the exports of each country change, that is, the rest of the world, which I am calling, and should call, R& W here. Exports to the US increase, so this...  
[00:41:41] We call this what we're saying is that the US, sorry, what we're calling this is that each country is a substitute for China, that's why exports increase in the US.  
[00:41:52] What we also see is that when the US imposes tariffs on China, the rest of the world imports a lot and exports much less to China. There are really different stories, but the one that seems most natural to think is that China demands  
[00:42:06] fewer inputs from the rest of the world. And finally, what we see is that the rest of the world, in addition to the US tariff on China, exports much less to the rest of the world. Or rather, it exports more to the rest of the world. This coefficient is a  
[00:42:19] bit noisy when we look at it by country; in some cases, it's very strong, but there is an increase in exports to the rest of the world, and this is what we feel from a negative supply curve.  
[00:42:30] Now, these are the responses from origin to destination. I did n't show you all the destinations to save space. What I didn't show you specifically is that in China, exports to the rest of the  
[00:42:43] world increase with this tariff. And then we have the other tariff, the one that China imposes on the US. We also have the other tariffs that the US...  
[00:42:53] And China imposes tariffs on the rest of the world. This is being in the regression, but I'm not showing the coefficients because the results here are generally noisier, and some economically speaking, if a  
[00:43:06] friend does it, the point estimates aren't strong either. But there are two things that stand out quite strongly: when China imposes tariffs on the US, as we might expect, the US exports less to China; and what happens very strongly is that  
[00:43:19] the rest of the world exports more to the rest of the world. It does n't seem to be penetrating the Chinese market, but what happens is that the rest of the world exports more to the rest of the world when China imposes tariffs on the US.  
[00:43:33] Now, these are the answers, and it's good the story we're telling about not answering if you don't have time. You wanted questions, and if not, I'll wait. And I'm not quite understanding the identification  
[00:43:47] of where you're coming from, how are you presenting something as an aggregate response in some way, and why? That's what confuses me. The following: there are commodities, and the substitution of commodities is obvious, like in the case  
[00:44:00] of soybeans, the soybean case you mention, and I wonder how much that might be affecting the aggregate, but I'll answer.  Now, and then, from the most basic  
[00:44:13] identification, all of this that we said is within a non- aggregated product. When I say exports increase, I mean exports increase in the products that receive the most taxes.  
[00:44:26] Imports from the typical country increase to the rest of the world in the products that receive the most taxes within a sector.  
[00:44:36] And now what I'm going to do is aggregate those responses. How do I aggregate them?  
[00:44:41] Using these rivers, using the coefficients, I can predict how much the export of each country to each destination changed in each product.  
[00:44:52] I'm telling you, it would be a problem if it happened in many products, and then it's not done at the product level. What I'm saying might be happening, but the products where this happens are  
[00:45:05] a smaller fraction, so they wouldn't affect the overall estimate. It's possible we can discuss this in a second because I don't know if I didn't understand completely, but what I want to emphasize is that here the elasticity  
[00:45:18] is only one elasticity that is common to all products. I'm assuming that all products have the same elasticity of response within a country.  
[00:45:26] And then, to add this, I make a prediction of how much the exports of each country to each destination changed in each product. And then, using the weight of each one, one of them is practiced.  These products in  
[00:45:40] each country's export basket or make an aggregate  This is because I know how much to export. This will give me a prediction for how much Argentine soybean exports to China will change. Then, I use the weight of soybeans in China's export basket and the weight of  
[00:45:54] Argentine exports to the rest of the world. So I do that aggregation, and when you do that aggregation, I average this table. I want to show you this table and a graph with heterogeneity. This table does this aggregation. Once I have these estimates by product  
[00:46:08] and by country, I can aggregate using the weights of the export basket, and it gives me an average that the US reduced its exports to China, and China to the US very significantly.  
[00:46:20] But it also shows that the rest of the world reduced its exports to China and increased them to the US, and fundamentally, the rest of the world increased them.  
[00:46:29] I'm taking that coefficient that told me the rest of the world increases its exports to the rest of the world based on the Chinese tariff and the American tariff. I combine and multiply them according to which products were taxed, which products received tariffs, and what  
[00:46:43] the basket is for each country. And I add it up, and I can do this aggregation, and then I combine all the coefficients, and it shows that the rest of the world increased Exports to the rest of the world, so  
[00:46:56] for reasons related to the world, it's as if aggregate trade has increased based on these coefficients.  
[00:47:05] And now, to close, and then we open up to all those questions, what we do is run this regression separately for each country. And now, if instead of having the rest of the world as an aggregate with the same coefficient, we allow these elasticities to  
[00:47:20] vary by country, like Argentina having a different elasticity than China, or Malaysia, when we run this regression, what I can do is rate each country if, every second, each country had a  
[00:47:33] positive response to the rest of the world or to the US. Based on the US, I can say that country is a substitute or complement to China, or operates with increasing or decreasing returns.  
[00:47:44] Since what I'm showing you are simply seven examples, Malaysia, which I've been mentioning because it ends up being a winner, is a country that, in these estimates, has a beta that is positive  
[00:47:57] towards the US when the United States imposes a tariff on China, which tells me that Malaysia is a substitute for China, but it also has a positive return to the rest of the world on the horizontal axis. So Malaysia has  Increasing returns and  
[00:48:10] it's a substitute for China, which are forces that will lead it to export a lot to Mexico. It's in that same field as South Africa and the Philippines. It turns out that it's a loser too. These are countries that are suggested to have increasing returns but are complementary to  
[00:48:24] China, so in this case it's worse for them. Then we can bring in all the countries. I don't remember exactly where Argentina is in this graph. It's operating as a substitute for China in the US and with  
[00:48:37] quite strong decreasing returns. Argentina, and then we have the similar graph when China's tariffs, and again we have this kind of  
[00:48:49] similar cloud. Malaysia, in this case, continued, sorry, in this case, more grace in exports.  
[00:48:58] In the case of exports to China, it appears as a complement to the US and operating with decreasing returns.  
[00:49:07] And obviously what we can do is, once we have these coefficients, we can give a ranking of winners of the trade war. Winners, in quotes, I mean that exports increased the most.  
[00:49:18] Well, there's a lot of heterogeneity, so some countries are more expected cases than others.  
[00:49:25] Among these, how many? This is doing this country aggregation.  By country, a country like Malaysia, which increased its exports by approximately 10 percent, appears alongside other countries like Turkey and Romania. These are countries that are very specialized in some type of machinery and hold  
[00:49:39] strong positions. Losers include the Philippines, Ecuador, and Israel. These are countries for which we can try to tell a mini-story. Argentina did n't fare too badly in this context. It appears relatively high  
[00:49:53] in this ranking, particularly as it is increasing its exports. Argentina is relatively faring better than the US or China in the rest of the world, and that's why it's growing more than exports  
[00:50:07] from the US or China. Finally, the message I wanted to give you, and this will refer to Irene's question earlier about why I'm interested in the specialization pattern, is this graph. What I  
[00:50:21] showed you here is the ranking of winners and losers, in quotes again.  
[00:50:25] When I incorporated two forces, the fact that countries have different specialization patterns between products, between sectors, and between products, but also that each country has its own sigma and its own elasticity of  
[00:50:38] demand and supply, there are two forces.  What I can do is create this same ranking of winners, but without allowing countries to have their own elasticity. That is, using the estimates I used for the first one when there was  
[00:50:51] no heterogeneity. This graph compares—excuse me, compares—both responses. On the horizontal axis, I have the prediction of aggregate exports for each country  
[00:51:04] when there is heterogeneity in the coefficients, and on the vertical axis, it's when there isn't. What we see is that the vertical axis shows much less telephony than the horizontal axis, which tells me that explaining this  
[00:51:17] ranking of winners and losers, and in particular the magnitude, depends a lot on taking into account that each country has its own response. It's not just that they are specialized in goods that receive high or low taxes, but that they respond differently based on those factors. And  
[00:51:31] that's the conclusion. We have a few more things we're working on, but I don't want to go into detail about what they are. I just want to make it clear what we learned. We learned that,  
[00:51:45] typically, in a trade war, the average country increased its exports— both the US and the rest of the world—in response to tariffs.  
[00:51:53] In a rather stylized model, it suggests that countries tend to substitute.  The US and operating with negative supply curves, although there is a lot of heterogeneity when we make that partition,  
[00:52:07] when we allow it in the estimation, and somewhat surprisingly, when we aggregate, we see that the trade war seems to have created trade opportunities, not only generating allocation.  
[00:52:18] We see that this ranking of winners and losers, what I was saying later, channels and values ​​in quotes, depends to the extent of these responses that are specific to each country, but also on the patterns of specialization. I have to say how we do these  
[00:52:31] aggregations, and I would be the first to, let's say, make this type of comment if I weren't the person presenting the paper. These aggregations that keep the fixed effects of the regression in a model, these effects would be endogenous, and we  
[00:52:46] are not identifying that. So that's it, that's the just-in-time work. I finish and I'll leave it there. Let them make the comments.  
[00:52:54] What professionalism, I love you! One question before moving on to the comments that I prepared: can I go back a minute to the graph, to that grass that we could  
[00:53:08] interpret as... in reality, everything is a matter of composition, and that the homogeneous part... in reality, what is happening is that...  Different compositions  
[00:53:20] at a more disaggregated level than six digits, and then, well, you manage to capture it in the data, but there's a reason why those sigmas aren't the same across countries of origin because the goods are different.  
[00:53:34] Ultimately, it's a composition effect. What happens is that you don't observe it. It could be a correctional effect, understood in a broad sense, even a differentiation related to  
[00:53:49] whether they are allied countries, which has to do with many things. But, to put it simply, there are characteristics; some are visible because they are different industries, and others are much smaller, and well, they are all  
[00:54:02] encompassed in that sigma. Yes, within that product, there's a subtlety: whether this story  
[00:54:13] depends on variations in tariffs within the six-digit category, and that variation isn't that high. So, the idea here would be that  
[00:54:27] within the same category 6, we have firms; we are much more disaggregated.  
[00:54:36] We would be saying that this firm, all these firms, are subject to the same tariffs within this category, but it's telling me that later, I remind you to study, there is a composition effect, which is why certain countries may be  
[00:54:50] Specialized in firms or products that are more substituted for the US than other countries, so it's not a characteristic of the country, but rather a characteristic of the firms or  
[00:55:03] products concentrated in the country. And one could imagine that these firms or products could be located elsewhere. And if this is 1 and 2, they were for that.  

## Discussion (00:55:15 – 01:11:05)

[00:55:15] Yes, I don't care. Well, Irene, it's all yours  
[00:55:26] in the comment. Well, and what does Pablo have to do?  
[00:55:37] And how do I have to do it? I have to open it first. I'm going to be clear, and then one thing is that I would have it somewhere. What do I have to choose and orient? I'm streaming. Let's see, let's  
[00:55:50] see, Windows, to edit.  
[00:55:59] Yes, yes, it looks big. Yes, well, perfect. Well, what I have prepared is quite brief. And how much time do we have? 10 minutes, 3. 15.  
[00:56:13] Give it a try. Perfect. Well, I already applied it more or less, Pablo, but to summarize, there is a tariff war between the US and China. It starts in July 2018 when the US  
[00:56:25] raises tariffs on many Chinese products, and then between July 2018 and December 2019 there are all the mutual tariff increases between the two.  
[00:56:39] Countries that are raising tariffs on increasingly larger groups of products.  
[00:56:47] In January 2020, there was a preliminary agreement, and little by little, tariff reductions began, along with an increase in trade flows.  
[00:56:56] So, the idea is to see what happened to trade flows during this period of the trade war between the US and China.  
[00:57:07] What do they want to see? Well, the new thing is that, on the one hand, they will obviously see the effects on exports between the US and China.  
[00:57:17] They will also see exports from the rest of the world to the US and China.  
[00:57:25] And the most novel aspect is that they will also look at what happens to exports from the rest of the world to other countries.  
[00:57:32] So, to show that there are effects in all these directions, and that tariffs that rise only between the United States and China end up  
[00:57:46] indirectly affecting exports from country X to another country X where there was no tariff change.  
[00:57:58] The empirical strategy is simple: they have bilateral trade data at the product level. The products are what are called omega values, six- digit values, as Pablo already explained.  This is quite disaggregated,  
[00:58:11] even the levels of disaggregation that follow, with which, when one refines it, it's not much that is refined in terms of making figures;  
[00:58:22] this is very disaggregated, and what they estimate, then, is several revisions for each exporting country and for each destination.  
[00:58:33] But they need to aggregate the destinations; they ca n't estimate for 50 different destinations, so they aggregate them into three: the United States, China, and egg remnants. What they estimate is  
[00:58:46] one regression for each of the exporters; in reality, there would be three for each one, if I understand correctly, because they would be for each exporter and for each of the three destinations. Then Pablo explained that there is actually a coefficient of the  
[00:59:00] ring and 36, but leaving the pretension aside, there would be three regressions for each exporting country. Here we have, then, the change in exports of product W,  
[00:59:14] how that is affected by the tariff imposed, or the change in tariff between the US and China, the change in the angel of Christ from China to the US, from the US to the exporting country, from China to the exporting country, and from the US  
[00:59:34] to all exporters, and premiums that are not...  Neither China nor the change as  
[00:59:45] above is the destination, it's fine because both combinations, so the effect is similar to everyone. I would  
[00:59:59] care which one it is, but while this is estimated for each exporting country and each destination, there are 3  
[01:00:14] more destinations than that, it's impossible because there isn't that much tariff change. So, in the rest of the world, tariffs aren't changing that much, so you ca n't estimate a response for each one. So, this is a lot of  
[01:00:28] information that they are estimating because, again, these are reactions that will be different for each country of origin and these three destinations. So, the question is how is this estimated?  
[01:00:41] Where are they getting the variability from? Where does it come from that they can estimate this? Well, first, this is a reduced form, that is, it's neither a supply equation nor a demand equation, it's an equilibrium situation, and  
[01:00:56] that's perfect, but it has the challenge that this review is swept to capture a specific phenomenon, that is, this tariff war that occurred at that moment. So, it's the reaction that  
[01:01:09] occurred to that tariff change, which cannot necessarily be extrapolated to another situation. So, for example, this couldn't be estimated using a long time series because that would  
[01:01:22] be  By combining many episodes, you would be combining many different equilibria, and 6 or econometrically it would be invalid. So something they resort to is a trick, in quotes, of saying,  
[01:01:36] well, what we're going to exploit is the variability between products. So instead of taking a time series, we're going to say, well, I'm going to concentrate on this episode.  
[01:01:45] Estimated reactions, which I don't know what it was called, to begin with, elasticity. It's something valid for studying what happened during this war, and what I'm going to use is that the  
[01:01:58] tariffs, the changes in tariffs, were different for the different products. So I can compare the tariffs that increased the most on the products and how they reacted. That's how  
[01:02:10] this regression is identified. The assumption, of course, is that these beta coefficients we just talked about are the same for all products from the same country and to the  
[01:02:24] same destination. So the reaction of soybeans is the same as the reaction of automobiles. That's the necessary assumption.  
[01:02:36] And then Pablo didn't talk much about the model, but they have a well- developed model in the paper.  
[01:02:46] What function is this model fulfilling? Because this is actually an empirical paper where they are estimating these reactions. So why do we need this whole model?  On the one hand, the fundamental model, to make it very clear what  
[01:03:01] assumptions are being made to carry out this estimation, first, the goods are differentiated by country of origin. This is necessary; that is, soybeans, even though they are a commodity, have to be differentiated. I don't want to get into  
[01:03:15] soybeans; there are thousands of products that are obviously much more differentiated. So the fundamental model is that goods have to be differentiated by country of origin, and that's how they enter the demand. There  
[01:03:29] have to be non- monochromatic preferences and substitution patterns that are also differentiated by origin. That is, the goods not only enter separately into the utility function,  
[01:03:42] but they also enter asymmetrically into the utility function according to their origin.  
[01:03:51] Then, on the other hand, if there has to be symmetry in substitution, cross-testing the jump we were just talking about, to ensure that the values ​​are the same between products in the  
[01:04:03] demand functions, goods of different origins have to be in the same way, but not through origins, but through products. That's why that sigma and prime didn't have an omega, because  
[01:04:18] those parameters have to be equal. The demand function, and the other thing is that prices are given so that the model, beyond being able to understand all this well, what we need to assume for the  
[01:04:32] estimation, also obtains a proposition 1 that Pablo didn't discuss, but that allows us to interpret the empirical results, and I'll come back to this later. Now,  
[01:04:45] I'm not going to comment on the tables, I'm not going to comment on the results, Pablo already did that, but I do want to show you some graphs that he also showed, and here I'm going to cheat a little because these graphs  
[01:04:58] aren't actually the regression results, but they do allow us to quickly and easily discuss the regression results. On the one hand, the basic results find that Chinese exports to the US... what you have to  
[01:05:13] look at here is the blue line; the gray line is prior to the war, and it's precisely used to make the argument that this was something that happened there, that before, the products didn't have different  
[01:05:26] price trends, and that therefore the empirical strategy is valid. So here we see this result, which is expected and obvious, that is, tariffs went up, exports from China to the US fell.  
[01:05:37] Demand is negative, that happens, the counterpart is the same. Here we have this blue curve.  
[01:05:47] US exports to China also fall. Furthermore, they quantify this better than this simple graph I 'm showing here. These are the three basic results, but the most  
[01:06:00] interesting thing about the paper is seeing what happens with the rest of the world. I found the name " Buy Standards" for other countries very clever; they were standing still and then the  
[01:06:15] war hit them. One thing they find is that exports from the rest of the world to the US increase. So, these vice versa.  
[01:06:27] When what's written is on average, then they have country-by-country results, which were those graphs that Pablo showed at the end. But if we take the rest of the world as a whole, the rest of the world  
[01:06:39] exports more to the United States, and this is the most notable thing. The rest of the world also exports more to the rest of the world because of changes in US tariffs.  
[01:06:52] When China's tariffs change, it 's the opposite. Yes, but to avoid complicating things, let's focus on this part.  
[01:06:58] So, on these results for the rest of the world, that's where proposition 1 of the model comes into play to interpret these These empirical results  
[01:07:12] are consistent with certain things happening within the model parameters. One is at the level of US demand, given that exports from the rest of the world to the US  
[01:07:27] are rising. This allows us to infer that, at the level of US demand, goods produced by China and the rest of the world are substitutes.  
[01:07:39] Yes, because when tariffs against China rise, Chinese exports fall, while exports from the rest of the world rise. This also  
[01:07:52] allows us to infer—and here the important result is this—that exports from the rest of the world to the rest of the world are increasing.  
[01:08:01] Here we see a positive interdependence in supply, which they interpret as meaning that, at the level of supply from the rest of the world, the aggregate slope is negative. This is because  
[01:08:14] all these exports are increasing, not only from the US but also to the rest of the world.  
[01:08:23] Ultimately, the important thing here is to show how interdependencies exist in exports and how changes in tariffs between China and the US have repercussions through  
[01:08:37] interdependencies that occur in demand.  As for the supply side and what's happening in the rest of the world, it seems a bit  
[01:08:52] surprising to me to use this terminology of an aggregate and negative slope, and it seems to me that this is a very specific way of referring to what's happening here. It's because  
[01:09:04] we have a competitive market, and I hope we can generalize this and think of it more generally as positive interdependencies in supply, where exporting to one country makes it  
[01:09:17] easier to export to another. This is a phenomenon that can occur at an aggregate level, but it can be driven by value chains, by things Pablo mentioned,  
[01:09:29] but it can also occur at the firm level. And here I want to take a second to comment on a result from one of my papers that is totally in line with this. It's a paper with Facundo Albornoz and  
[01:09:43] Manuel Ornelas. What we look at here is an episode that occurred in 1997 when the US suspended tariff preferences for Argentina, that is, it raised tariffs on  
[01:09:56] Argentine products. And what we found there is that Argentine firms Using firm-level data is a different approach, but we found compatible results because these Argentine firms that stop  
[01:10:09] exporting to the US also stop exporting to the rest of the world. This indicates that there are economies of scale, some kind of fixed cost, something that generates positive interdependence between destinations.  
[01:10:22] In our case, with PayPal, they can't see interdependence between products because they're assuming that all products move the same way. In our case, looking at the firm level, we find that for  
[01:10:36] products, it's the opposite: at the product level, there's substitution and negative interdependence. The Argentine firms that continue exporting to the US substitute one product for another; that is, they  
[01:10:48] stop exporting products where tariffs have increased and instead start exporting other types of products.  
[01:10:59] Well, I'll leave it here, and congratulations to Pablo for it because we're... well, thank you very much, Irene. Great, Pablo, do you want to reply, Irene? If  

## Q&A (01:11:05 – 01:38:12)

[01:11:11] possible, go later. Yes, much better than a lot of clues about how to present the facts and ideas of what we  
[01:11:26] do, so I'm going to adopt it. I mean, it all seems very consistent with the way in which...  We thought about it, there's just one  
[01:11:38] thing I wanted to clarify when describing the components of the model,  
[01:11:50] when describing why we use the model. We use the model to establish the assumptions that motivate the regression, that justify this specification.  
[01:12:00] There's one that you mentioned, which I think we need to explain better: what are given prices? So, as an allusion to this, that we're in an equilibrium relationship, well, there we have to do some juggling  
[01:12:14] to be able to take the model from an equilibrium, to differentiate all the equilibrium conditions when tariffs change, to obtain a relationship in changes that  
[01:12:28] is then consistent with what we run a series of assumptions on. But when we do that experiment, it's scanning the prices because, in principle, there aren't. So the first  
[01:12:43] regression that I show, let's say, is one of all prices. Well, on the other hand, yes, but in principle, what I'm doing is, given  
[01:12:57] the prices of all the other commodities that will end up entering as an error term in the regression at the product level, I'm going to choose the price of the export good. Let's imagine I'm running the  
[01:13:11] regression of how much  Argentina exports to the rest of the world. I'm going to look at it, I'm going to let the price of Argentine goods adjust so that there is equilibrium in the market for Argentine goods in the world, given all the  
[01:13:23] prices. So, the prices that I'm keeping constant would be the prices of all other goods except the price of the good from the country I'm looking at.  
[01:13:35] So, there I create equilibrium in that market, and when I create equilibrium, the market is what allows me to interpret the stigma of that country and the idv of that country in the beta coefficient. So, here, that's why that beta is like a combination of the supply and demand coefficients of that country,  
[01:13:53] and then if there are all the other effects of the other countries, these effects that enter into the error term, we can find conditions such that they are still changing. That term is quite intuitive, and when it will be [a] ...  
[01:14:25] Well, from Argentina, when I show you the equation for Argentina, how I get it, and the price is adjusting in the macro, it's like it gave me a detail of how we implemented it, but it's good and under control, also totally agree, I had it  
[01:14:39] in mind, and as obviously, you're exposing the different papers, like in the relationship, and I totally agree, this is analysis, and a little related to the question you asked me, that is, a lot of this action that's happening at the firm level, Wedding Prod,  
[01:14:53] and well, there are these regressions that you do that show that they are consistent with this type of interdependence, that there is complementarity, the firm leaves and not only stops contributing to one country but exports to the other side, so it's like the symmetric effect,  
[01:15:06] let's say, losing market share, you lose there, but at the firm level, and we can't capture that, so thank you very much for the discussion. I have a couple of questions/comments, I do  
[01:15:21] n't know if I'll be able to answer them, let me do a little bit of oil, Díaz, a little bit, like, why not, I don't understand this either, so maybe I can't say it so clearly, it  
[01:15:34] confuses me a lot, it seems to me, in principle  The natural way for me to do this would be to do it by product, estimate it by product, and then add in some  
[01:15:48] way what does the estimation product by product mean?  
[01:15:53] Product by product, not that an observation is a product.  
[01:15:58] Clearly, what would one gain and lose if one does it by product?  
[01:16:07] In principle, one would think one captures much more, like the general equilibrium in that product; there is more consistency in all the elasticities about what is happening with that product. What one  
[01:16:20] loses is the interdependencies with other products and the richness of substitution patterns that you have between countries for the same product, which I think is essential.  
[01:16:35] Could one capture it, or could one capture it much less, or would have much less freedom to capture that? So, in a way, you are privileging that heterogeneity in substitutability  
[01:16:49] between countries, even for the same product. You gain in that, saying that even between cars from Argentina and Brazil they are more similar, and I allow that substitution to be  
[01:17:02] different from that between Brazil and Thailand or Argentina and Thailand, to say something.  
[01:17:11] You gain in that possibility of having richer patterns of substitutability, but on the one hand, you lose this...  It may differ by product, but there's also a  
[01:17:25] consistency issue that I'm not clear on. To what extent is there consistent consistency between the different regressions? What's implicit in the underlying substitutability patterns that manifest in the different regressions?  
[01:17:40] For example, if a question is the same, a specific question would lead to the rest of the economies of scale, or the interpretation of economies of scale, whether there's consistency between the different  
[01:17:53] regressions in estimating the economies of scale a country might have. I did n't quite understand that. I wonder if it's somehow imposed or not. So, given all this confusion, and this is a general question of approach, the  
[01:18:08] more concrete question would also be: is n't there an intermediate way, let's say, neither at one extreme nor the other, to restrict it a bit and find  
[01:18:22] something in the middle? The idea is that perhaps by going too far to the extremes, more can be lost than by doing something intermediate where perhaps they estimate a few digits but  
[01:18:37] obtain a regression for each group of products, for example, and where the restriction is that the substitutability patterns are the same for certain... I don't know, to put  
[01:18:51] certain restrictions on that. Having some freedom, but less than what exists now in the city, the patterns of structures between countries, but more freedom between products.  
[01:19:06] We're going to talk about several things. You said the first one, and as a mechanical matter, let's say, how to identify this type of effect of variation  
[01:19:21] between products. So, different products have different tariffs. So, there's a sense in which I do n't see very well what other source of  
[01:19:36] variation can be exploited other than that between different products at different tariffs.  
[01:19:42] And what we're looking at is the response that a country, or different countries, have between products. So, that's how it is.  
[01:19:55] Then there's the question of how flexible those responses can be. They could never vary at the level of the product itself, but they could vary by groups of products. We did something like that. In  
[01:20:09] fact, we have this whole analysis partitioned. There's a vision of this paper that is much more ambitious and larger, which is to  
[01:20:21] estimate economies of scale by sector. And what we were doing was this analysis with a sector index. So, suddenly you have that the betas vary  
[01:20:35] by product and by sector. And in fact, there's a version in which we thought, if I go to the model— the model, since I said it very quickly—by Steve and a micro-foundation  This is one of  
[01:20:50] the typical offers that has two parts. This micro-foundation has one part that relates to a parameter that tells you how easy it is to reallocate factors within a country between products, and another part that tells you about  
[01:21:03] economies of scale. In fact, as we initially conceived it—and there are many previous versions that were like this—we thought that this would be a function of the country; certain countries have  
[01:21:16] more flexible labor markets. But the range was a function of the sector; certain sectors have stronger economies of scale than others. So in that version, the beta is a sum of a country component and  
[01:21:29] a sector component. So that version, I think, is more flexible. It's not intermediate with what I'm showing; it's more flexible,  
[01:21:41] but it captures the notion that there's something country-specific and something sector-specific. And that seems to me something we should revisit. Well, the  
[01:21:56] story is that at the beginning we had that; it held us back a lot. We did n't move on to other things, but we refined others a bit to justify eliminating it. But that version would capture something of what you're saying, it seems to me, and it would be interesting to look at it again.  
[01:22:13] And in fact, I think we have to. Now you mentioned something else that I'm very aware of, which is the restrictions between regressions of these coefficients. In other words,  
[01:22:25] to what extent is it seen that, well, here's the coefficient, it's the same, and then when you look at the regression, this exercise, this coefficient varies  
[01:22:37] between several regions. But I wouldn't be talking about regressions I'm running, meaning it doesn't vary depending on whether I'm looking at the Chinese tariff or the American tariff. I mean, it  
[01:22:50] varies by origin and destination. But when I show them all, let me rephrase that: when I run  
[01:23:07] the regression, there's a coefficient for each tariff, yes, but the coefficients that underlie this regression—what you see, let's say, the curve—are the same in the beta that comes in here, in the beta that comes in here. So, what you're  
[01:23:22] telling me is that the results are consistent with an economy of scale, which is the same in the US coefficient as in China. And in fact, the regressions show you that when we  
[01:23:36] allow each country, everything is... When I show you this table, if I show you the table that's common to all countries—not country by country, but as if I 'm telling a story that's somewhat consistent—what makes it so...  
[01:23:51] The reading is consistent with this idea that there are demand flowers coming from the side of white chili, Pablo, one little thing, I mean, of course, you, perhaps that wasn't taken into account, and voices, island economies  
[01:24:04] of scale in exports, yes, the effect of economies of scale, and that question that sometimes only one of the product is only one at the product level,  
[01:24:17] but at the level of realism, let's say, its product level is not defined, but at the level of Espino, it's all demand, let's say, and it's all demand that this derives from substitution or productive linkages, that's how we're thinking about it, but it's  
[01:24:31] not, I mean, it's as if the effect of demand is added, let's say, if only the scale only depends on how much to export in total, well, the question is, how much of  
[01:24:45] this story of the US putting China, if what I'm seeing, what happens between China and the US, we understand, the United States exports less, imports less  
[01:24:58] from China and exports less to China, and the US turns out to export more to the rest of the world, and also China exports more to the rest of the world, I'm not showing it with respect to the US tariff, but then I  
[01:25:12] love the rest of the world and you  I said that the story I'm going to tell you—I could tell you others, in fact—but the story that seems most consistent with the whole configuration of coefficients is that the rest of the world operates with economies of scale. For the rest of the world,  
[01:25:26] China is a substitute, which is why the US increases. This is interpreted as me saying that it gives you scale, a lot, it gives you size because you're growing. The US is positive for you overall.  
[01:25:38] At the same time, you're exporting more to the rest of the world. So, in addition to the model, this is telling me that you have a negatively sloped curve, economies of scale. And now comes the trick, which is: but then you should export more to  
[01:25:52] China too, but you're exporting more to China.  
[01:25:56] And that's where we get this idea that this has to be happening because China is demanding less. Besides all the negative demand from China, which comes from productive elements, there's one interpretation, of  
[01:26:11] course, there's another interpretation, which is a bit like Irena's, which is to say, "Oh, but well, maybe these spillovers are different layers. You want to be in the US and grow from the US. It has a different style, a  
[01:26:25] different complementarity on the supply side for your productive processes."  The rest of the world is almost entirely dependent on China, yes, that may be, but there are things that suggest to me that this productive reorganization might be happening,  
[01:26:39] and one is the following: with this tariff, the US tariff that China imposes on China, China is actually responding to this tariff by exporting much more to the rest of the world. This is  
[01:26:53] the response to the tariff that the US imposes on China; China is exporting much more to the rest of the world. China  
[01:27:07] stops exporting to the US and exports more to the rest of the world, and that in fact makes it very difficult for the rest  
[01:27:20] of the world to compete. So that's what makes it so surprising that the exports of the rest of the world increase. The rest of the world has to become more competitive. The rest of the world is competing with China, and that suggests to me  
[01:27:35] that some forms of competitiveness are coming because they are talking about scale, and the reason that the rest of the world is somewhat affected is also because of the Chinese, because China is demanding less.  Now, what I  
[01:27:47] wanted to tell you was something about this story of defendants and productive parking lots.  
[01:27:56] We were looking at some data on that, and it's difficult to see the input-output matrices at the level where they're disaggregated to six digits.  
[01:28:03] But one thing we were doing just a couple of days ago, we just started doing it, is that if you look at the input-output matrix from the perspective of the Chinese, and you look at the Chinese customs data on what  
[01:28:17] the same firms import and export, and in China what you look at is the matrix, like the total output matrix, which is about the products that Chinese exporters import, and I saw how loaded it is, all the way to the diagonal,  
[01:28:31] and there you can see quite a bit that, well, 30% of Chinese exporters import the same product, which could be telling you that when these guys export less to the  
[01:28:44] US, they also import less of that product. It's like their manager is going in that direction, but it's different in terms of how to make the story compatible between  
[01:28:57] the different coefficients, and well, you have to do some juggling, but in light of the  Thinking about it this way, we can have a story where the amount of offers remains consistent and is the same and consistent with all the coefficients.  
[01:29:11] Pablo, I have a question related to this. I mean, this story about intermediate inputs seems perfectly plausible to me.  
[01:29:22] I believe it perfectly. But thinking about it, trying to understand a little more, you have the rest of the world, and we're trying to figure out if the US exporting more makes it easier for them to  
[01:29:36] export to other places. But on the other hand, we have that they're exporting less to China, and isn't that also a release of resources that makes it easier to export elsewhere? So when you have this  
[01:29:50] proposition, which would actually be a kind of partial derivative, so to speak, you're not changing the US and China at the same time, but you 're alluding to something  
[01:30:02] important. That's why I think that once you bring up this idea of ​​the juice product, it becomes a  
[01:30:16] force in itself that can explain everything, because, well, not everything, but it can lead us to say, "Ah, but what's really happening here?"  It's that because you're exporting less to China, you're exporting more to the rest of the  
[01:30:30] world. That's pretty much what you're saying, right? Exactly. And that force is there, and I think it's a quantitative issue. What I do want to say is that it's somewhat on my left; it's  
[01:30:43] purely a quantitative issue. But I'm left with the fact that there's a very large increase in Chinese exports to the rest of the world.  
[01:30:51] So, you have to... I mean, you're exporting less from China from a certain base. You have to do some aggregation to see, but these Chinese exports to the rest of the world are huge and are increasing with a coefficient that is more or less the  
[01:31:04] same, in fact, with a different sign, which means that exports to China are decreasing.  
[01:31:09] So, well, you're facing that greater competition there.  
[01:31:15] But I agree. Qualitatively, one could tell a story where the low demand from China is also promoting its exports to the rest of the world. What we should really see is how strong these  
[01:31:28] input-output matrices are by origin, not from China to the different origins, but against the US. The data I just gave you was from any origin, but if it appears, I think it's a super good point.  
[01:31:45] There are others that...  They compete, explanation of this effect of economies of scale. I honestly didn't have a product for anyone in this story anymore, it wasn't even difficult.  
[01:32:00] I think there's a story you can tell with a story, and very frequently the comment came up in even more preliminary versions of productive linkages,  
[01:32:13] productive sales, but there's always something that's a bit subtle, which is that if they were open productive links, it has to be that a product uses itself in production because what we're identifying is the product that the US gives to  
[01:32:27] China, it exports less to China, it doesn't have to be that this product is also used in production, so that's the story that I always thought could be consistent and compete, explanation, I think there's that one, then there's another one that has to be, which seems to  
[01:32:40] me to be like between bidder and routing, which is like this: I've been to China to export what you can't because the United States is that China exports well to Korea and Korea  
[01:32:54] exports to Vietnam and Vietnam exports to the US, so cross-trade increases from the rest of the world to the rest of the world because before Korea exported to Vietnam and now it exports to the US, but the value increases 0, that 's another one, I don't know if I remember well, but  
[01:33:13] many times it's the categories  Rock arts and then yes, maybe some have and others don't have the six- digit categories, maybe yes, if there were a way to  
[01:33:28] distinguish sectors, like dividing between those that have more and those that have less, according to what some explicitly say, includes the parts, they are included there, and then maybe if there were a way to do what is done, it is  
[01:33:42] a way to test those parts that are also for that, yes, totally, also separate a little by products that are classified as inputs with others that are not, that's how it has been, the name of the  
[01:33:55] description is 16 and not totally, they are things that we have to distinguish, you're going to try to find like the big tuxedo and there is another one, yes, yes, a partition would have to be made by  
[01:34:10] that type, it can also be done a little by sectors, that is, when you do it by sectors, some sectors naturally, well, have more, with great, I don't know, here, here we have sectors that are much more, I mean, so  
[01:34:24] that partition can also be done, but if it happens like these two stays, it would be good to be able to, Neil Down, which ones, if they would take out gray, for example, and this I imagine doesn't have as much  
[01:34:39] weight because there are fewer positions, I do n't know, in the  Statistics, imagine that I imagine that what happens in adding to the sun weighs less, but if you were to take it out, if you were to show that I don't know, it's not a farmer effect, and that's some  
[01:34:53] st, how to think about this, if it's done by sector, I think it would give us, if I have active, but we had, how do they estimate by sectors, we have to  
[01:35:06] regenerate, well, let's take it out like that, the country, and that way we show who showed something, no, no, no, well, thank you  
[01:35:25] very much, and thank you, I persevered at the beginning, I think I had the illusion that it could  
[01:35:36] be in English, but if not, it happens that this audience finds it difficult, our audience, to follow, in general, many find it difficult to follow anything, mathematics,  
[01:35:49] most of the presentations are shadows, and if not, there are no statisticians and econometrics, but if you  
[01:36:00] were enslaved, Irene, mandalas light, that if you, the hands, but I clarify that there are two, it's line, but you did n't see them, but I mean, I mean, you make me, I  
[01:36:15] liked how the phrase, until then, what will help me is to look again to remember, go ahead, go ahead, I'll send it, but well, what I wanted to say, excuse me, one  
[01:36:35] more second, but with the contempt, give a little  Also, something I was going to say, without going into too much detail, is that this is what ultimately gives you an offer, and allows you to talk about an offer  
[01:36:48] at the origin level, not, for example, making different pricing decisions according to different destinations or things like that. I understand that if the prices don't vary bilaterally, then you're talking about being  
[01:37:03] dependent on the supply curve, and that stems directly from having those prices given, and not, on the other hand, a price inclusive of the market.  
[01:37:12] There's no price in, into the market, like Air France, but it doesn't change with tariffs, only given by trade costs, of course. And so, does this allow for a supply curve of this type of thing? It's something else entirely, as if it has been requested to  
[01:37:27] complement it, because if I give you a supply curve that has different elasticity for each destination, it starts to give me a greater degree of freedom to explain. I could have done whatever I wanted, so that's restricted.  
[01:37:41] Well, thank you very much. Many presentations for me.  
[01:37:52] We're going to upload the presentations, so you'll also be able to see Irene's comments as a presentation, things like that.  
[01:38:06] The controls, so she no longer has the power to upload them. We're thrilled.