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

# Tariffs, Investment, and the Missing Trade Collapse

Authors: Michael E. Waugh, Andrea Prestipino, Andrea Raffo, Francesco Ferrante  
Discussant: George A. Alessandria  
Video: https://www.youtube.com/watch?v=wEt4iKiNNZQ&t=9287s


## Talk (02:34:47 – 03:21:30)

[02:34:54] Okay, so I'm happy to put things down.  
[02:35:32] And Enrique for putting my paper on the program.  
[02:35:35] Francesco's over there ready to answer the hard questions.  
[02:35:39] The disclaimer applies since we're all Federal Reserve employees.  
[02:35:42] This is our views and not the views of the federal system and so forth.  
[02:35:47] So the way I like to start talking about this paper is like I'm really big into origin stories.  
[02:35:52] So like, you know, Batman is like cool and stuff.  
[02:35:54] But okay, so like where did this paper come from was actually at the end of 2025, you know,  
[02:36:01] a discussion between me and Andrea.  
[02:36:03] And then I know Andrea was having discussions with Andrea and Francesco about what did we get wrong  
[02:36:09] after Liberation Day.  
[02:36:12] So we don't need to go through it, but I will like recap it is, you know, we put Terrace on the United States.  
[02:36:21] They were very large.  
[02:36:23] I mean, historically large.  
[02:36:24] And then very quickly economists as a profession started pronouncing, you know, various degrees of doom and glue.  
[02:36:33] And then, you know, when you got to the end of 2025 and looked at it, things seemed okay.  
[02:36:40] We didn't seem to really see the doom and glue.  
[02:36:42] And so the wall this Wall Street Journal article, I think like summarized it's well, in fact, I think our president circulated this around.  
[02:36:49] I don't know.  
[02:36:50] And and and then was like in some of the meetings is like staring at me, Mike, like what do you think?  
[02:36:55] And then so like why everyone got, you know, Trump's tariffs wrong.  
[02:37:01] And so I think the best way to kind of like see this is actually just looking at the trade flow data.  
[02:37:07] So what I have here is the US effective tariff rate.  
[02:37:13] So this is just like computing the in the simplest way possible.  
[02:37:16] We're just taking like total duties collected divided by volume.  
[02:37:20] All right.  
[02:37:21] And that gives us a measure about basically how how restrictive trade is the tariff is.  
[02:37:26] And then it's hovering around say about 3%.  
[02:37:28] And then there's some stuff happening in February and then we have liberation day right here.  
[02:37:32] And then at peak, we're almost at 12 percentage points of effective tariff rate in the United States.  
[02:37:38] All right.  
[02:37:38] So this isn't quite smooth Holly level.  
[02:37:41] So that's like about 20%.  
[02:37:42] But this is a, you know, the highest we've ever seen since smooth Holly before the Great Depression.  
[02:37:47] And then subsequently, there's been this Busan truce with China, which lower tariffs.  
[02:37:52] And then the Supreme Court struck down the IEPA tariffs.  
[02:37:56] And now we're down, I think about like 8%.  
[02:37:58] I think when we project out, it's going to go down a little bit lower.  
[02:38:01] But this is a big shock to US trade policy.  
[02:38:06] So on the other panel, what I have here is, you know, real imports.  
[02:38:11] Right.  
[02:38:11] So I have both like it's important to look through is both we have the quarterly series  
[02:38:16] and then the annual averages.  
[02:38:18] And so it was growing.  
[02:38:20] And then here's liberation day.  
[02:38:22] And there's a couple of things I want to point out is for this first do that annual  
[02:38:25] averages, you know, trades actually grown right through this.  
[02:38:31] That's kind of like you go back to the Wall Street Journal article.  
[02:38:34] It's like, hey, you know, these trade got economists were on Twitter saying trade's  
[02:38:38] going to collapse and all these bad things are happening.  
[02:38:40] And then we actually had higher trade levels than we were before.  
[02:38:44] When you start looking at the figure as well, you start seeing some things  
[02:38:47] that are going on in the sort of late to George's work is that we see this  
[02:38:52] massive spike right before liberation day.  
[02:38:55] So importers, the story always hears importers for saw the the tariffs coming.  
[02:39:00] They started accumulating importing stuff to accumulate  
[02:39:05] inventories to sell abroad.  
[02:39:07] So that's like obviously sell later.  
[02:39:10] That's obviously an important story.  
[02:39:12] But then there's this other issue that's floating on right here.  
[02:39:16] Right.  
[02:39:17] Imports just kept rising over this whole time.  
[02:39:20] So the whole paper is basically we want to kind of dissect this and think  
[02:39:23] through like what are the different mechanisms behind this and then more  
[02:39:27] broadly the US macro economy.  
[02:39:29] So let's hope the clicker works here.  
[02:39:31] OK, so what am I going to do here?  
[02:39:32] I mean, the big kind of question basically is like that Wall Street Journal.  
[02:39:36] I mean, the take on the Wall Street Journal thing is like, why was the US  
[02:39:39] economy and US trade so resilient through this time period?  
[02:39:43] All right.  
[02:39:43] So the way the talks going to work and basically, and this is kind  
[02:39:46] of like the contribution as well that we're doing is I'm going to talk  
[02:39:49] through some data aspects of this.  
[02:39:52] And one of the things that like when Andrea and I were first talking  
[02:39:55] about this at the end of 2025, what we thought was interesting was to like  
[02:39:59] think through trade in terms of in use.  
[02:40:02] So like consumption, intermediates and capital goods.  
[02:40:07] Now, one of the things that I'm going to show you is that the  
[02:40:09] incidents of the tariffs hit those different in uses very differently.  
[02:40:15] In particular, consumption was hit very hard, intermediates in the middle  
[02:40:19] and capital goods are basically left untouched completely with respect to the tariffs.  
[02:40:25] Then I'm going to show you some things about trade and AI, which suggests that  
[02:40:30] there's a like a massive kind of like import surge related to AI,  
[02:40:35] trade and AI related products.  
[02:40:37] OK, so that kind of tells us like some like starts to shape some of the shocks  
[02:40:41] that we want to put into a model and decompose things.  
[02:40:45] So what we're going to write down is like, and this is kind of new to me.  
[02:40:48] So that's why we got Francesco here is like new Keynes, you know,  
[02:40:51] binocani model.  
[02:40:52] So we got a lot of stickiness, a lot of adjustment costs,  
[02:40:57] all the bells and whistles are in here.  
[02:41:00] The way to think through this is we want to use this as an accounting device  
[02:41:04] and then fit it to data or fit it to some series and then decompose  
[02:41:09] what different shocks did to the US economy.  
[02:41:11] And the shocks we're basically thinking about is what role the tariffs play.  
[02:41:16] And again, we're going to be emphasizing these different components  
[02:41:19] of consumption, intermediates and capital.  
[02:41:22] The other thing that we're going to when we look through the data,  
[02:41:24] we're going to see this investment boom and we got to somehow wedge  
[02:41:27] that into the model.  
[02:41:28] And so we're going to put this marginal efficiency of investment shock in here.  
[02:41:32] So that's like in the KT plus one equals one minus delta KT plus I.  
[02:41:37] There's going to be a shock that hit in the eye there.  
[02:41:39] So per dollar spin, I'm going to get more capital out of it.  
[02:41:43] That's going to be our pledge for AI.  
[02:41:45] I'll talk more about this and why we have other ways to deal with this as well.  
[02:41:50] The other thing, I'm not going to talk much about it, but this was a lot of the  
[02:41:54] conversation around the time during the liberation day was the dollar.  
[02:41:57] So the dollar started moving in ways that we didn't typically expect.  
[02:42:01] In fact, the dollar depreciated after liberation day.  
[02:42:03] We want to at least have that as a potential force when the tariffs go on  
[02:42:08] around the model.  
[02:42:10] Okay, so we're going to have this model with these different shocks.  
[02:42:14] And then we want to feed these shocks in to hit a couple of series and  
[02:42:18] then inspect what's going on with our economy.  
[02:42:22] So again, it's like use the models in accounting device.  
[02:42:25] So one of the things that does come about this is we're going to hit a couple  
[02:42:28] series, we're going to leave other series untargeted.  
[02:42:31] In particular like GDP consumption inflation and in aggregate imports.  
[02:42:36] And then the model is actually going to do a very good job replicating all  
[02:42:40] these features.  
[02:42:40] I mean, it's not going to be like point for point, but  
[02:42:42] it's going to get that broad fact that the US economy is very resilient.  
[02:42:47] Inflation is going to go up, imports are going to go up, consumption is going to  
[02:42:50] go down, it's all going to look real good.  
[02:42:52] All right?  
[02:42:54] And then we're going to look through these different shocks and say, what's up?  
[02:42:58] So what do we see is basically when we ask this question,  
[02:43:02] like go back to the Wall Street Journal article, what did we get wrong?  
[02:43:05] Well, what we got wrong was we're missing how important and  
[02:43:09] how strong this AI investment boom is.  
[02:43:12] It completely masks any negative effects of tariffs in the model.  
[02:43:18] All right?  
[02:43:18] So this is kind of like the way to think about this is when, I don't know,  
[02:43:22] you see your favorite person talking about the US economy and say, hey,  
[02:43:26] it's really strong, tariffs didn't seem to do anything.  
[02:43:29] We got to kind of trace it back to, we got to think through this shock.  
[02:43:33] Right?  
[02:43:35] What's that?  
[02:43:37] Yes?  
[02:43:39] Yes?  
[02:43:40] Yeah.  
[02:43:41] No, it is.  
[02:43:42] No, the AI, I mean, they're separate out.  
[02:43:53] Well, it's everything.  
[02:43:54] It's like, we'll talk about this myself.  
[02:43:58] But I would prefer, honestly, I prepared a slide deck for 35 minutes so  
[02:44:08] people can talk a lot, by the way.  
[02:44:10] So because it said on the website, 35 minutes, yeah.  
[02:44:20] Yeah.  
[02:44:22] Yeah.  
[02:44:26] It is what you see.  
[02:44:29] Yeah, yeah, yeah, exactly.  
[02:44:30] So and we're going to show, I'm going to put numbers on the doom and gloom.  
[02:44:32] So it's going to like, output's going to go down by about like a little less than  
[02:44:36] 1%.  
[02:44:37] So it's going to be a, tariffs are going to be big here,  
[02:44:40] partly because we have capital.  
[02:44:42] It's going to be a big component.  
[02:44:44] So there was a lot of doom and gloom.  
[02:44:46] It's just we didn't see it.  
[02:44:53] Something just like, yeah, I'm going to show you, I'm going to show you this.  
[02:44:56] I got this.  
[02:44:58] And then one of the things that Matt, so you might say like, okay,  
[02:45:01] okay, that's fine.  
[02:45:03] Sometimes I wake up at night, go like, dude, this is obvious.  
[02:45:06] But like it, what's not obvious is when you start thinking through the policy  
[02:45:10] implications of this is, and this is where the New Keynesian model actually is  
[02:45:15] like really important here is, you know, the output inflation trade off is  
[02:45:20] actually way more subtle than just like tariffs or supply shock.  
[02:45:24] In particular, like capital goods here with the New Keynesian frictions are  
[02:45:28] actually going to act like a demand shock.  
[02:45:31] So output's going to go down.  
[02:45:33] Inflation's going to go down in response to the tariff on the capital goods.  
[02:45:38] Okay, and so that's an important lesson for policy makers when we think about the  
[02:45:42] structure of tariffs being put on, how should a monetary policy authority  
[02:45:47] respond to that?  
[02:45:48] All right, so let's get going.  
[02:45:51] So just like to recap about like how resilient the US economy was.  
[02:45:55] We have a couple sides of some macro aggregates.  
[02:45:58] First here we have GDP.  
[02:46:00] I'm going to highlight when we talk about GDP, we're reconstructing GDP basically  
[02:46:04] the normal way net of structures.  
[02:46:06] So structures are not in here.  
[02:46:08] So if you know anything about the US economy right now, if you're trying to  
[02:46:11] sell a home, hopefully you don't.  
[02:46:16] If you're trying to sell a home, it's really bad right now.  
[02:46:19] Okay, so like we pull structures out and then this is what GDP looks  
[02:46:24] like relative to trend and since liberation day, it's like it's  
[02:46:29] booming relative to trend.  
[02:46:31] You look at investment, again this is investment net of structures in here.  
[02:46:35] It's completely booming.  
[02:46:44] Once we have a linear trend relative to 2015 to 2019.  
[02:46:48] Yeah, and then down here we have the exchange rate.  
[02:46:53] So right now the dollar was appreciating until liberation day and  
[02:46:57] then right after it depreciated.  
[02:46:59] And then here we have inflation and then inflation what we put up was  
[02:47:03] core PCE relative to the survey professional forecasters.  
[02:47:07] So inflation is about 90 basis points of relative to what we anticipated enough.  
[02:47:12] Yeah, yeah, yeah, yeah, yeah, yeah, yeah, exactly.  
[02:47:20] So I mean, because there are aspects of the US economy that are not doing it  
[02:47:24] and it's particular structures.  
[02:47:25] So yeah, yeah, we have like the intellectual property component in that.  
[02:47:36] Yeah.  
[02:47:37] This is the trade element of this and this is what I was talking about when you  
[02:47:42] look at tariffs and start breaking it down by in use.  
[02:47:45] So I have here in parallel, so this is consumer goods, this is intermediate goods,  
[02:47:51] and this is capital goods, tariffs and imports.  
[02:47:55] A thing to think about is total goods trade.  
[02:47:57] This is about a third, a third, a third.  
[02:48:00] All right, so there's kind of equally bend size buckets here.  
[02:48:04] So what's really striking about this is the structure of the tariffs that were imposed  
[02:48:11] were heavily oriented towards consumer goods.  
[02:48:14] So you see that like right, it's hovering around 4% and then after liberation,  
[02:48:19] they had spiking up to nearly 20% on consumer goods and then coming down to like 12.  
[02:48:27] This is an effective...  
[02:48:28] Tariff effective, yeah.  
[02:48:31] I'm a bother.  
[02:48:40] The question.  
[02:48:50] Or yeah, it's always collected duties over volume.  
[02:48:55] Now you can do other things on this, which I've done on my website,  
[02:48:58] www.tradewartracker.com, Adrian's like, yeah, that's a good website,  
[02:49:03] which has the detailed statutory rare tariffs on it and you can look at the statutory thing  
[02:49:08] and you'd come up with the same thing.  
[02:49:15] So I mean, this is the tariff rate, like here it's high,  
[02:49:19] you have the intermediate goods, it's kind of in the middle,  
[02:49:22] and then this is the interesting one that capital goods, yeah, tariffs went up on capital goods,  
[02:49:28] but nowhere to extend as intermediates or consumption goods.  
[02:49:33] And then you kind of look at, well, what did that do to trade?  
[02:49:36] So imports for consumer goods, in a lot of ways, the way I kind of think about this  
[02:49:40] is this looks like our standard kind of trade models, right?  
[02:49:43] Tariffs went up a bunch, trade collapsed, it's starting to recover a bit  
[02:49:49] as tariffs have come down.  
[02:49:50] This is kind of like exactly the kind of normal story we'd expect.  
[02:49:55] When we look at intermediates, there we start to see the inventory front-rolling.  
[02:50:01] And again, I would highlight about the way we constructed the data,  
[02:50:05] the way we built this data set is we put pharmaceutical products  
[02:50:08] into intermediate product here, right here.  
[02:50:11] And so this is a big component of the front-running right here.  
[02:50:15] But what's happening is these companies saw these tariffs  
[02:50:18] feed into their production processes and they start importing this,  
[02:50:22] very aggressively ahead of time and building inventories.  
[02:50:25] And then over time, yeah, at the quarterly frequency, trades collapsed.  
[02:50:30] But if we look at the annual average, so the red dot, and  
[02:50:33] we integrate over all of 2025, still trade is up over that time period.  
[02:50:39] So it's like we pulled in and then dropped down.  
[02:50:42] And we take the linear trend 2015 to 2019 projected out, it's relative to that thing.  
[02:51:00] This is the interesting thing is that even though the tariff,  
[02:51:07] it's a slightly higher tariff right here, is that imports of capital goods  
[02:51:11] have given completely surging over this whole time period.  
[02:51:14] Okay.  
[02:51:15] Yeah, so that's, yeah, so this is the next slide on this.  
[02:51:22] So in response to some of the policy makers questions that we're facing at the bank,  
[02:51:29] one question is like how important is AI in this?  
[02:51:32] And so earlier this year, I wrote this article, which the basic idea is like,  
[02:51:38] how can I measure AI's impact in this?  
[02:51:40] So when you look at the trade data, the way it's structured is there's  
[02:51:44] like these HS10 codes.  
[02:51:45] So there's about 20,000 HS10 codes.  
[02:51:48] HS10 codes have detailed descriptions about what's associated with each one of those kind of products.  
[02:51:56] And you can get more information, you can merge it in with like NAICS and blah, blah.  
[02:52:00] Okay, so the idea was, why don't we?  
[02:52:07] What's AI?  
[02:52:08] I know it's like that circular.  
[02:52:10] But like, I mean, I did some other things leading into this, but like we can get to like talk about that later.  
[02:52:16] But the idea is what you do is you construct a prompt and then you go through every single code and say,  
[02:52:22] do you think this would go into a data center?  
[02:52:25] What's its relevance?  
[02:52:26] Okay, so you whip through the whole thing, it takes about a day.  
[02:52:30] The Fed still hasn't paid me back for like the $500 it took to do this.  
[02:52:34] But anyways, they like, I did this thing and then what it does is it categorizes every single HS10 code by like how relevant it is.  
[02:52:46] I just did a simple high, medium, low.  
[02:52:48] And then it breaks up the high stuff into what elements of a data center would go to.  
[02:52:54] So like here, for example, this is high AI relevance.  
[02:52:57] Is it compute hardware?  
[02:52:58] Is it electrical power, networking, telecom, cooling and HVAC?  
[02:53:03] I have some other stuff there too.  
[02:53:05] And then low AI relevance.  
[02:53:07] So one of the things that you first see very quickly when you do this kind of thing is like,  
[02:53:12] most of the stuff that's declared to be high AI relevance is basically a capital good.  
[02:53:18] All right.  
[02:53:18] So now we can start to put these pieces together that like the capital goods are surging.  
[02:53:23] And a lot of this is like going up level.  
[02:53:32] But what we know is that it's likely to be going into a data center.  
[02:53:40] Yes.  
[02:53:43] Or I think that the rates, which are interesting.  
[02:53:46] So most of this stuff was relatively low tariff in particular like compute hardware.  
[02:53:51] So this is basically semiconductors.  
[02:53:53] That was explicitly exempted by the Trump administration  
[02:53:57] during after liberation day.  
[02:53:59] Networking, telecom, extent.  
[02:54:01] But there's other stuff was not exempted like electrical power stuff was not cooling HVAC was not.  
[02:54:07] And then when you start looking at the growth rates, it's just like crazy.  
[02:54:11] I was showing this to my advisor like a couple weeks ago, like Ravi's like those numbers are wrong.  
[02:54:16] I was like, I didn't even like hesitate.  
[02:54:18] But like, no, these numbers are right.  
[02:54:20] It's like high relevance is like absolutely surging.  
[02:54:23] You know, even you look at stuff that like I like cooling in HVAC for the following reason  
[02:54:28] is that the tariff went up a lot.  
[02:54:31] 10 percentage points.  
[02:54:33] It's barely changed.  
[02:54:34] We're still bringing this stuff in.  
[02:54:36] Right.  
[02:54:37] And then you can compare to low, low AI relevance is it's not very much capital.  
[02:54:42] And like these aren't capital goods.  
[02:54:44] High terror and then trade is falling.  
[02:54:47] Okay.  
[02:54:48] And then just to like put like like further show this stuff is if you look at the aggregate  
[02:54:53] AI capital goods series, it's like crazy.  
[02:54:57] Like so you look at so the blue line is AI relevant and this is value relative to 2023.  
[02:55:03] So it's not the detrended thing relative to 2023.  
[02:55:07] The value of the stuff is up 173 percent.  
[02:55:10] Whereas the non AI relevant.  
[02:55:12] I mean, it's like so small relative to this kind of search here.  
[02:55:16] So just this is just capital goods of trade.  
[02:55:21] Trade imported capital goods.  
[02:55:23] I break it up into AI relevant and not AI relevant.  
[02:55:27] This right now, this in 2023 was about I think it's about 35 percent of capital  
[02:55:34] goods trade.  
[02:55:35] Now it's almost upwards of 50 percent of capital goods trade through this kind of stuff.  
[02:55:40] And then I think partly it's because of because of how big this thing is.  
[02:55:47] Yeah, so there is a collapse here.  
[02:55:49] I want to say it's maybe around it might be around 10 percent something like that.  
[02:55:53] But it's just like orders of magnitude larger.  
[02:55:55] That's one of the thing.  
[02:55:57] And so again, it kind of emphasizes this point is that there's this.  
[02:56:00] We had these tariffs taking place, but we had this just absolute massive surge.  
[02:56:06] In in capital goods in particular related to AI stuff.  
[02:56:10] So now we want to kind of decompose it and like trace out what different shocks we're doing during this time period.  
[02:56:18] Okay, so we're going to have like I mean for me is actually a pretty big model.  
[02:56:23] And by the way, this is like the only slide I have on the model.  
[02:56:26] So if you have like, yeah, for reasons like oh my God, that guy's from Minnesota has no equations in the papers.  
[02:56:31] Okay, so the thing is is that the if you want to ask questions, this is about model.  
[02:56:38] So this is this is the place to do it.  
[02:56:40] But so let me just like tell you kind of like overview about how it's going to work.  
[02:56:44] So it's two country New Kansy and open economy, US and rest of the world.  
[02:56:48] The households, everything is going to be like Armington aggregator over like, you know,  
[02:56:53] a bundle of domestic and imported goods of a consumption good.  
[02:56:56] Okay, and then these consumers are going to also have habits and their preferences.  
[02:57:01] So that's going to make, you know, you know, consumption not going to be respond so fast in response to shocks.  
[02:57:07] They're going to be supplying labor.  
[02:57:09] They're going to have the traditional New Kansy and Union sectors.  
[02:57:12] So wages are going to be sticky.  
[02:57:15] They're going to be saving in bonds and capital, but the US bond is the only internationally traded asset out here.  
[02:57:21] All right, the firms are basically building output with labor, capital and materials.  
[02:57:29] And then we have a CS production function over that particular element.  
[02:57:33] And then each one of these things, the investment goods that go to build capital or the materials are going to be again,  
[02:57:41] Armington aggregators of the domestic and the foreign good.  
[02:57:46] Okay, layered on top of this are all the New Kansy and elements and the other elements that go into kind of like medium scale models.  
[02:57:54] So we have sticky prices.  
[02:57:56] We have adjustment costs on investment.  
[02:57:59] We have adjustment costs on the importing the stuff as well.  
[02:58:04] So that's going to give us a difference between the long run trade elasticity and the short run trade elasticity.  
[02:58:08] We'll talk a little bit about that.  
[02:58:12] So the price, the way the production chain is working is like you're producing like some value added that value added goes into a differentiated sector.  
[02:58:21] In that differentiated sector, you have Rogenberg price stickiness.  
[02:58:25] And then that produces that good.  
[02:58:27] It goes through either into consumption, into investment or into materials.  
[02:58:33] Does that answer your question?  
[02:58:34] Yeah. Wages are not flexible. Wages are sticky.  
[02:58:39] So that's the you supply labor to the unions and it's the union deal.  
[02:58:44] Does that make sense?  
[02:58:48] Yeah, yeah, yeah. So okay.  
[02:58:49] So important elements about what's going on here.  
[02:58:51] Okay, so important elements are so we're going to have these different end uses.  
[02:58:56] So there's a good that is going to go into consumption, a good that's going to go into investment,  
[02:59:00] a good that's going to go into materials.  
[02:59:02] As that good goes into those different elements, they're going to be tariff at different rates.  
[02:59:06] Right. So that's the tariff heterogeneity element.  
[02:59:09] Okay. Here's the invoicing thing.  
[02:59:12] We're going to do the dominant currency paradigm.  
[02:59:14] So, you know, foreign exporters in the United States are sticky in dollars.  
[02:59:20] US exporters into the foreign country, sticking dollars as well,  
[02:59:24] but then flex in the in the foreign price.  
[02:59:26] All right.  
[02:59:27] And then we we're going to put inventories in here again because we want to like I as you see,  
[02:59:33] you can't talk about this without inventories.  
[02:59:36] We're going to do this approach by I mean George might complain about this.  
[02:59:41] We're going to put this approach by Alcler, Longley and Straub.  
[02:59:44] I mean, we're doing this because we love George and we think his stuff is awesome.  
[02:59:48] But we also want to like wedge it into like the New Keynesian framework.  
[02:59:52] The way they do this is basically it's kind of weird, but it works is you take a CS.  
[02:59:59] What is being supplied to the final producer value added is material services.  
[03:00:05] Material services is a CS production function of a flow of materials and the stock.  
[03:00:13] So the stock and the flow are like kind of wedged together.  
[03:00:16] What that means is that the person providing the material services values the stock.  
[03:00:22] That's a way to get it in there.  
[03:00:24] All right.  
[03:00:40] That's not like the stock.  
[03:00:42] So it's like just think about the standard like capital accumulation equation,  
[03:00:46] but there's a shock on I.  
[03:00:48] And so per dollar spent, you get more capital out.  
[03:00:53] We're going to call that the investment shock.  
[03:00:56] Yeah, we have it.  
[03:00:58] We'll make this come up.  
[03:00:59] We have a two sector model where we kind of more think more explicitly about AI  
[03:01:03] and people consuming AI and building it out.  
[03:01:06] So a little bit more involved.  
[03:01:08] We might go into that do that as our baseline in the future.  
[03:01:11] But yeah, I said one thing that keeps up me up at night is that this paper is  
[03:01:38] completely obvious.  
[03:01:39] Number two is that that's the second thing is Kevin Hasett shows up on  
[03:01:42] Fox and Friends and says says these guys missed the point.  
[03:01:48] The terrace caused the I shock.  
[03:01:50] We're going to take a stand in the model that those are completely orthogonal  
[03:01:54] things and they just happen to show up at the same time.  
[03:01:57] So I mean, what's that?  
[03:02:01] Super intelligence is what we're working with here.  
[03:02:04] Yeah, yeah, yeah.  
[03:02:17] Yeah, that's a thing.  
[03:02:19] I mean, like this is a yeah.  
[03:02:22] So the way to think about it is there's like once it goes through that  
[03:02:25] differentiated thing, there's like one good.  
[03:02:28] And then it comes through the border and I say, I'm using that one good  
[03:02:31] for consumption, it gets slapped with that tear.  
[03:02:36] The guy was as I'm using for intermediates and get slapped with a different tear.  
[03:02:39] So it you know, we're assuming that the guys in the model are reporting.  
[03:02:44] You know what I mean?  
[03:02:45] Yeah, yeah.  
[03:02:50] Yeah, but yeah, yeah, we're the same thing.  
[03:03:13] Yeah, yeah.  
[03:03:14] So we're missing like the energy that you guys have in that for sure.  
[03:03:17] And then I mean, I think that's why it yeah.  
[03:03:21] I mean, I think that's also why I alluded to the two sector model is that  
[03:03:25] that allows us to like break up investment and to investment going into AI  
[03:03:29] and then investment going into like regular capital goods.  
[03:03:32] And then again, if you remember that table, like semiconductors are zero.  
[03:03:38] Right.  
[03:03:38] So there is a sense of which in within like and that's kind of to your point  
[03:03:41] within investment goods is there is heterogeneity around there that matters as well.  
[03:03:46] OK, so I'm going to kind of just like now talk through some mechanisms  
[03:03:50] about how them to like keep in mind through impulse response functions.  
[03:03:54] So I'm going to do like three things.  
[03:03:55] I'm going to look at, you know, basically shock the model with a 10 percent  
[03:04:00] tariff and highlight a couple of features.  
[03:04:02] So this first slide is about consumption.  
[03:04:05] So we're just going to put a 10 percent tariff on the consumption goods  
[03:04:08] and see how the economy responds.  
[03:04:10] So like the thing is there's two things to take away from this whole.  
[03:04:15] I mean, there's six panels here, but two things to really take away from  
[03:04:18] is first one is inflation.  
[03:04:21] OK, so the consumption tariff with the invoicing basically just one for one  
[03:04:27] passes through right into the consumer.  
[03:04:29] So if you look at like the consumption, the import share in consumption,  
[03:04:34] it's about 6 percent.  
[03:04:36] So then you do a 10 percent tariff times 6 percent, just like a first order  
[03:04:42] approximation will get you an increase in the price level about 60 basis points.  
[03:04:48] That's exactly what happens right here.  
[03:04:50] So it's just like boom right in.  
[03:04:52] And then it's a one time thing too.  
[03:04:54] And so then you see like the gerbil and the snake kind of go through over  
[03:04:57] four quarter inflation and then it just like completely dies out.  
[03:05:00] All right.  
[03:05:01] So that's like our typical thing about like I think when the first  
[03:05:04] we were talking about this at the bank is like when the tariffs first came out.  
[03:05:08] Oh, this is what we'd expect to see happen with inflation.  
[03:05:11] The second thing I want to highlight about yeah, no retaliation.  
[03:05:16] This is pure like just trying to impulse response kind of stuff and see what happens.  
[03:05:20] The other thing is if you look at GDP, the GDP response here is actually quite small.  
[03:05:26] So this is actually why I think we'll come back to this investment is quite important.  
[03:05:31] But the GDP response here actually looks like if you look at, for example,  
[03:05:35] this paper by Pablo Fiegelbaum and Anna Kondowal in the Brookings just this last spring,  
[03:05:41] it was like a revisit of like what are the effects of Trump tariffs?  
[03:05:44] That is what I would call a standard kind of like trade model that they used.  
[03:05:49] And they came up with an estimate of, you know, 11 basis points on GDP from the Trump tariffs.  
[03:05:56] That's almost exactly what this is showing.  
[03:05:58] So just through kind of standard like Armington channels alone,  
[03:06:02] you basically don't get much in terms of the output costs.  
[03:06:06] So the way to think about this first thing is like consumption is the tariff on consumption  
[03:06:10] is basically all about inflation.  
[03:06:12] When you look at inventories or materials, here we're going to do a slightly different shock.  
[03:06:17] We're going to add in one quarter news.  
[03:06:21] What comes out of this?  
[03:06:22] So this is that 10% impulse response.  
[03:06:26] Here is where the inventory mechanism shows up, right?  
[03:06:30] So these guys see that the price of their stuff that they're going to sell is going to be  
[03:06:34] or buy is going to be more expensive in the future.  
[03:06:37] They pull purchases in and then hold the higher stock and then reduce how much they buy later.  
[03:06:43] All right?  
[03:06:43] So it's like that inventory mechanism here just again to really important to get the trade  
[03:06:49] dynamics right around the tariff shock.  
[03:06:53] These things actually have relatively mild effects.  
[03:06:56] Lation again, because it's materials, it's being passed through all the way.  
[03:07:00] Like it's on a chain before it gets to the consumer.  
[03:07:03] It's much more mild.  
[03:07:05] And then again, the output and the consumption investment response are pretty small here.  
[03:07:11] Now the big one is actually investment.  
[03:07:14] And this is like coming from the trade side kind of like surprised me a lot about how  
[03:07:18] important investment is in the model in response to this.  
[03:07:21] So again, this is the same 10% shock, but now it's on the investment good.  
[03:07:27] Okay?  
[03:07:28] So there's a couple of things to notice here.  
[03:07:30] This first investment collapses a lot very strongly.  
[03:07:34] This is just like standard neoclassical growth model investment dynamics.  
[03:07:38] Right?  
[03:07:39] I want to lower my capital stock a bit.  
[03:07:42] Investment is the flow.  
[03:07:43] So I got to drop my flow a lot.  
[03:07:46] Consumers okay with it.  
[03:07:47] They're going to smooth consumption through it.  
[03:07:49] Okay?  
[03:07:52] Yeah.  
[03:07:53] So that's another one.  
[03:07:55] Lessons can be a bit.  
[03:07:57] So our balance is a bit of a lot.  
[03:08:03] Lessons sort of are a bit of a lot.  
[03:08:10] The 30% of our investment is coming from the ball.  
[03:08:15] So it's not getting like the 3% of the graph is coming from the growth.  
[03:08:21] That's something that we need to be in mind here.  
[03:08:24] So then this is where New Keynesian stuff comes involved is here output is demand  
[03:08:30] determined.  
[03:08:31] Right?  
[03:08:32] So investment is falling.  
[03:08:34] So output needs to fall.  
[03:08:37] Hours needs to fall.  
[03:08:39] So then you get this big collapse in GDP in response to this.  
[03:08:44] And then this is where the deflation component comes in here.  
[03:08:47] So remember hours are falling and then with the sticky wages and the sticky  
[03:08:51] prices that ultimately means marginal costs are falling.  
[03:08:56] And then as these guys through the Rotenberg pricing start to reset their  
[03:08:59] price, they lower their price in response to this.  
[03:09:03] So what's like important about the investment thing is you get a large  
[03:09:07] output cost and this like deflationary impulse response.  
[03:09:12] Yeah.  
[03:09:15] No.  
[03:09:27] Yeah, I didn't do that.  
[03:09:28] I don't know if we've done that in the past, like looked at that.  
[03:09:34] We have the adjustment costs in there.  
[03:09:35] That's the other thing in there.  
[03:09:36] Yeah.  
[03:09:42] That's about monetary policy about how monetary.  
[03:09:44] Yeah, it's not choking it off completely in that moment.  
[03:09:48] And then you see a little bit of it.  
[03:09:49] Like I don't know if you're like the Barrow Keen kind of thing is like  
[03:09:52] that's the traditional.  
[03:09:54] You see a little bit of it, but it doesn't.  
[03:09:57] It's not though.  
[03:10:01] Yeah.  
[03:10:02] So right now, everything's the same.  
[03:10:04] That's something like later we might go explore more.  
[03:10:07] But today it's the same.  
[03:10:08] Alas, this, yeah, I got it.  
[03:10:11] Okay.  
[03:10:12] So here's the other thing is like if you, I mean, I'm still learning this  
[03:10:15] by the way.  
[03:10:16] So like, but one of the interesting things about the tariff on investment is  
[03:10:20] it actually looks like this like marginal efficiency investment shock.  
[03:10:25] So if you look to like it's like just young premature tabalini, like they  
[03:10:29] made this big point about that as a driver of business cycles.  
[03:10:32] And then when you kind of look, you can see it in the equations, but here  
[03:10:35] you see it in the impulse responses.  
[03:10:38] We put a similarly scaled MEI shock lined up with the tariff shock  
[03:10:43] and they basically trace out each other one for one.  
[03:10:46] The only difference is actually consumption.  
[03:10:48] And that's makes sense because the tariffs being rebated to the consumer.  
[03:10:52] In this case where there's no tariff to be rebated here.  
[03:10:56] So this is kind of the point is that the MEI shock and that tariff shock  
[03:10:59] are moving like they're hitting the same exact kind of margin.  
[03:11:03] And then that's also shaping our story about when we think about the data.  
[03:11:06] Yeah.  
[03:11:30] You know, you can go ahead and just move on.  
[03:11:32] Like based on like the data I was showing you, we wanted something really  
[03:11:34] forceful on investment without spilling over like say how productivity  
[03:11:38] shock would do it.  
[03:11:39] And that was the choice of the MEI shock.  
[03:11:42] That's again kind of why we're thinking more about we might go to  
[03:11:44] sector model where we can cleave these things out.  
[03:11:47] But again, I think we do think it's interesting that like when you have  
[03:11:51] the tariff thing, the tariff shock, it does look like the MEI shock  
[03:11:57] and it does give rise to this deflationary impulse.  
[03:12:03] So I got like a not too much time surprisingly.  
[03:12:07] So here's the quite we'll whip through some pictures about how this maps  
[03:12:11] the data.  
[03:12:12] So the way we're going to do this is we're actually going to get this  
[03:12:15] is like a filtering exercise.  
[03:12:16] So the way this filter is going to work is we're going to feed in  
[03:12:19] tear shocks, the actual tariffs that we see.  
[03:12:22] So at Q4 2024, we're going to announce to the agents in the model  
[03:12:27] the entire path of the tariffs.  
[03:12:29] So Biden does bad on the debate.  
[03:12:31] We kind of know Trump's going to get elected.  
[03:12:33] That's the idea.  
[03:12:35] What's important is they do need to know after, you know, what's  
[03:12:38] going to eventually go.  
[03:12:39] So we're going to have persistence of point nine five.  
[03:12:41] So eventually tariffs are going to go away with some persistence.  
[03:12:45] Then we're going to have these two shocks.  
[03:12:46] We're going to have the MEI shocks and the and the UIP shocks.  
[03:12:50] These are going to be like repeat unanticipated shocks with a  
[03:12:54] perceived persistence of again point nine five.  
[03:12:56] So it's like I wake up, there's any me I shock.  
[03:12:59] I think it's going to drift down.  
[03:13:00] I go the next period.  
[03:13:01] Another me I shock is going to drift down.  
[03:13:04] How am I going to discipline those shocks?  
[03:13:06] Well, what I'm going to do is I'm going to pick the sequence  
[03:13:09] of shocks to hit the aggregate investment path.  
[03:13:12] Okay.  
[03:13:12] So the invest, so I'm going to get investment right.  
[03:13:15] Same idea with the UIP shock is I want to get the path of the real  
[03:13:19] exchange rate to be the same thing that I see in the data.  
[03:13:22] And then once I do this, I'm just going to ask some questions  
[03:13:24] about like what drove macro aggregates, what accounts for the  
[03:13:28] dynamics of trade and so forth.  
[03:13:31] So this is like kind of what comes out of this or like what  
[03:13:34] we fed into the model.  
[03:13:35] So these are the same tariff series.  
[03:13:37] This is the investment series.  
[03:13:39] So we're exactly matching it.  
[03:13:40] And then this is the real exchange rate.  
[03:13:42] It's flipped here.  
[03:13:43] So exchange rate, real exchange rate appreciates and then  
[03:13:46] depreciates over that time period.  
[03:13:48] So those are the forcing variables that we're putting into this.  
[03:13:52] And this is the black area.  
[03:13:53] So nothing else on this is like is basically has been targeted.  
[03:13:58] I mean, we have I but okay.  
[03:14:01] GDP, we get a boom in GDP.  
[03:14:03] That's basically equivalent to what we see in the data.  
[03:14:06] We kind of miss this like drop at the first quarter, 2025.  
[03:14:09] But overall, that looks pretty good.  
[03:14:11] Consumption declines, imports.  
[03:14:15] Again, that's kind of like the main thing.  
[03:14:16] We want to make sure we like could we replicate the path of  
[03:14:20] imports that we saw in the data.  
[03:14:22] We basically get it.  
[03:14:22] We kind of undershoot and I'll come back to this in a second.  
[03:14:26] We even get like net exports over GDP.  
[03:14:28] Looks pretty good.  
[03:14:30] And then the one kind of thing I would say kind of miss is  
[03:14:33] that so this is relative to inflation.  
[03:14:37] We're kind of like too soon on inflation relative to what we  
[03:14:40] see to the data that actually like reminds me like your guys  
[03:14:43] is paper like we have the network structure where it delays stuff.  
[03:14:47] So that's like compliments us so well, but that's kind of how it works.  
[03:14:50] I think the other big miss that we do is we do struggle with  
[03:14:54] individual trade components in a way.  
[03:14:56] And that's why we we're leaning towards going to the two sector model  
[03:15:00] is you know, we kind of like we don't get the collapse  
[03:15:04] in consumer goods as much intermediate looks good,  
[03:15:08] but capital goods is is we're undershooting it.  
[03:15:11] So there's a sense in the trade data.  
[03:15:13] We don't completely like get that whole component.  
[03:15:15] Again, we have fixes for this.  
[03:15:17] We haven't like highlighted it, but but that's a little.  
[03:15:24] Yeah, it's just the share.  
[03:15:25] And that's like maybe George is going to sit.  
[03:15:27] Yes, I think because it seems like within capital or investment goods.  
[03:15:31] There's like different shares as well.  
[03:15:33] And those are moving as well.  
[03:15:35] So we filter this through and then we ask like, well,  
[03:15:38] what are these contributions to this?  
[03:15:39] And again, like think about this like big question is like, well,  
[03:15:42] what did we get wrong?  
[03:15:43] What were we missing?  
[03:15:44] Right. So like when we look at GDP, so remember,  
[03:15:47] we're basically getting the path of GDP roughly right.  
[03:15:50] And then we can decompose it into these different shocks.  
[03:15:53] And then the first thing is the so we have the different shocks  
[03:15:56] are colored tariffs are blue.  
[03:15:58] Me is the red.  
[03:15:59] The UIP is green.  
[03:16:01] I don't know if anyone else is like color blind.  
[03:16:03] That's in the bank.  
[03:16:04] That's a problem sometimes.  
[03:16:05] But if it's gray, that's OK.  
[03:16:08] So but like the thing that comes out super clearly is that  
[03:16:12] GVs like completely been being driven by this I mean, I shock.  
[03:16:15] And then like seven and walked out.  
[03:16:17] But like back to her question is that I'm not is the question is like,  
[03:16:21] well, if we hadn't had the I mean, I shocks, what would have happened?  
[03:16:25] Right. Modulo Kevin has its like commentary on Fox and friends.  
[03:16:30] Well, the tear shocks with it.  
[03:16:32] This is almost like 60 to 70 basis points of GDP.  
[03:16:35] So this is actually a fairly sizable like decline in GDP,  
[03:16:39] especially relative to kind of standard trade models coming from from from the tariff.  
[03:16:45] Again, you can see here is like inflation.  
[03:16:48] Where is that all about?  
[03:16:49] That's all about basically being driven in in in the tariffs.  
[03:16:53] You know, investment is driven by the I shock, despite the fact there's a drag  
[03:16:58] coming from tariffs.  
[03:16:59] So even though we had a relatively smaller tariff back to your point,  
[03:17:02] we had a relatively smaller tariff on investment goods.  
[03:17:06] Yeah, we're just like cranking through it, given how productive investments are  
[03:17:10] right now and then consumption falls in response to this.  
[03:17:15] Let's see what. Yeah.  
[03:17:37] Because you're saying the I shock is just move in inflation up a bit.  
[03:17:41] You think, yeah, because this is a part.  
[03:17:43] No, that's why it's like in my head, it's like  
[03:17:46] positive in the I shock should create inflation.  
[03:17:50] Yeah, yeah, yeah. And that's consistent with that.  
[03:17:53] Very small. OK. Yeah. Yeah.  
[03:18:11] Dynamics to trade here.  
[03:18:12] So what do we see is, you know, what's the driving like imports?  
[03:18:17] Well, again, the key kind of thing is like, well, there's a fight  
[03:18:20] between the tariffs and the I shock.  
[03:18:21] And that's what's basically propping everything up.  
[03:18:24] And this becomes very transparent when you look at break it down by consumer  
[03:18:28] goods, capital goods, intermediate.  
[03:18:30] It's like the tariff just kills that in the I shock props up capital goods.  
[03:18:35] And then here you kind of see the funny thing is like the tariffs  
[03:18:38] actually increase imports in that first quarter.  
[03:18:41] Well, because that's the inventory dynamics right there.  
[03:18:49] Well, if we hadn't had this in me I shock or this A.I.  
[03:18:52] Investment boom, what would have been the impact is you can this breaks that down.  
[03:18:56] So this is just focused on the tariff component and then shading it into  
[03:19:00] different like consumption, intermediates and investment.  
[03:19:05] And so what's really important here.  
[03:19:07] And again, this like really kind of shaped my change my mind  
[03:19:09] when I thought about trade models is like how important investment is here  
[03:19:14] is, you know, the collapse in GDP attributed to tariffs is almost  
[03:19:18] completely coming all through like the tariffs relative.  
[03:19:21] And remember, that's a relatively small tariff, only about like five, six percent.  
[03:19:26] But it's just like hitting back to Mark's point, a large base.  
[03:19:29] And then it feeds into where we get those like neoclassical dynamics,  
[03:19:33] the flow versus stock.  
[03:19:34] And so that gets a large decline in output right here in response to the  
[03:19:39] investment tariffs.  
[03:19:40] And then back to Fabrizio's point, we're making this thing is like here.  
[03:19:43] You see it's relatively small, the M.E.I.  
[03:19:46] component from the investment in tariffs, but it's right there.  
[03:19:50] OK, so I got the number one.  
[03:19:52] So let's see what we got here left.  
[03:19:57] But this is kind of the stock of everything I've shown you.  
[03:20:00] And then then I'll wrap up with one more point.  
[03:20:02] We got the like can match untargeted pass, you know,  
[03:20:07] tariffs alone or stagflationary.  
[03:20:09] There was the doom and gloom.  
[03:20:11] But, you know, that point that investment move offsets this and then and then so forth.  
[03:20:17] The thing is, it's like moving forward and I'll wrap up with this slide.  
[03:20:20] It's like I think I mean, this is kind of Andrea echoing, you know,  
[03:20:24] we've been talking about this.  
[03:20:25] We've seen like policymakers.  
[03:20:28] I was at a conference in Stanford.  
[03:20:29] I think Arvin was there.  
[03:20:31] It's like, you know, they kind of stood up there and was like, hey,  
[03:20:34] everything was like, OK, the US economy is resilient.  
[03:20:37] And then it's worth kind of thinking through is like, well,  
[03:20:40] what's the underlying policy and what's the shocks hitting the US economy  
[03:20:44] and separating those two things.  
[03:20:46] And we think that's really important is because if the shock dissipates,  
[03:20:51] if this AI boom collapses, then we're left with these negative effects of the policy.  
[03:20:56] And I think that's really important for us to kind of keep in mind.  
[03:20:59] And then like I said, it's like, well, what how should a monetary policy  
[03:21:03] authority respond to a policy like, you know, the stuff that that president  
[03:21:07] put Trump president Trump put on in place during liberation day.  
[03:21:10] And then, you know, we kind of illustrated some point.  
[03:21:13] It's not just simple as like it's just a straight supply shock.  
[03:21:15] But, you know, how you might want to respond to it is different.  
[03:21:19] So I'll wrap up there.  
[03:21:20] And I look forward to seeing George's comments.  
[03:21:23] Yeah. Yeah.  

## Discussion (03:21:30 – 03:38:55)

[03:21:51] Thank you.  
[03:22:15] So.  
[03:22:18] So.  
[03:22:30] Love that.  
[03:22:31] Inventories love that.  
[03:22:32] Don't like any way of which that they did these things.  
[03:22:34] But I love that they're thinking about those.  
[03:22:36] There's trade composition.  
[03:22:37] There's this really nice old paper by Mark Tenbell.  
[03:22:39] You are a couple of papers that thought about like trade composition and business  
[03:22:43] cycle fluctuations.  
[03:22:44] More people should know that work.  
[03:22:46] Then things go off the rails.  
[03:22:47] There's sticky prices, sticky wages, dominant currency pricing.  
[03:22:51] Forget about habit investment adjustment costs.  
[03:22:53] Then we got tariffs, MEI shocks, UIP shocks.  
[03:22:56] There's a part that does a six quarter fitting.  
[03:22:59] Then there's something with 200 years of data.  
[03:23:00] There's just a lot in this paper.  
[03:23:02] Well, huh?  
[03:23:05] Yeah. Yeah.  
[03:23:05] So AI didn't correct this.  
[03:23:07] I should get Mike's LM.  
[03:23:09] All right.  
[03:23:10] So, but like when you take a step back, what does the paper have?  
[03:23:14] It has a clean, intuitive accounting of recent trade, right?  
[03:23:18] And it's consistent with everything we learned about the great trade collapse 15 years ago.  
[03:23:23] So what do we learn from the great trade collapse?  
[03:23:26] The first thing was like inventories were super important.  
[03:23:29] OK. Just without a doubt.  
[03:23:31] Here it's kind of working the other way in the great trade collapse.  
[03:23:35] Inventories are magnifying the decline in trade.  
[03:23:39] Here they're kind of like reducing the decline in trade in some way because of their kind  
[03:23:43] of stockpile into future shocks.  
[03:23:45] And the second thing, there's like a really important role for the composition of demand.  
[03:23:49] OK. Those are kind of the two lessons from the great trade collapse and they kind of jump out through here.  
[03:23:55] All right.  
[03:23:56] So I'm going to focus on the part of the paper that says missing trade collapse because there's no mention of inflation.  
[03:24:01] And then the title and I only read the title.  
[03:24:03] So I'm going to kind of think about like the question, are the recent US trade dynamics really import dynamics?  
[03:24:09] Are they surprising?  
[03:24:10] I think this is what was keeping Mike up at night.  
[03:24:12] Right. And so I'm going to say no and yes.  
[03:24:17] OK. And for different reasons.  
[03:24:18] So I kind of think about that.  
[03:24:19] So let's just start out like everyone in this room knows this.  
[03:24:22] When you think about an Armington demand model, that's kind of the model we almost always use.  
[03:24:27] You can think about the value of the goods I buy from a particular source and being sent  
[03:24:33] to a destination in time t depends upon some relative prices and some total demand.  
[03:24:38] Right. So the amount of cars we buy from Japan depends upon the total number of cars  
[03:24:43] and what's the relative price of those cars.  
[03:24:45] And then that tariff and that trade costs are the things that are kind of leading us to substitute away.  
[03:24:50] And those states are telling how like the tariffs are a change in tariff and move us away.  
[03:24:56] So the aggregate data, I got to sum up across all goods and sources to get like some measure.  
[03:25:03] And if I want to do like changes over time, I've got to come up with like some weighted average of tariffs.  
[03:25:09] Mike has a weighted average. We're all comfortable with that.  
[03:25:12] I need some weighted relative prices and I need some measure of like weighted final demand.  
[03:25:16] Right. That's going to be the key thing.  
[03:25:18] And I think this is the part that like, you know, we can all kind of.  
[03:25:23] I think Mike, the whole point in this paper was you need to kind of do the waiting correctly.  
[03:25:27] Okay. So I'm going to, I'm going to kind of just take this equation.  
[03:25:31] This is the testing equation and there's there's kind of two ways of thinking about taking that equation to the data.  
[03:25:37] Okay. So the top way is like, hey, let's be, let's be macro economists.  
[03:25:41] Let's bring that measure of demand. Right.  
[03:25:43] So what are the purchases of Japanese cars relative to total cars?  
[03:25:47] Total imports relative to total spending on tradeable goods.  
[03:25:50] How does that depend upon like the tariff, which we saw went up a bunch and then some substitution margins.  
[03:25:55] That's like the macro way of doing this. And I think that's how these guys are approaching it.  
[03:25:59] There's a micro way of doing this, which is to basically go and like look at good level tariff variation.  
[03:26:05] And that good, those good level tariff variations are going to kind of like pull out the demand effects.  
[03:26:11] And then we kind of see the substitution. Okay.  
[03:26:13] But in the end, I'm going to be thinking about this is like, is there something funny about that data once that control for demand?  
[03:26:19] Is the, is the trade elasticity the response to tariffs? Is it moving in a weird way?  
[03:26:24] Okay. So let's, let's just start out like this is the picture that Mike showed.  
[03:26:28] I guess I didn't do any detrending. This is just raw deviations from 24 Q4.  
[03:26:35] And this is just raw. And I think, you know, people are looking at that final point and saying, look,  
[03:26:39] we put all these big tariffs and trade actually went up a year later. Okay.  
[03:26:44] Of course, my theory said I had to divide through vice and stuff.  
[03:26:48] So let's start dividing through by stuff. Well, actually, first thing you have to do is there was this weird thing about gold.  
[03:26:53] So I'm going to try to take the gold out, which again is completely inventory stockpiling, right?  
[03:26:58] So I'm going to take the gold out. That's going to make the surge a little bit smaller and the collapse a little bit smaller.  
[03:27:03] But then I want to divide by demand. So now this is like goods, final sales.  
[03:27:09] And Mike showed me this, this variable a long time ago and I always divide everything.  
[03:27:14] Now it's starting to look like actually there is a trade collapse, right?  
[03:27:17] Like we've kind of substituted away. So now I'm going to do what Mike did is he's going to tell me like, oh, and here's the tariffs, right?  
[03:27:24] I'm going to, I'm going to create like a different measure of final demand that kind of equally weights goods consumption,  
[03:27:31] equipment investments and ICT. That thing doesn't look like it's moving much beforehand.  
[03:27:37] So I'm pretty much, I think I've got like a pretty stable kind of measure and now I'm seeing like a bigger decline over time.  
[03:27:44] So we've substituted away from foreign goods. But the whole question is why is like this final demand thing kind of booming, right?  
[03:27:51] Okay, so we can convert this into elasticities. These are like the changes in trade relative of that trade measure relative to tariffs.  
[03:28:01] You know, everything's we're substituting away. If you just use the raw numbers that that elasticity flip sign at the end,  
[03:28:09] if you start controlling for demand, these things are kind of low close to one. If we do this other measure, it's a little bit higher,  
[03:28:15] but it's still kind of a low number. Okay.  
[03:28:18] All right. So, so what I when you talk about like, you know, is there should I expect to collapse? Well, I'm seeing a collapse.  
[03:28:25] People are substituting away, but we need to control for the demand part, right? Okay.  
[03:28:30] And then this is this is the thing that's driving everything might kind of showed this. This is real capital goods imports just kind of taken off, right?  
[03:28:37] I put in the total value of it at the end of the sample. It's like 1.3 trillion dollars, I guess, whatever. I don't even know what the numbers are.  
[03:28:45] This is real ICT investment, which is kind of boomed and look, it's half of real capital imports.  
[03:28:51] Some of this stuff is in real equipment. Okay.  
[03:29:01] So there is something to like, why are we buying so much of this, this ICT stuff.  
[03:29:07] Okay.  
[03:29:10] The other way that we could sort of look at these elasticities, and there's a paper by a con at all.  
[03:29:16] I'm like the final author of that, Alexandria. And it's called trade dynamics when tariffs aren't tariffs. And the idea is to look at these trade elasticities, the change in the value of the good source by the tariffs with these time fixed effects for sources and goods.  
[03:29:34] So you can think about the demand effect that Mike's talking about, which show up in that GT fixed effect. All right. So if we're if we're buying a lot of goods for ICT, we'd be buying it from all sources and then we'd be tilting away from sources.  
[03:29:45] We're tariffs went up relatively more. And so we want to kind of understand how does this theta move around? Are we substituting away more or less than usual?  
[03:29:53] All right. So this is what we see like before these tariffs went in place, the short run trade elasticities were around like 2.5, 2.6. Okay, so 10% tariff trade falls 25%.  
[03:30:07] When these tariffs went in place with Liberation Day, this trade elasticity fell here. I've got these as negatives, right? So kind of half as big. So it's kind of interesting.  
[03:30:17] So there is like missing. There is like a missing trade collapse. And it's basically to the substitution away from tariffs. Okay.  
[03:30:25] So what's the explanation for that? Well,  
[03:30:29] 2.32 tariffs, which like there was no possibility of refund in the IEPA tariffs, the trade elasticities are very different. So the IEPA ones which are completely refunded.  
[03:30:42] Those are the ones that are driving that kind of change in trade elasticities.  
[03:30:47] And so I think the thing we want to be thinking about is it firms are kind of responding to like tariffs in a different way because those tariffs weren't like fully expected to be real. They were going to get refunded.  
[03:30:59] Then some of the results should be revisited. So I will say that if you do like a, a media wine scene kind of like pass through regressions, you also find the products that we're getting those IEPA tariffs, there's less pass through into producer prices.  
[03:31:13] So, okay. So now I want to add like inventories into the model. Right. So I gave you an import demand equation. And now once you add inventories, those change in imports are going to depend upon my, like again, think about it.  
[03:31:26] We're importing Japanese cars. The amount of Japanese cars are importing and depend upon how many Japanese cars I sell and how whether I sold them out of inventories or whether I'm building out my inventory.  
[03:31:36] Right. That's going to be the way we're going to be adding inventories into the model.  
[03:31:52] Inventories are more important in international trade than domestic trade. Part of that is because there's larger fixed costs, longer riskier delivery shipping times.  
[03:32:02] So when you actually look at the data, it looks like there's roughly five months of imported inventories relative to sales compared to, for imports relative to domestic goods.  
[03:32:11] So inventories are actually the way they've calibrated the model. Inventories are pretty, pretty small in their model. And the data, they're much bigger.  
[03:32:17] Part of it's because inventories get held in three different places. So when you do like inventories over final sales, it's a much bigger number, bigger stock.  
[03:32:26] Well, they kind of give you like some something that looks like a forward looking pricing decision, right? They, they change your pricing and ordering ordering, right?  
[03:32:35] You're going to be thinking about like pricing over the discounted marginal cost of a future, the replacement value of the inventory. So what's the marginal value of inventory?  
[03:32:45] So you're thinking about doing sell today or they sell tomorrow. So that's where like future tariffs are coming in.  
[03:32:52] There's also going to start leading to stockouts. So some, some goods like the price that we see you can't buy at. So like the CPI is going to be mismeasured.  
[03:33:00] That was an issue with COVID. And then the SS ordering is going to depend upon like actually the distribution of ideas and credit shocks and inventories. Okay.  
[03:33:08] We'll get incomplete pass through and the distribution of inventories are going to matter.  
[03:33:12] So,  
[03:33:15] the first thing I have here is pretty unusual. It's not SS it's not a stock out of ordinance and input is more valuable. If like, I saw the way I think about it. If I if we put a gallon of gas and for BCOs car, that gallon is more valuable for BCO if there's 100 gallons  
[03:33:35] at the gas tank of the at the at the gas company, then if there's he got the last gallon, it's kind of a weird thing. And then there's also like a bliss point that these guys have for for for inventories, which given the types of shocks we're thinking about that are reorganizing global supply chains, you might want to think about like letting  
[03:33:54] that bliss point be in dodging this in some way. Okay.  
[03:33:57] So let me kind of just jump ahead because I don't want to I don't want to talk about that stuff.  
[03:34:02] So, so now I'm going to kind of drill down in the micro data and I think for Mike showed like when tariffs are going to go up and stock up and then stock down right because I still want to think about elasticity.  
[03:34:14] So this is splitting up the food and beverages industrial supplies, capital goods automotive parts consumer goods.  
[03:34:20] You see for every good kind of a boom bust boom bust, but it doesn't quite have that inventory flavor which was boom over shooting and then coming back.  
[03:34:31] So if you start thinking about tariff elastic, is there's something or there's something a little bit funky and taking the model to the data, but you should really be thinking about inventories in every sector.  
[03:34:40] Okay. This is what this was going to say. All right.  
[03:34:43] Two last points.  
[03:34:47] We're going to invest in this net exports.  
[03:34:50] It's crazy how these things are kind of moving in opposite directions. This sort of tells me if we're thinking about like flows and net trade flows, we should be thinking about models with inventories.  
[03:34:59] This, if I go to like a longer series, this correlations like zoom in up here with the tariffs going up, whether it's big money.  
[03:35:12] It turns out in some of the natural models, the way you model inventories, whether it shows up in like stockpiling of imports or like relative trade flows matters in the way that you model the inventories and I think here it's showing up mostly in imports instead of like net exports.  
[03:35:28] I mean, it also shows up in that exports.  
[03:35:30] All right. The last thing I think I want to say is this is me being Kevin Hassett.  
[03:35:39] But paper is thinking about and I had to sort of dig in there because there's no like a household budget constraint in the paper.  
[03:35:46] Mike was very lean on like equations today.  
[03:35:49] But in this model, we're putting these tariffs in place and there's lump sum rebates of it.  
[03:35:56] We could think about taking those tariffs and like subsidizing investment.  
[03:35:59] I mean, there was a there was a tax bill that went through that was doing doing things.  
[03:36:07] So I think it's useful to start thinking about this is like the Charles Engel point, right?  
[03:36:11] If we think about this tax reform is doing something special for investment and that's being financed by tariff revenue.  
[03:36:19] This is probably the way we want to do it in terms of doing these kinds of analyses.  
[03:36:23] One more thing like, I mean, this model has enormous inefficiencies because of the markups, right?  
[03:36:30] And double marginalization.  
[03:36:31] And so this is a really inefficient model.  
[03:36:33] I don't know if you guys did the usual trick where you kind of make it efficient with like subsets in the background.  
[03:36:38] So you if you're going to put an extra tax, you might want to do something with the revenue.  
[03:36:43] So what I did is I got rid of the sticky prices.  
[03:36:45] I did this poll. I did their policy.  
[03:36:48] I calculated the welfare gain.  
[03:36:50] If we did their policy and like did like the investment subsidy and it's about 1% of US welfare.  
[03:36:58] If I do the lump sum, the US is worse off by about 0.15%.  
[03:37:02] So it's very, very different.  
[03:37:03] But we can look at the transitions.  
[03:37:08] It's like a path that might gave me for tariffs.  
[03:37:11] I'm going to keep them fixed because I actually think they'll probably stay fixed.  
[03:37:15] And then I've got one where the dashed line is like where I'm doing an investment subsidy with their tariff revenue.  
[03:37:21] And the other is I'm doing the lump sum.  
[03:37:23] So you can see for BDSO, this is the stock piling.  
[03:37:26] Sorry, not the stock piling.  
[03:37:27] This is the investment boom at a time, right, that you were asking about and then stuff is falling off.  
[03:37:32] But then if I'm if I'm subsidizing investment, we get a big kick in investment, which it could be there any any eye shock.  
[03:37:39] Obviously, I've got to finance that with reduced consumption.  
[03:37:43] And this is actually very expansionary for the economy.  
[03:37:46] So it's a completely different way of thinking about what's going on.  
[03:37:50] I think probably some like the truth is someplace in between.  
[03:37:54] I think the thing that's, you know, important is to think if we think about this as subsidizing all investment, then we shouldn't see it just be showing up as investment in AI, right.  
[03:38:07] When you dig into it, it looks like aspects of the bill would favor AI over non AI, but you know, I think that's like an open question.  
[03:38:16] So let me, I'm out of time.  
[03:38:18] Yeah, let me just show like one last picture.  
[03:38:21] I was looking at inflation.  
[03:38:24] This is real inventory goods over final sales.  
[03:38:26] This is like the COVID period where inventories like fell sharply.  
[03:38:30] Right.  
[03:38:31] And then this is when inflation took off.  
[03:38:33] Here's the kind of boom and inventories in the bus.  
[03:38:36] I don't know what's going to happen next.  
[03:38:38] So I think it's useful. I think it would be useful to kind of take your model and like bring it to this period before we start also looking at at this period.  
[03:38:47] But yeah, so let me just stop there. Thank you.  

## Q&A (03:38:55 – 03:41:30)

[03:39:14] Oh, yeah, super asking.  
[03:39:19] It just took a few days for me to set up.  
[03:39:23] Yeah.  
[03:39:25] I went to school.  
[03:39:27] I learned some extra sciences on the market.  
[03:39:32] I'm sorry.  
[03:39:35] It was very embarrassed.  
[03:39:37] The whole time.  
[03:39:40] The aggregate demanded that.  
[03:39:42] It was costly.  
[03:39:44] That's what I say.  
[03:39:46] Yeah.  
[03:39:50] Yeah.  
[03:39:52] Yeah.  
[03:39:54] Yeah.  
[03:40:02] Yeah.