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. = # Optimal Tariffs under Geopolitical Risk: A Theory of Strategic De-risking Authors: Ina Simonovska, Ahmad Lashkaripour Discussant: Matilde Bombardini Video: https://www.youtube.com/watch?v=U_ssYRB6uPs&t=9149s ## Talk (02:32:29 – 03:03:15) [02:32:31] alright, we are very excited to have Ina presenting on [02:32:52] optimal tariffs on a geopolitical risk [02:32:54] you have 30 minutes [02:32:56] great, thank you so much for putting this paper on the program [02:32:59] we're really excited to be here [02:33:01] this is joint work with Ahmed Lashkaripur [02:33:03] who's sitting right over there [02:33:05] and we'll gladly take your questions [02:33:07] okay [02:33:08] so the motivation for this paper [02:33:11] is a twofold [02:33:13] so first we've seen overall a shift [02:33:16] in the global trading system [02:33:19] and specifically in the global trade policy system [02:33:24] so on the one hand [02:33:26] over the past two decades [02:33:28] the export restrictions have effectively quadrupled [02:33:32] so what I'm showing you here is different types of export restrictions [02:33:36] it's a count [02:33:38] so it's an extensive margin, it's just a number [02:33:40] and you can see that they've basically risen from about 3,000 [02:33:44] to 13,000 in 20 years [02:33:47] and a lot of these are concentrated in inputs [02:33:51] some critical inputs such as rare earths, semiconductors [02:33:55] medical supplies during COVID, etc [02:33:57] and then on the other hand [02:33:59] from the US point of view we have seen [02:34:01] what appears to be a response [02:34:04] so focusing on the first Trump war [02:34:09] so 2018-2019 [02:34:11] an interesting observation arises [02:34:13] so these were the section 301 tariffs [02:34:15] that were introduced in China [02:34:17] and if you look at the growth of China's import share [02:34:21] five years prior to these tariffs [02:34:24] we've seen that for the section 301 tariffs [02:34:27] that growth has been a positive 4% [02:34:30] whereas for the non-301 tariffs [02:34:33] so the non-China tariffs [02:34:35] we've seen a decrease in the import share [02:34:37] by 3% or so for the United States [02:34:40] and this looks at categories that were traded [02:34:43] in both the beginning and the end here [02:34:46] so this started to get us thinking about [02:34:49] whether the US is imposing tariffs [02:34:53] in order to deal with strategic [02:34:56] foreign supply disruption [02:34:59] and one way to kind of summarize that [02:35:02] is that we're seeing that somebody that we admire [02:35:05] said herself, Janet Yellen [02:35:07] so this is not just rhetoric coming [02:35:09] necessarily from politicians [02:35:10] it's coming from economists [02:35:11] and she said we cannot allow countries [02:35:13] to use their market position in any one raw material [02:35:16] to disrupt our economy or exercise [02:35:18] unwanted geopolitical leverage [02:35:20] so therefore we're going to try to formalize [02:35:23] these types of discussions in a model [02:35:26] and we're going to try to ask whether [02:35:28] the US tariff policy is a welfare-improving [02:35:31] or a geopolitical risk [02:35:33] so let me lay out the details [02:35:35] of the background of the model [02:35:37] and how we're going to answer this question [02:35:39] I'm going to lay everything out [02:35:41] with two countries and a single sector [02:35:44] and then I'm just going to extend it to multiple sectors [02:35:46] because the extension is going to be very natural [02:35:49] but I wanted to kind of derive all the results for you [02:35:51] in a simple environment [02:35:53] so we're going to have two countries [02:35:55] home and foreign, CSD man aggregator [02:35:57] with trade elasticity epsilon [02:35:59] and we're going to have round about production [02:36:01] so what does that mean? [02:36:02] so suppose you're producing a variety of omega [02:36:04] in country N [02:36:06] how is that produced? [02:36:07] so you have the productivity level [02:36:10] in that country [02:36:11] and then you have labor [02:36:13] LN [02:36:14] and inputs [02:36:17] a set of inputs MN [02:36:19] that are combined in a CS fashion [02:36:21] so beta here's the labor share in production [02:36:24] and there's going to be two wedges [02:36:26] so the first one is something [02:36:27] you can control over [02:36:28] so these are your iceberg trade costs [02:36:30] DNI [02:36:31] they're technological [02:36:32] and then we're going to have two policy instruments [02:36:35] the first one is the import tariff [02:36:37] so Tau NIM here [02:36:40] means that it's the import tariff that country I [02:36:43] so think of the home country [02:36:45] the United States imposes on imports from country N [02:36:47] and then export taxes Tau NIX [02:36:50] which are the taxes that the foreign country [02:36:54] N imposes on exports to the home country I [02:36:58] so just to kind of summarize [02:37:02] how the model is going to work [02:37:04] before I show you some equations [02:37:06] we have retail firms that are purchasing [02:37:08] and bundling consumer goods [02:37:09] so that's our aggregate consumption [02:37:11] production firms were purchasing bundled [02:37:13] intermediate inputs [02:37:14] so that was the M that we saw previously [02:37:17] and each firm is optimistic [02:37:19] meaning they treat aggregate variables [02:37:21] as given so think monopolistic competition [02:37:24] now here's where the interesting part is going to come in [02:37:27] so there's going to be geopolitical sourcing [02:37:29] externality and the way that it works is as follows [02:37:32] so the unit prices for consumption [02:37:34] and intermediate inputs [02:37:37] depend on the foreign export tax [02:37:39] Tau NIX [02:37:41] which itself will optimally as I'll show you [02:37:44] depend on the foreign's aggregate market share [02:37:47] but this aggregate market share [02:37:49] is of course determined in equilibrium [02:37:51] by each firm's individual sourcing decision [02:37:54] however an individual firm being optimistic [02:37:57] does not internalize the fact that [02:37:59] it affects the overall aggregate market share [02:38:02] which then drives export taxes [02:38:07] so therefore terrorists will potentially [02:38:09] have a corrective role here [02:38:12] right so let me go through [02:38:14] a little bit of details of the model [02:38:16] so we can see how this intuition plays out [02:38:20] so this standard CES [02:38:22] the buyers prices for a good [02:38:24] from country N [02:38:26] sold in country I [02:38:27] are driven by the productivity [02:38:29] which is a summary of the firm's [02:38:31] productivity and country's productivity [02:38:33] and then we have the two policy instruments [02:38:35] which are the import, TRF and the export tax [02:38:38] and then we have the input cost [02:38:40] so what is the input cost again [02:38:41] it has the bilateral trade cost [02:38:45] which is the iceberg [02:38:47] and then naturally the wage rate [02:38:49] and the price of inputs [02:38:50] because labor and intermediate inputs [02:38:52] are both used for production [02:38:54] the aggregate trade share is very standard looking [02:38:57] coming from a CS demand system [02:38:59] defined in lambda and I [02:39:01] meaning this is how much [02:39:02] country N sells to country I [02:39:05] again just to make sure that everybody remembers [02:39:08] the notation because these are objects [02:39:10] that are going to show up [02:39:12] this is the import TRF [02:39:13] and this is the export tax [02:39:15] so I'm going to skip through this [02:39:18] because it's very standard [02:39:20] so nothing interesting here [02:39:22] aggregate equilibrium [02:39:24] general equilibrium conditions apply [02:39:26] we have the aggregate price index [02:39:27] labor market clearing [02:39:28] and I just want everyone to make sure [02:39:31] that they understand that we're looking at balanced trade [02:39:33] so no kind of deficits [02:39:35] no nothing interesting there [02:39:39] alright now in order to lay out the results [02:39:42] two objects are useful for me [02:39:46] so the first one is the export supply elasticity [02:39:49] epsilon FHS [02:39:51] the export supply elasticity in this literature [02:39:53] measures the degree of the importers [02:39:56] so the buyers market power [02:39:58] it'll be an object that will be showing up a lot [02:40:01] when we look at our optimal tariffs [02:40:03] but I want to also bring in the import demand elasticity [02:40:07] so the import demand elasticity on the other hand [02:40:09] I think of as the mirror image [02:40:11] of the export supply elasticity [02:40:13] and this object measures the degree of the exporter's market power [02:40:17] which in traditional trade theory does not show up [02:40:20] in the optimal tariff [02:40:22] but as we'll see in our model with geopolitical risk [02:40:25] it... [02:40:27] so I'm going to lay out the problem as follows [02:40:30] I'm first going to show you what the home countries [02:40:33] the United States will do [02:40:35] if the foreign country is completely passive [02:40:37] and then we're going to bring in the interesting point [02:40:40] where the foreign country may or may not act strategically [02:40:43] and solve everything by backward induction [02:40:46] okay so starting off with the simple [02:40:49] or think of it as the benchmark case [02:40:51] where the foreign country is passive [02:40:53] the only thing that the home country is doing [02:40:55] is to maximize its own welfare [02:40:57] which is just the real income [02:41:00] so it's a... [02:41:03] income over the price index [02:41:06] the first proposition that comes out of this [02:41:08] is something that should look familiar to [02:41:10] to people who have studied this literature [02:41:12] which is that the optimal tariff [02:41:14] is just the inverse of the export supply elasticity [02:41:18] the only thing that's perhaps different [02:41:20] from what you've seen before is that [02:41:22] here the export supply elasticity [02:41:24] is reflecting this extra term [02:41:26] which comes out due to the input output [02:41:29] structure of the model [02:41:31] okay so buying... [02:41:33] sort of buying the fact that we're [02:41:36] we're buying foreign goods [02:41:39] in order to produce our home good [02:41:41] that means that we're [02:41:44] we're going to... [02:41:45] when we impose a tariff [02:41:47] we're going to effectively raise the price [02:41:49] of the foreign good relative to the home even more [02:41:52] than we would otherwise [02:41:54] so that [02:41:56] this formula is very familiar [02:41:58] in terms of intuition and it basically [02:42:00] captures the terms of trade effect [02:42:03] so what are the terms of trade effect [02:42:04] seeing in these models [02:42:06] is that because there's just a single factor of production [02:42:08] which is labor [02:42:10] the only way that we can manipulate these terms of trade [02:42:12] or the only way that we can maximize our utility [02:42:14] is to just [02:42:16] change the relative price of our [02:42:18] one factor which is labor [02:42:20] or the wage rate [02:42:21] compared to the other country [02:42:23] okay [02:42:24] now [02:42:25] let's bring in the [02:42:27] strategic part [02:42:28] so the way that we set this up [02:42:30] and you know we're kind of open to [02:42:32] to comments here but this is how [02:42:34] how we started thinking about it [02:42:36] so we're going to set up as a Bayesian [02:42:38] Stackelberg game [02:42:40] so I'm going to guide you through two [02:42:42] steps [02:42:44] that we're going to use in order to set up the model [02:42:46] and then solve everything with a backward induction [02:42:48] so let's kind of see if it makes sense [02:42:50] so we're going to have nature [02:42:52] draw the type of [02:42:54] the foreign country that's going to be captured [02:42:56] by a parameter eta which is drawn from [02:42:58] a probability distribution [02:43:00] which is drawn from a distribution [02:43:02] G and this eta [02:43:04] basically represents the non-pecuniary [02:43:06] cost of non-cooperation [02:43:08] so think of it as like a penalty [02:43:10] for being bad [02:43:12] right for not [02:43:14] cooperating and then [02:43:16] the timing is as follows [02:43:18] so in the first period [02:43:20] the home country will commit [02:43:22] to an import tariff [02:43:24] denoted by T [02:43:26] of course it observes the distribution [02:43:28] of this object eta [02:43:30] but not the realization itself [02:43:32] and then in the second [02:43:34] period after this [02:43:36] realization is [02:43:38] takes place so after the type [02:43:40] of the foreign countries revealed [02:43:42] observing the tariff and [02:43:44] this type realization [02:43:46] the foreign country will pick a regime [02:43:48] this regime can be cooperative or non-cooperative [02:43:50] which basically [02:43:52] comes down to the following if it's cooperative [02:43:54] then the foreign country chooses [02:43:56] to not introduce [02:43:58] an export tax so the optimal [02:44:00] export tax is zero if [02:44:02] it's non-cooperative then [02:44:04] it chooses an export tax that [02:44:06] maximizes its own welfare [02:44:08] and let's set up the problem of the foreign [02:44:10] country so that we see what that [02:44:12] looks like so what is the foreign [02:44:14] country trying to show so [02:44:16] far so they're simply trying to maximize [02:44:18] their own real income [02:44:20] net off that [02:44:22] non-pecuniary cost of non-cooperation [02:44:24] which only occurs [02:44:26] in the case that they're non-cooperating or they're bad [02:44:28] so then the foreign country [02:44:30] is effectively playing a best [02:44:32] response which is [02:44:34] a bundle of [02:44:36] the export tax itself [02:44:38] and the probability of [02:44:40] non-cooperation or the probability [02:44:42] of being bad which is effectively [02:44:44] the probability that the [02:44:46] realization of this non-pecuniary [02:44:48] cost is such that [02:44:50] it is lower than [02:44:52] the net [02:44:54] welfare [02:44:56] that the foreign country will earn by introducing [02:44:58] an export tax [02:45:00] versus not [02:45:02] okay so [02:45:04] I'm going to show you some properties [02:45:06] of this solution in a second [02:45:08] and I just want to make one definition [02:45:10] before I do so so [02:45:12] lambda FH which is the amount [02:45:14] of goods or the share [02:45:16] of goods that we the home country [02:45:18] are purchasing from abroad is going to be [02:45:20] a key element here [02:45:22] that will [02:45:24] show up in the solution of the foreign country's problem [02:45:26] and therefore in the home country's problem [02:45:28] so let me call that object lambda [02:45:30] so think of lambda then as being [02:45:32] the import share 1 minus lambda [02:45:34] must therefore be the domestic [02:45:36] the share of expenditures [02:45:38] okay [02:45:40] so what we show in the paper is that [02:45:42] if the foreign country ends up [02:45:44] taking an adversarial stand [02:45:46] then it'll set [02:45:48] an optimal export tax [02:45:50] that is given by this expression [02:45:52] here which of course [02:45:54] if you go back to [02:45:56] our definition of the export supply [02:45:58] elasticity it exactly mirrors [02:46:00] that expression [02:46:02] so why is that so that's the familiar learner [02:46:04] symmetry result which is that [02:46:06] you're setting an export tax that is equivalent [02:46:08] to an import [02:46:10] in this case the resulting welfare [02:46:12] gain for the foreign country [02:46:14] is just the increase [02:46:16] in welfare that they get from [02:46:18] introducing the steriff over [02:46:20] not and this export tax [02:46:22] will be applied as we said earlier [02:46:24] if that difference is bigger [02:46:26] than the cost of non-cooperation [02:46:28] which itself is [02:46:30] parametrized by the function [02:46:32] G now [02:46:34] we show two things [02:46:36] which is that both the export [02:46:38] tax itself [02:46:40] and the probability that [02:46:42] a country will end up being [02:46:44] bad are increasing [02:46:46] in the import share [02:46:48] of the United States [02:46:50] from China so the idea here is that [02:46:52] the more that the US is [02:46:54] buying from China [02:46:56] the higher is the optimal [02:46:58] export tax that China is going to set [02:47:00] and the higher is the probability [02:47:02] that China will do so [02:47:04] so that's true under the following conditions [02:47:06] first under [02:47:08] parametric assumptions that make a lot of sense [02:47:10] empirically [02:47:12] we show that greater dependence will strengthen [02:47:14] the foreign market power [02:47:16] in the bad state of the world [02:47:18] or the adversarial state raising the [02:47:20] optimal export wedge [02:47:22] and then we also show that as long as [02:47:24] the foreign countries welfare [02:47:26] is like super modular export taxation [02:47:28] gains increase with lambda and therefore [02:47:30] the probability of an export restriction [02:47:32] will also increase in lambda [02:47:34] putting those two things together [02:47:36] allows us to solve then the problem [02:47:38] using the certainty [02:47:40] equivalence [02:47:42] principle so let me define [02:47:44] the certainty equivalent foreign wedge [02:47:46] tau bar L to be [02:47:48] the deterministic export tax [02:47:50] that leaves the home country [02:47:52] as well off as it would [02:47:54] if we were to solve that [02:47:56] problem in the stochastic [02:47:58] space so remember this is a stochastic [02:48:00] problem because we don't know whether the foreign country [02:48:02] is going to be good or bad [02:48:04] so we'll be solving this [02:48:06] using the certainty equivalent which is basically [02:48:08] solving it in expectation so [02:48:10] the certainty equivalent we show [02:48:12] can be approximated by the following quadratic [02:48:14] function and when we go to the [02:48:16] quantitative part of the model [02:48:18] we'll be parameterizing it and putting some numbers on it [02:48:20] okay so this then allows [02:48:22] a foreign [02:48:24] policy problem of a country that [02:48:26] does take an adversarial stand [02:48:28] to basically be to maximize [02:48:32] it allows the home country's [02:48:34] problem in this [02:48:36] case of a bad foreign country [02:48:38] to simply be to maximize [02:48:40] their own welfare subject to the [02:48:42] same equilibrium constraints that we saw earlier [02:48:44] and the additional constraint [02:48:46] that the foreign country's export [02:48:48] tax is equal to this [02:48:50] equivalent export tax tau bar that we just [02:48:52] defined now [02:48:54] remember that I showed you [02:48:56] that it's true that [02:48:58] both the probability [02:49:00] and the export tax itself are increasing [02:49:02] in the import dependence therefore the certainty equivalent [02:49:04] is also increasing in the [02:49:06] import dependence which [02:49:08] that means that due to this [02:49:10] increase it's going to be the case that [02:49:12] lowering [02:49:14] the exposure of the home country to the [02:49:16] foreign country will reduce the foreign [02:49:18] export wedge and therefore we will want [02:49:20] a TRF to accomplish [02:49:22] that task for us so that [02:49:24] gets us to the optimal macro [02:49:26] TRF under geopolitical risk [02:49:28] which is made up of two components so [02:49:30] the first component is the standard one [02:49:32] that we already saw is the terms of trade component [02:49:34] and the second one is the geopolitical [02:49:36] risk correction [02:49:38] which reflects [02:49:40] two things so first it reflects [02:49:42] the elasticity of demand [02:49:44] which as I defined earlier [02:49:46] is reflecting [02:49:50] the market power of the [02:49:52] buyer in this case [02:49:54] and the second one is [02:49:56] the change in the [02:49:58] or the growth [02:50:00] in the certainty equivalent [02:50:02] so this second [02:50:04] term is the new term relative [02:50:06] to standard theory and [02:50:08] specifically as you can see it depends [02:50:10] on the [02:50:12] exposure of the home country to the foreign [02:50:14] country unlike the terms [02:50:16] of trade component now what's [02:50:18] interesting about this result is that when we [02:50:20] take it to multiple sectors [02:50:24] everything goes through very easy to set up [02:50:26] and solve [02:50:28] the results [02:50:30] to me at least when we first started [02:50:32] working on this was surprising so [02:50:34] what is surprising about it so the first [02:50:36] part which is the terms of trade part [02:50:38] is uniform so it's not [02:50:40] sector specific why because [02:50:42] we're just trying to manipulate the terms [02:50:44] of trade through the wage effect [02:50:46] but it is the second part here [02:50:48] the risk correction part that is sector [02:50:50] specific and [02:50:52] looks exactly like what we saw [02:50:54] earlier except that [02:50:56] it has sector [02:50:58] variable so it's reflecting both [02:51:00] sectoral [02:51:02] of trade and this is something that we've seen in [02:51:04] Mateo's work right it reflects [02:51:06] the sectoral demand [02:51:08] elasticity and this [02:51:10] correction which is [02:51:12] due to the export tax [02:51:14] so what we're going to do [02:51:16] next is to try to take this to the [02:51:18] data and see [02:51:20] first of all what types [02:51:22] or what the optimal [02:51:24] tariffs should look like for the United States [02:51:26] we're going to do it for Europe [02:51:28] as well and then second [02:51:30] we're going to take this to the data and see [02:51:32] what the tariffs have actually looked like [02:51:34] for the United States in 2018 and see [02:51:36] whether they match up with [02:51:38] the theory that we propose [02:51:40] right in order to take [02:51:42] the model to the data a lot [02:51:44] of the data is standard we just need [02:51:46] import shares [02:51:48] GDP expenditure but we do need [02:51:50] one parameter which [02:51:52] is a parameter that relates [02:51:54] to the geopolitical risk [02:51:56] and that's the parameter zeta so what is [02:51:58] zeta so it's the geopolitical [02:52:00] risk elasticity which is basically [02:52:02] measuring how fast the probability [02:52:04] of the foreign country restrict trade [02:52:06] or puts an export tax rises [02:52:08] with the home country's import [02:52:10] dependence or import share [02:52:12] where is that coming from so it's coming [02:52:14] from straight from our assumption [02:52:16] on what [02:52:18] the [02:52:20] probability is [02:52:22] or what the distributional [02:52:24] assumptions that we put on the [02:52:26] on the cost data so just [02:52:28] inverting that gives us [02:52:30] a hazard function pi which is [02:52:32] a power function of [02:52:34] lambda and zeta is [02:52:36] just measuring that elasticity [02:52:38] or that coefficient in the power [02:52:40] function so naturally when you log it [02:52:42] it gives us a very easy [02:52:44] estimating equation where we can [02:52:46] estimate that [02:52:48] zeta by regressing [02:52:50] the log of the hazard [02:52:52] on the import share [02:52:54] we're going to take the log import share [02:52:56] here because we want to [02:52:58] say what type of [02:53:00] restriction has China imposed [02:53:02] given the import share [02:53:04] that we observed the previous year [02:53:06] from the United States [02:53:08] United States import share from [02:53:10] China in the previous year [02:53:12] and we're going to run this [02:53:14] simple linear regression [02:53:16] where we're going to have [02:53:18] importer sector fixed effects [02:53:20] and time fixed effects [02:53:22] and the hazard data is going to [02:53:24] come from the GTA and always see [02:53:26] the export restriction database [02:53:28] so given the level of aggregation [02:53:30] that's available in that database [02:53:32] we're going to aggregate the hazards [02:53:34] to the [02:53:36] we're going to aggregate the [02:53:38] hazard to the [02:53:40] to the HS6 [02:53:42] digit level data by using [02:53:44] the export share [02:53:46] for each good within that category [02:53:48] and running this regression [02:53:50] we're going to do several [02:53:52] specifications I just don't have time [02:53:54] to show you all of them [02:53:56] we're going to do OLS [02:53:58] we're going to do [02:54:00] PPL [02:54:02] PPL [02:54:04] PPL [02:54:06] PPL [02:54:08] sorry [02:54:10] and [02:54:12] we're going to do it for [02:54:14] the US and for Europe [02:54:16] so the number that comes out of it [02:54:18] in the baseline year of 2018 [02:54:20] for a 16 aggregated sector [02:54:22] for the United States and for [02:54:24] EU at 0.29 and 0.23 [02:54:26] so we're going to combine these [02:54:28] numbers together with import shares [02:54:30] for the United States [02:54:32] vis-a-vis the rest of the world [02:54:34] the European Union vis-a-vis the rest of the world [02:54:36] and a total expenditure [02:54:38] in GDP together with [02:54:40] the United States [02:54:42] to the US and the EU [02:54:44] and the US [02:54:46] to trade elasticity for now [02:54:48] across sectors of four [02:54:50] and value added shares for [02:54:52] intermediate inputs [02:54:54] that we compute from the [02:54:56] input-output database for both [02:54:58] the US and the EU [02:55:00] the result is as follows [02:55:02] the optimal [02:55:04] tier for the United States and the EU [02:55:06] without the geopolitical risk [02:55:08] adds another 5 percentage [02:55:10] points so about 16% [02:55:12] of the optimal tier [02:55:14] so it's not trivial [02:55:16] but what's interesting about it is [02:55:18] when we take it across sectors [02:55:20] so doing the exact same exercise [02:55:22] and calibrating the model [02:55:24] to generate optimal tier [02:55:26] at different sectors [02:55:28] what I'm plotting here is [02:55:30] the result for the United States and the EU [02:55:32] as a function of the foreign market share [02:55:34] in that sector [02:55:36] all that the optimal tier [02:55:38] of [02:55:40] excluding the geopolitical part [02:55:42] which is just the terms of trade portion [02:55:44] is roughly 26% [02:55:46] and it's the same across sectors [02:55:48] and then the optimal [02:55:50] tier of including this geopolitical risk [02:55:52] ranges from [02:55:54] about 27% [02:55:56] to all the way [02:55:58] 36 or 37% [02:56:00] for some sectors so this is about [02:56:02] 50% extra [02:56:04] kick [02:56:06] that comes out of the geopolitical [02:56:08] portion so it's [02:56:10] quantitatively it's quite large [02:56:12] and more interestingly it's really [02:56:14] accounting for this [02:56:16] cross sectoral variation [02:56:18] so then the big question is [02:56:20] okay you have this model [02:56:22] this is how you think about the world [02:56:24] does it make sense? Is this what we see [02:56:26] in the world? [02:56:28] So in order to [02:56:30] test the predictions of our model [02:56:32] and to be the data it's important [02:56:34] to recognize that there are alternative [02:56:36] theories to what optimal [02:56:38] tariffs look like so the two leading [02:56:40] alternatives here being [02:56:42] the first original [02:56:44] which is the terms of trade [02:56:46] theory that basically says that you're [02:56:48] setting tariffs that are exactly [02:56:50] equal to the inverse of the [02:56:52] export supply elasticity therefore [02:56:54] tariffs track importer market power [02:56:56] Brutalima and Weinstein tested this [02:56:58] in 2008 so we're [02:57:00] looking at the [02:57:02] market for the [02:57:04] export supply elasticity at the sectoral [02:57:06] level the day estimated in that paper [02:57:08] and then the second one is of course [02:57:10] protection for sale where [02:57:12] the theoretical model that comes out [02:57:14] of Grossman and Hulman was tested [02:57:16] or taken to the data by Giovanni [02:57:18] Maggi and Penny Goldberg so [02:57:20] this test would say that [02:57:22] protection should reflect [02:57:24] lobbying so therefore for [02:57:26] sectors that are organized the [02:57:28] penetration of [02:57:30] the sector [02:57:32] or the country relative to the other [02:57:34] country so that is effectively [02:57:36] going to say that the tariff should be [02:57:38] related to the import demand [02:57:40] elasticity once again it's something [02:57:42] that we take estimated from [02:57:44] the data and finally our model [02:57:46] will tell you that [02:57:48] you have the export supply elasticity [02:57:50] you have the export demand elasticity [02:57:52] but it should really be the growth [02:57:54] in this dependence or the growth [02:57:56] in the tariff and the [02:57:58] distribution of the tariff [02:58:00] and the share from China that [02:58:02] should drive the optimal [02:58:04] tier so therefore we run the [02:58:06] following regression we're going [02:58:08] to take the change or the growth [02:58:10] in tariffs during [02:58:12] the 2018 episode specifically [02:58:14] we're going to start by focusing [02:58:16] on the section 301 tariffs which [02:58:18] are the China specific tariffs so [02:58:20] just to make sure that everybody [02:58:22] is on the same page here during [02:58:24] the different stages of tariff [02:58:26] increases right there was the [02:58:28] first two stages [02:58:30] were not China specific [02:58:32] so they were kind of the same for everybody [02:58:34] so those are your section 232 to [02:58:36] 01 tariffs we're going to add them back [02:58:38] in a second then you have the third [02:58:40] trench which is the China specific one [02:58:42] and then there was the fourth one that was [02:58:44] kind of threatened but then it was pulled [02:58:46] back okay so we're really focusing on that [02:58:48] third one which is 301 okay [02:58:50] so we're looking at the change in [02:58:52] those tariffs from before to [02:58:54] after they were implemented and we're [02:58:56] regressing that on the growth [02:58:58] in the import [02:59:00] share from China [02:59:02] in the five years prior [02:59:04] to 2018 controlling [02:59:06] for of course the export supply [02:59:08] elasticity and the import demand [02:59:10] elasticity which would be predicted [02:59:12] by the alternative theories [02:59:14] so the baseline [02:59:16] here is the [02:59:18] result that where we're looking at [02:59:20] the change in China's [02:59:22] import share out of total [02:59:24] imports for the United States [02:59:26] and in fact [02:59:28] that's positive significant [02:59:30] which is what our theory would predict [02:59:32] interestingly the exports applied [02:59:34] import demand elasticity are not being [02:59:36] picked up at all so [02:59:38] the estimates are very small [02:59:40] and insignificant we're doing this [02:59:42] both at the Hs10 level [02:59:44] and Hs4 digit level a caveat [02:59:46] here I should make sure that everybody understands [02:59:48] is that for [02:59:50] export supply elasticity we only have [02:59:52] observations at the Hs4 digit level [02:59:54] that's what [02:59:56] broad and we now have estimated so then [02:59:58] when we go to Hs10 digit level which is [03:00:00] where we observe the tariffs we basically [03:00:02] have to assume that it's that same [03:00:04] the number applies to all the categories [03:00:06] within that bracket [03:00:08] and then [03:00:10] on top of that we run [03:00:12] I'm not sure that placebo is the best [03:00:14] word for it here we're still going back [03:00:16] to the point where we're looking at [03:00:18] tariffs changes on the same goods [03:00:20] but from non-China suppliers [03:00:22] because in principle [03:00:24] here how much or what our [03:00:26] import shares from China should not be [03:00:28] telling us how we're changing tariffs [03:00:30] on other countries [03:00:32] in this episode and indeed [03:00:34] we're seeing that those are not [03:00:36] significant so then in PNLB [03:00:38] we do the exact same thing but we're [03:00:40] now on top of [03:00:42] looking at the change in the [03:00:44] total import share of China out of total [03:00:46] imports we're also scaling by the total [03:00:48] import penetration in order to get the [03:00:50] actual scale effect [03:00:52] and indeed the coefficients still line up [03:00:54] and then finally what we do is to [03:00:56] add back those [03:00:58] trench one and trench two tariffs [03:01:00] which are the section 232 and section [03:01:02] 201 tariffs so remember these are not [03:01:04] China specific so what we're seeing [03:01:06] here is that relative to the baseline [03:01:08] the coefficient falls [03:01:10] and becomes weaker and indeed that is [03:01:12] what we expect because these are not [03:01:14] China specific tariffs so [03:01:16] they should not be [03:01:18] correlated with how much we're [03:01:20] buying from China [03:01:22] so [03:01:24] overall we see these results [03:01:26] are showing support for a [03:01:28] theory that builds on strategic [03:01:30] decoupling [03:01:32] due to supply or risk [03:01:34] and finally we [03:01:36] do one more exercise where [03:01:38] we use USTR communications [03:01:40] during the [03:01:42] 2010 to 2025 period so here [03:01:44] we're taking every USTR [03:01:46] release and we are [03:01:48] classifying it into six economic [03:01:50] frames where the ones that are [03:01:52] specific to our mechanism are geopolitical [03:01:54] or centriturgical exposure and [03:01:56] what we're seeing is that pre-2017 [03:01:58] the [03:02:00] mentioning of these [03:02:02] terms was [03:02:04] relatively low [03:02:06] so 6.9% [03:02:08] of the USTR [03:02:10] was in 2018 that jumped [03:02:12] by [03:02:14] roughly four or five fold and [03:02:16] indeed that jump is roughly [03:02:18] concentrated in the March to [03:02:20] July 2018 window [03:02:22] which is the window between the [03:02:24] section 301 announcement and the [03:02:26] first wave of section 301 [03:02:28] tariff list. [03:02:30] So we find this [03:02:32] rhetoric to be moving with [03:02:34] strategic risk together as [03:02:36] we're going to [03:02:38] look at the next [03:02:40] slide. [03:02:42] So to conclude [03:02:44] of what we're showing in this paper [03:02:46] is that the optimal tariff is made [03:02:48] up of two components the standard [03:02:50] component we're used to seeing [03:02:52] which is the terms of trade [03:02:54] manipulation and a strategic risk [03:02:56] correction term. This risk [03:02:58] correction term is larger in [03:03:00] sectors with higher import [03:03:02] dependence which seems to line [03:03:04] up with the [03:03:06] standard [03:03:08] and the discussion is ## Discussion (03:03:15 – 03:18:16) [03:03:16] Matilda Bombardini. [03:03:27] So this is working and this is [03:03:29] the other. [03:03:31] Thank you so much for inviting me [03:03:33] to discuss this paper. I had a lot [03:03:35] of fun thinking about it and [03:03:37] hear my thoughts and I [03:03:39] apologize if they're confused [03:03:41] still. [03:03:43] So I love papers that start with [03:03:45] the sort of fact that [03:03:47] I'm going to go into the [03:03:49] next slide. [03:03:51] So I'm going to go into the [03:03:53] next series and we'll see [03:03:55] I'll tell you [03:03:57] whether that's the [03:03:59] conclusion I reached but [03:04:01] that has the [03:04:03] China import share here [03:04:05] and the tariffs that the [03:04:07] U.S. is imposed on China so [03:04:09] meaning that you impose [03:04:11] higher tariffs where the [03:04:13] China import share by the way [03:04:15] is divided by total imports. [03:04:17] It's not imports from China [03:04:19] divided by absorption. [03:04:21] So that's something that I [03:04:23] will come back to. [03:04:25] So and you know I kind of [03:04:27] described that yes if you [03:04:29] think about import [03:04:31] absorption it's true that [03:04:33] Grossman and Helpen sort of [03:04:35] predict sort of the [03:04:37] inverse because it's Ramsey [03:04:39] taxation so it's [03:04:41] inverse and so on and the [03:04:43] other thing is that [03:04:45] the U.S. [03:04:47] is the most [03:04:49] important part of the [03:04:51] process and I'll [03:04:53] predict that. [03:04:55] So just a very quick summary [03:04:57] of so that sort of [03:04:59] motivates their that together [03:05:01] with the increase in export [03:05:03] restrictions motivates this [03:05:05] sort of the model where as [03:05:07] a reminder you have two [03:05:09] countries it's an Armington [03:05:11] and Stackelberg [03:05:13] it's called sort of [03:05:15] where instead of doing [03:05:17] simultaneous moves [03:05:19] of home and foreign home moves [03:05:21] first and sets the tariffs [03:05:23] and then nature draws this [03:05:25] non-cooperation cost and [03:05:27] foreign either cooperates or [03:05:29] doesn't based on that [03:05:31] then they derive this sort of [03:05:33] hazard and so then [03:05:35] both the hazard and so the [03:05:37] main result is that both [03:05:39] the tax that foreign imposes [03:05:41] and the probability of imposing [03:05:43] this export tax sort of are [03:05:45] increasing in what? [03:05:47] In homes dependence from [03:05:49] foreign and so they say this is [03:05:51] a new sort of term that's [03:05:53] sort of they call geopolitical [03:05:55] risk that adds to the [03:05:57] standard terms of trade sort [03:05:59] of mill bicker [03:06:01] like formula and [03:06:03] then what the nice [03:06:05] thing is that they have a lot [03:06:07] of data that they bring to [03:06:09] that which is sort of the [03:06:11] estimate this hazard function [03:06:13] the probability of foreign's [03:06:15] export restrictions increasing with [03:06:17] home dependence and then they do a [03:06:19] quantification and [03:06:21] importantly they go back to match [03:06:23] that pattern that we saw at the [03:06:25] beginning sort of showing that this [03:06:27] model predicts that [03:06:29] you sort of impose [03:06:31] higher tariffs on [03:06:33] sectors that on which you're [03:06:35] going to be able to [03:06:37] get the export tax are going to be [03:06:39] higher. Okay, so that's the paper. [03:06:41] All right. [03:06:43] So I'm going to make two points [03:06:45] only. I mean I'm going to [03:06:47] sprinkle a sort of smaller ones along [03:06:49] the way but the first is [03:06:51] really the nature of this [03:06:53] geopolitical risk correction. [03:06:55] I'm going to ask the following [03:06:57] question which I didn't have time [03:06:59] to ask them because I didn't [03:07:01] prepare in time. [03:07:03] So I'm going to go back to [03:07:05] the paper and I'm going to [03:07:07] go back to the paper with just [03:07:09] tackle board game. [03:07:11] No risk, no intermediates. [03:07:13] Okay, so I'm going to sort of [03:07:15] solve that model. [03:07:17] And then I'm going to kind of [03:07:19] push and kind of suggest that [03:07:21] we should, you know, it would be [03:07:23] useful to have all that sort of [03:07:25] imports from China versus total [03:07:27] import dependence. You should have [03:07:29] more than one country because then [03:07:31] you're going to have a lot of [03:07:33] more than one country. [03:07:35] So I'm going to go back to the [03:07:37] paper first. [03:07:39] Okay, so I was really trying to [03:07:41] figure out what was going on here, [03:07:43] sort of see how far, you know, [03:07:45] what do you really need to get [03:07:47] that result of this dependence. [03:07:49] So I sort of [03:07:51] stripped down, I removed [03:07:53] intermediates, no roundabout, [03:07:55] no hazard, forget about the [03:07:57] extensive margin of whether you [03:07:59] want to get that result. [03:08:01] That's the same thing. [03:08:03] And I'm going to preserve the [03:08:05] timing, which is the key here, [03:08:07] which is that home sets a [03:08:09] tariff, then foreign sets a [03:08:11] tariff. [03:08:13] What would this deliver? [03:08:15] I think it would deliver, sort of [03:08:17] a similar thing, meaning that [03:08:19] when you have [03:08:21] foreign, we'll still set [03:08:23] a tariff that's [03:08:25] the inverse of the export [03:08:27] and it's going to depend on the [03:08:29] sort of import dependence of [03:08:31] home. [03:08:33] So you still get that result that [03:08:35] the foreign tax, tariff, [03:08:37] whatever you want to call it, is [03:08:39] increasing in home dependence, same [03:08:41] as the full model. [03:08:43] Why? So this kind of reveals that [03:08:45] you don't need intermediates, [03:08:47] you just need consumers, [03:08:49] consumers are optimistic by [03:08:51] definition, so they don't internalize [03:08:53] all this functionality, which is why [03:08:55] we're having tariffs in the first place. [03:08:57] That's my interpretation, maybe I'm wrong. [03:08:59] So that's [03:09:01] stage one. [03:09:03] So what about [03:09:05] the second mover? [03:09:07] And then home will [03:09:09] also try [03:09:11] to anticipate this by setting [03:09:13] a higher tariff on [03:09:15] sectors in which it has a higher dependence [03:09:17] because it anticipates that the [03:09:19] tariff is going to be higher on those sectors. [03:09:21] And so [03:09:23] again, [03:09:25] you know, the term, [03:09:27] so I've just redirived in a very simple [03:09:29] version, I mean it's just a [03:09:31] collapsed version of that model, [03:09:33] essentially a sort of a [03:09:35] mille-bicard type terms of trade part [03:09:37] and then what I'm really calling [03:09:39] a first mover premium, [03:09:41] which [03:09:43] I mean it's not geopolitical risk, [03:09:45] but it's sort of a, it's just due to [03:09:47] the fact that instead of a Nash [03:09:49] you have, you get to move first. [03:09:51] And that can anticipate, [03:09:53] you know, can undo some of the [03:09:55] damage that foreign will do later. [03:09:57] And so I think [03:09:59] but again, I didn't have time [03:10:01] to ask you, I think this still delivers [03:10:03] figure two, sort of the [03:10:05] motivating figure. [03:10:07] And so I asked Claude [03:10:09] to draw it for me. [03:10:11] And so I just [03:10:13] set up this model and so it did deliver it. [03:10:15] So, and so the similar [03:10:17] slope to the model and [03:10:19] so I think you can fit that. [03:10:21] So I guess my question is [03:10:23] can we just call this Stackelberg premium? [03:10:25] So, you know, can we call this [03:10:27] sort of the Stackelberg first mover premium? [03:10:29] Which is absent under Nash. [03:10:31] And then, so then [03:10:33] it really is about timing [03:10:35] which we know matters a great deal [03:10:37] because it's all about commitment [03:10:39] and all that and so we know it matters. [03:10:41] So it's just, you know, is this [03:10:43] the right assumption? You know, is this [03:10:45] the right assumption? You know, if you think [03:10:47] about the tariffs versus [03:10:49] the first rare earths [03:10:51] export restrictions, maybe, [03:10:53] you know, I'm not, you know, [03:10:55] but in any case, you know, [03:10:57] playing with the timing and being sort of [03:10:59] clear about what delivers what [03:11:01] is great. [03:11:03] This slide is not to the fact [03:11:05] that the organizers are who they are. [03:11:07] And so, [03:11:09] and so [03:11:11] this, [03:11:13] so here [03:11:15] I think [03:11:17] the second paper, the one that [03:11:19] Matteo [03:11:21] taught me to call the outside option paper [03:11:23] which I think everybody else [03:11:25] calls it concussion and something else [03:11:27] and fragmentation or something like that, yeah. [03:11:29] But I think of it, it's useful [03:11:31] to think about it the outside option and the first [03:11:33] the chronometrical one is the inside option one. [03:11:35] That's the, yes. [03:11:37] So, [03:11:39] so, and so [03:11:41] I think that would deliver a similar thing because [03:11:43] in their paper, [03:11:45] remember the outside option, so [03:11:47] what's the story there is that, let's say, [03:11:49] the non-Hegemon, [03:11:51] I mean here you would have to decide who's the [03:11:53] Hegemon here, but [03:11:55] let's see that, let's say that [03:11:57] US moves first and then China moves [03:11:59] second. [03:12:01] China there would have the, let's say China [03:12:03] now is A Hegemon. [03:12:05] It could cut, you know, [03:12:07] if you make yourself dependent on China, [03:12:09] then China can withhold their inputs [03:12:11] and extract [03:12:13] various wedges [03:12:15] from you. [03:12:17] And so then they would [03:12:19] then you would try to reduce that, exactly. [03:12:21] So if you then, you make yourself less dependent [03:12:23] on China because you want less to be [03:12:25] extracted later on. [03:12:27] Then there's a whole issue whether China can [03:12:29] commit to not extract it later on, which is [03:12:31] something that you guys have to deal with, so I'll just leave that to you, but [03:12:33] I think in that [03:12:35] case you would still get that the tariff [03:12:37] would be increasing in impredependence [03:12:39] so that kind of [03:12:41] goes back to the question of whether [03:12:43] the existing theories [03:12:45] explain Figure 2, which [03:12:47] I think it would, but I didn't have [03:12:49] time to talk to both these guys and so [03:12:51] okay. [03:12:55] Oh, and then, how am I doing [03:12:57] the time? [03:12:59] Oh, I may not even [03:13:01] use them all. [03:13:03] So one thing [03:13:05] actually, let me go back. [03:13:07] So what I like about [03:13:09] the fact that [03:13:11] relative to the CMS [03:13:13] paper, that you do [03:13:15] have retaliation, you have [03:13:17] on-path [03:13:19] not punishment, but you have [03:13:21] on-path export [03:13:23] taxes, which [03:13:25] in your [03:13:27] paper, it's [03:13:29] a threat which [03:13:31] shouldn't materialize. [03:13:33] And so that's a nice [03:13:35] thing that [03:13:37] in INA [03:13:39] and HAVMAD's paper, you do [03:13:41] have the export tax [03:13:43] gets placed on [03:13:45] exports. And so, and that's [03:13:47] why they can estimate all this hazard [03:13:49] function and so on. [03:13:51] And so, I mean [03:13:53] also the difference is that [03:13:55] here it's a well for maximizing [03:13:57] choice to input the tax. [03:13:59] So you do it because you're [03:14:01] just exploding in terms of trade. [03:14:03] Whereas, you know, when it's [03:14:05] a threat to cut off inputs, [03:14:07] that's a cost for the hegemon. [03:14:09] So you don't want to do it. [03:14:11] And you don't do it because, you know, [03:14:13] you'll hold it as a threat. Whereas here, you actually do it [03:14:15] on-path. Okay, so I guess [03:14:17] then you can sort of try to [03:14:19] distinguish because figure [03:14:21] two is delivered by, I think, [03:14:23] both sort of [03:14:25] theories, but [03:14:27] this on-path punishment or [03:14:29] on-path sort of [03:14:31] aggression of some sort [03:14:33] is realized. [03:14:35] So maybe one could try to distinguish [03:14:37] the two theories in that way. [03:14:39] Okay, so [03:14:41] I was [03:14:43] so I [03:14:45] maybe [03:14:47] I could use more discussion of [03:14:49] sort of the extensive versus intensive [03:14:51] dimensions of this threats of like [03:14:53] because they both [03:14:55] both the probability of non-compiration [03:14:57] and the size of the export tax are both [03:14:59] independent. [03:15:01] And so I would [03:15:03] eventually the problem is described in terms of [03:15:05] certainty equivalent. So I wasn't so clear. [03:15:07] It'd be great [03:15:09] to know [03:15:11] in normative terms whether these two have [03:15:13] you know, sort of conceptually different roles. [03:15:15] Of course in positive [03:15:17] terms it's important to know, you know, [03:15:19] we want to explain the world and so we want to explain [03:15:21] both the probability of imposing something [03:15:23] and how much you're going to impose it. [03:15:25] But I guess [03:15:27] it's important to know that [03:15:29] in normative terms [03:15:31] it's important to know that [03:15:33] normatively it'd be nice to have a discussion. [03:15:35] And then I had a quibble about calling [03:15:37] x equal to zero then the cooperative regime. [03:15:39] I guess [03:15:41] the word cooperative always has a flavor of [03:15:43] efficient or joint utility maximizing [03:15:45] which x equal to zero is not given that you put [03:15:47] a tariff but that's just [03:15:49] quibble. [03:15:51] And then finally I'd like to sort of comment on this [03:15:53] sort of the [03:15:55] question that you have. [03:15:57] So I think [03:15:59] you can play with not just putting [03:16:01] you know, having a domestic share but it's a share [03:16:03] I think of what's produced in the friendly [03:16:05] countries that really matters for the [03:16:07] export supply elasticity. [03:16:09] And so I think there you get [03:16:11] to break this sort of [03:16:13] which you do in the empirics, which you [03:16:15] absolutely do in the empirics. [03:16:17] But I think conceptually it'd be [03:16:19] useful to do it in the theory as well [03:16:21] so that you can break sort of [03:16:23] the whole thing down overall. [03:16:25] Okay. [03:16:27] This is again [03:16:29] sort of another sort of little [03:16:31] sort of toy model [03:16:33] in which [03:16:35] this is with three [03:16:37] countries where if you have [03:16:39] sort of a hundred percent [03:16:41] sum in if you have [03:16:43] sort of you know three countries [03:16:45] this is where [03:16:47] it's all [03:16:49] China and then that's what you have [03:16:51] now which is sort of [03:16:53] essentially it's only one [03:16:55] country but here if you have two [03:16:57] countries with China is about a third [03:16:59] then you get [03:17:01] term the risk driven [03:17:03] tariff premium would be smaller. [03:17:05] So I think quantitatively [03:17:07] makes sense if there's another country. [03:17:09] Alright, so [03:17:11] interesting paper I had a lot of fun [03:17:13] and I look forward to having many [03:17:15] questions answered [03:17:17] in what follows. [03:17:19] Thank you so much. ## Q&A (03:18:16 – 03:32:52) [03:19:14] There's more questions so [03:19:16] is there another equilibrium [03:19:18] which is the repeated game cooperative equilibrium [03:19:20] where [03:19:22] US doesn't put a tariff and then [03:19:24] China doesn't put a tariff and then [03:19:26] like the digital signal cycle [03:19:28] and the free trade is sustained [03:19:30] with the repeated game cooperation [03:19:32] and if you [03:19:34] there is one then [03:19:36] there are ways to rationalize [03:19:38] the breakdown of the cooperative equilibrium [03:19:40] at that time perhaps [03:19:42] like the Chinese capital G [03:19:44] maybe there's like any crease in the Chinese [03:19:46] export restriction use [03:19:48] maybe not some evidence that China [03:19:50] like started getting draws [03:19:52] that make it more prone to a GDA [03:19:54] and that will bring down the [03:19:56] cooperative equilibrium at that time [03:19:58] because like a time series of the [03:20:00] restriction. Emily? [03:20:02] There are three questions [03:20:04] closely related I think so [03:20:06] on the section 301 tariffs [03:20:08] the section 301 tariffs were imposed [03:20:10] on these big lists of goods [03:20:12] but the coupon rate [03:20:14] the official ad for the add-on [03:20:16] was the same within each one of those buckets [03:20:18] so it was 25% [03:20:20] or 10% or 5% [03:20:22] so there wasn't any [03:20:24] differentiation within those buckets [03:20:26] across goods which seems problematic [03:20:28] and I'm actually wondering [03:20:30] are you working in the percentage increase [03:20:32] in the tariff rate or the percentage [03:20:34] point increase in the tariff rate [03:20:36] that could explain what's going on [03:20:38] and the risk maybe related to that [03:20:40] also when you're looking at benchmark tariffs [03:20:42] so I think this is pretty [03:20:44] real one if I understood correctly [03:20:46] that's a really cool result [03:20:48] but it would seem to be closely [03:20:50] correlated with the pattern of [03:20:52] what are the labor intensive products [03:20:54] and where does China have to compare to the ban [03:20:56] so that's a worse price [03:20:58] I kind of like to see it run [03:21:01] I'll ask one and then we should give you time [03:21:03] to answer [03:21:05] mine is a little bit like outside of the models [03:21:07] I'm wondering what you think about it [03:21:09] so [03:21:11] it was pretty cool to see that the increase [03:21:13] is on the import share [03:21:15] I think Matilda mentioned you would have imagined [03:21:17] in a model it should have been [03:21:19] the share over total absorption [03:21:21] including domestic production [03:21:23] what I suspect my curiosity is [03:21:25] it could be [03:21:27] that politicians are spending a lot of time [03:21:29] and attention on just how much do we import [03:21:31] from China irrespective [03:21:33] of whether we have a huge capacity [03:21:35] to use it domestically [03:21:37] and that becomes salient [03:21:39] even if that probably is not what [03:21:41] a model would predict [03:21:43] and that generates an artificially strong correlation [03:21:45] the Europeans for example are obsessed [03:21:47] about what we import from China [03:21:49] irrespective of how much of that [03:21:51] are we actually producing domestically anyway [03:21:53] which for dependency [03:21:55] or for fear of [03:21:57] retaliation later should really be the right concept [03:21:59] and I'm wondering [03:22:01] what your thoughts are on it [03:22:16] so should I just answer [03:22:18] you or also I don't know what the rules are [03:22:20] because you can't get at you [03:22:22] yeah so a couple of things [03:22:24] and then I can take over [03:22:26] so Matilda thanks for [03:22:28] a great discussion and for raising all of those questions [03:22:30] so [03:22:32] the simplified model [03:22:34] that you [03:22:36] showed makes a lot of sense that's kind of how we started thinking about [03:22:38] things absolutely the input output [03:22:40] stuff doesn't matter we put it there for [03:22:42] quantitative purposes because [03:22:44] we just want the tariffs [03:22:46] to make [03:22:48] more sense and this [03:22:50] also gets at a little bit of Giovanni's question [03:22:52] which is like that first part [03:22:54] the terms of trade part [03:22:56] why it's uniform so actually Giovanni [03:22:58] you could potentially [03:23:00] break that if you made these [03:23:02] betas like the input [03:23:04] the input output things [03:23:06] to make them sector specific [03:23:08] but [03:23:10] the variation that you would get is very very small [03:23:12] so it turns out that it's like actually really hard [03:23:14] to get [03:23:16] a lot of variation in the terms [03:23:18] of trade effect even with [03:23:20] with fixed factors etc [03:23:22] so quantitatively that's not [03:23:24] easy in theory yes but when you put the numbers [03:23:26] on it it's actually it's pretty [03:23:28] difficult so we're finding [03:23:30] that we do really need this [03:23:32] what we call the geopolitical part or maybe [03:23:34] Matilda has a different [03:23:36] word for it to get the [03:23:38] variation okay [03:23:40] and that kind of gets [03:23:42] to the first mover premium [03:23:44] and I think Ahmed wanted to say something on that [03:23:46] so I'll let him do that [03:23:48] and then [03:23:50] regarding [03:23:52] the connection to [03:23:54] Mateo's work which [03:23:56] we do with co-authors which we do [03:23:58] side but today I [03:24:00] was not [03:24:02] siding because of time [03:24:04] of course so I would say that [03:24:06] yes on the intensive margin [03:24:08] there are some of the predictions but we do have this extensive margin [03:24:10] this [03:24:12] probability part that actually does [03:24:14] matter it matters [03:24:16] both in theory and quantitatively [03:24:18] and [03:24:20] I'm going to let Ahmed say something about the theory [03:24:22] but on the quantitative side [03:24:24] again it's like generating a lot of the variation [03:24:26] once again it gets to the [03:24:28] same answer to Giovanni like it's just not [03:24:30] that easy to get this type of variation [03:24:32] that we see that we see in the data [03:24:34] okay I answer Giovanni [03:24:36] andre [03:24:38] lacked import share [03:24:40] aha so [03:24:42] we had a big conversation with Ahmed yesterday while I was flying [03:24:44] I would love to do that [03:24:46] the part where we're trying to test it [03:24:48] from the China side which is what you're saying [03:24:50] it's [03:24:52] we have not been able [03:24:54] to find the data [03:24:56] to do that because this GTA [03:24:58] data and stuff it's a count [03:25:00] it's not [03:25:02] an intensive margin measure [03:25:04] but if you have suggestions on that [03:25:06] I would love to know because [03:25:08] yes I also would like to do that [03:25:10] and then [03:25:12] Ahmed do you want to handle the part about [03:25:14] the cooperative [03:25:16] game because this is your baby [03:25:18] you know this stuff better than me [03:25:20] and then [03:25:22] Emily [03:25:24] so [03:25:26] the stuff about China [03:25:28] having comparative [03:25:30] advantage so again like [03:25:32] here the way that I thought about [03:25:34] it was that we were [03:25:36] really doing that [03:25:38] through the protection for sale part [03:25:40] but I guess maybe [03:25:42] you think otherwise so perhaps [03:25:44] we can talk about it a little bit more after but [03:25:46] it means throwing in another control [03:25:48] which I would love to know what that is [03:25:50] so yeah [03:25:52] go ahead [03:25:55] yeah so again thanks a lot for all the comments [03:25:57] just [03:25:59] one point on the choice of the model [03:26:01] I think we want [03:26:03] I think it needs clarifying that [03:26:05] having the input output loop [03:26:07] has this added benefit that [03:26:09] it gives a clear sense that [03:26:11] input output linkages lower the optimal [03:26:13] tariff so we know the gross formula very well [03:26:15] and this is kind of a [03:26:17] interesting new [03:26:19] amendment to that formula [03:26:21] that shows that input output loops have this effect [03:26:23] it's not [03:26:25] necessary really for the mechanism but it just [03:26:27] makes it more general but in terms of [03:26:29] the extensive margin the pie [03:26:31] the risk part [03:26:33] so we were concerned about [03:26:35] two things one in reality we see [03:26:37] tariffs that are applied [03:26:39] on goods that never experience [03:26:41] an export restriction so we want [03:26:43] to be able to rationalize this in the data [03:26:45] so our model does predict [03:26:47] that so you will see goods that are [03:26:49] subjected to the tariff because [03:26:51] ex ante they could have [03:26:53] received an export restriction but ex post [03:26:55] the draw doesn't justify that [03:26:57] the other thing is that the extensive [03:26:59] margin actually gets you this [03:27:01] big variation [03:27:03] with the intensive margin across sectors [03:27:05] and going back to Giovanni's [03:27:07] point so if I had [03:27:09] added specific factors, heterogeneous [03:27:11] betas into the model and then [03:27:13] simulated this optimal [03:27:15] tariff problem [03:27:17] I would get very little variation [03:27:19] so these GE implied [03:27:21] sector level [03:27:23] export supply elasticities typically are not [03:27:25] going to be very large so what is nice [03:27:27] about the model is that we get this [03:27:29] meaningful variation across sectors [03:27:31] and in [03:27:33] my experience it's difficult to break [03:27:35] uniformity you kind of get quasi [03:27:37] uniformity but this kind of breaks [03:27:39] the uniformity result [03:27:41] but I think going back to Emily's [03:27:43] point I think one thing that we really do [03:27:45] need to add is what you mentioned because [03:27:47] in your paper in your restud paper [03:27:49] you guys show that this [03:27:51] input output stuff [03:27:53] means that you want to have targeted [03:27:55] protection based on comparative advantage [03:27:57] and we don't have that so I think as [03:27:59] another [03:28:01] covariate I think it would make sense to add [03:28:03] factor intensity or [03:28:05] some measure of revealed comparative advantage [03:28:07] to kind of control for that mechanism [03:28:09] so that's a great point thank you [03:28:11] why don't we [03:28:13] collect a second round of questions and [03:28:15] you still have a few minutes [03:28:17] Peter [03:28:19] Thanks very much I know that [03:28:21] things can get very complicated [03:28:23] very quickly but in much of what we've discussed [03:28:25] today we've omitted [03:28:27] concerns about domestic politics [03:28:29] and of course we all know that [03:28:31] specifically with the tariff it has [03:28:33] tremendous domestic distribution [03:28:35] consequences [03:28:37] there are gators and there are losers [03:28:39] what was particularly interesting about [03:28:41] the 2018 tariffs was that the [03:28:43] domestic politics of these were acute [03:28:45] that in fact [03:28:47] the administration followed what was [03:28:49] at the time relatively standard procedure [03:28:51] by holding hearings allowing firms [03:28:53] to lobby for exceptions to those tariffs [03:28:55] and indeed there's [03:28:57] plenty of evidence to suggest [03:28:59] that politically connected firms located [03:29:01] in Republican districts that were making [03:29:03] large amounts of contributions [03:29:05] all received exceptions from these [03:29:07] tariffs because they wanted to keep the price of their [03:29:09] inputs low right so [03:29:11] my question I guess to you is in [03:29:13] what sense, what direction is that bias [03:29:15] you're finding right is by [03:29:17] ignoring the politics do [03:29:19] do we feel stronger [03:29:21] about the story you want to tell [03:29:23] or are we meeting an important part [03:29:25] of the analysis here that should be taken to Kiddilip? [03:29:27] Shau? [03:29:29] I was interested in maybe a slightly narrow [03:29:33] point on estimation of Zeta [03:29:35] you're estimating this probability that [03:29:37] foreign districts with respect [03:29:39] to import dependents at home [03:29:41] that was a dress that you want [03:29:43] targeted for this country but not that [03:29:45] country they don't want a variation [03:29:47] of export restrictions [03:29:49] they want to live with little trade [03:29:51] data that you use and I think [03:29:53] at least in agricultural markets which could just [03:29:55] be inspecial part of what happens [03:29:57] is that domestic prices go up [03:29:59] so Indonesia for example will put in [03:30:01] a poverty ban for exports [03:30:03] no matter where they go [03:30:05] and I think that to the extent that you have [03:30:07] this common across all trade [03:30:09] partners action that might dampen [03:30:11] actually and cause you to understate [03:30:13] potentially the U.S. [03:30:15] why don't we stop here [03:30:17] we give you two minutes to wrap it up [03:30:19] and then... [03:30:21] Let me come back to your point [03:30:23] and then I'll give Ina the microphone [03:30:25] so I think one thing about [03:30:27] surge in imports from China [03:30:29] that I think that's definitely [03:30:31] part of the story so one of the goals [03:30:33] was can we rationalize [03:30:35] tariffs [03:30:37] that mitigate [03:30:39] strategic risk [03:30:41] without adding any extra [03:30:43] preference to the government's [03:30:45] objective function [03:30:47] like if the government [03:30:49] just simply values [03:30:51] real GDP [03:30:53] in both countries [03:30:55] can we still rationalize [03:30:57] what's happening [03:30:59] yes but absolutely we're going to [03:31:01] miss out on a lot of going back [03:31:03] to the other question [03:31:05] political economy motives [03:31:07] for policy that are also playing in the background [03:31:09] and also driving these tariffs [03:31:11] so I think [03:31:13] we think that this doesn't give a big picture [03:31:15] but it does give a nice positive picture [03:31:17] of why even purely [03:31:19] well for maximizing governments [03:31:21] who do not care necessarily about [03:31:23] the surge of imports from China [03:31:25] or another adversarial partner [03:31:27] may want to [03:31:29] restrict imports because [03:31:31] firms fail to internalize this excessive [03:31:33] dependence issue [03:31:35] yeah so thank you [03:31:37] for the comment on the GTA [03:31:39] maybe we can talk offline but [03:31:41] did we not exclude [03:31:43] I think we had a specification where [03:31:45] we exclude the agricultural goods [03:31:47] and so that did not [03:31:49] seem to matter a lot [03:31:51] and then of course so the big question [03:31:53] the way that so [03:31:55] Ahmed and I did some back and forth on [03:31:57] what you were seeing yesterday so [03:31:59] exceptions [03:32:01] indeed were granted [03:32:03] especially for inputs [03:32:05] and this is something that [03:32:07] we could potentially [03:32:09] actually model and [03:32:11] there is data on it [03:32:13] so I think that it was a [03:32:15] hairy [03:32:17] kind of thing that the [03:32:19] Ahmed had dealt with before but absolutely [03:32:21] I actually think that what you said is important [03:32:23] so I'm glad that Ahmed gets to [03:32:25] hear it because now we can [03:32:27] do it because I do actually think [03:32:29] we should bring it in and but I cannot [03:32:31] answer to you how it's going to play out [03:32:33] because I know that data [03:32:35] is there and we can put in the model and I want to do that next [03:32:37] so thank you for bringing that up I appreciate it [03:32:39] thank you okay fantastic [03:32:41] thank you okay so now we're [03:32:46] on lunch break which is outside [03:32:48] and we reconvene at 1.30