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

# Population Aging and the Realignment of World Trade

Authors: Joseph Kopecky  
Discussant: None  
Video: https://www.youtube.com/watch?v=w6ObGDwelyg&t=10718s


## Talk (02:58:38 – 03:32:28)

[02:58:38] Looks perfect. Thank you. That working uh properly.  
[02:58:42] >> Hold on. I need to uh Yep. Okay, you can see the slides properly.  
[02:58:48] For some reason, my other screen changed over to a There we go.  
[02:58:52] >> Yeah, now we can see. Good. Thanks. >> Perfect. Thank you.  
[02:58:56] >> So, maybe wait move this >> another two minutes or so.  
[02:58:59] >> Okay, great. >> Wait. Yeah.  
[02:59:06] >> Sorry, just rearranging my screens now. All good.  
[03:00:38] Okay, so we're just on time. Um, so welcome back everybody. I'm going to be moderating for the next couple of sessions. Um, just a reminder for those of you who are joining us, each  
[03:00:50] presenter gets 30 minutes a presentation including five minutes at the beginning with no questions. Uh, and then clarifying questions are allowed. Um and at the end we have 10 minutes of broader discussion. So our next presenter is Joe  
[03:01:05] Copeki telling us about population aging and realignment of world trade. Take it away.  
[03:01:10] >> Great. Thank you so much. I hope you can hear me. All right. Um thanks so much for including this paper. Really just an incredible lineup of uh papers and presenters. So I'm very excited to to hear what what you all think. So uh yeah, this paper is about uh population  
[03:01:24] aging and trade. And let me be in the right place. There we go. And so, you know, we know workforces are aging and importantly for this paper uh in particular, they're aging at different rates around the world. And so, the sort  
[03:01:39] of big bold question is how this is going to affect trade. And so, not surprising to anyone who's been watching today, of course, we've got kind of workforces aging. And so the left panel here is kind of normalizing to 2014  
[03:01:52] looking at some of these major economies how their workforce size uh is going to change. So this is kind of the working age population uh change and you know there's a few countries where it's getting bigger, quite a lot of countries where it's getting smaller. Most of  
[03:02:07] these have peaked kind of relatively recently uh you know with the boomers aging out and of course these kind of longer term shifts that we've been talking about a lot today. And then of course the right panel is actually the aging of that workforce itself. And these are kind of two different uh  
[03:02:21] trends that are going to be at the center of all the results I'm going to show you is one this kind of raw size of the workforce, this input uh that's going to go into production and trade.  
[03:02:31] Uh and then one about this tilting of that workforce itself. Uh so how that workforce composition is changing. And so just to kind of briefly summarize the channels and kind of the implications that I'm going to be talking about this  
[03:02:45] workforce size that left picture uh is going to turn out to be really important for the volume of trade how much countries are trading with each other.  
[03:02:53] Um and that age skill mix is going to have some implications for the volume but really is going to be more about the composition of what's being traded. Uh and so what I'm going to try to show you today is some model results that try to back out how big those things are. Uh,  
[03:03:07] and then I'm also going to show you an extension that builds in in a slightly stylized way an age wealth profile that allows for sort of non-balanced trade by having these net foreign asset positions  
[03:03:20] be non zero. Uh, that's going to matter. It's going to be able to replicate some sort of sensible ways of thinking about uh these external balances uh but isn't going to turn out to be actually all that important for the first two points.  
[03:03:33] And so I I wanted to do that largely because I wanted to know how important that might be on a first order way for things like the volume of trade and the composition. Uh but we'll see that and I'll give a little more details when we get there. So what am I doing here? Uh  
[03:03:47] constructing multis sector eaten cortum style trade model. Uh and really this is just a caliando pero uh version of this ricardian eaten cortum trade where we've got 30 economies uh 29 and one rest of world 20 tradable goods and one  
[03:04:02] non-tradable service good uh input output linkages between these different sectors going to be calibrated to this world import output database uh and what's new what goes beyond that caliando pero framework is that there's  
[03:04:16] going to be age dependent skills so we have homogeneous labor being replaced in that model with three age dependent task inputs. So a country's age structure is going to exogenously shift here how much  
[03:04:31] of that endowment of labor it has in these kind of age specific skill inputs.  
[03:04:36] uh going to do counterfactuals with this exact hat algebra feed the age structure shoe through and be able to show you uh a comparison from a fixed kind of 2014 economy where I've got kind of the last  
[03:04:48] year of my input output data and uh and 2050. I mean I do a lot of horizons but these are going to be the comparisons I show you a lot uh today. So everything else is kind of fixed. There's a lot going on kind of in this time series.  
[03:05:01] you know, a lot of uh policy things, you know, technology things, trade costs, whatever. Uh just trying to back out this demographic component cleanly. So, not really trying to let any of that move around. So, there's a huge literature uh that this comes from.  
[03:05:15] Really, the main thing that you'll want to know and that if you haven't thought about kind of aging and trade maybe would be new are these first two kind of bullets. Uh and really the first one is the one that motivates the paper. Um so  
[03:05:27] this Kai Stoyenov and Sugtoyenov uh papers looking at this age skill mix and how having different age structure also is going to affect your comparative advantage based on uh a literature of uh  
[03:05:41] skills in occupations, skills that are going to be sort of increasing with age, skills that are going to be decreasing with age and then trying to figure out how sectors use those occupation skills together. uh and so those are empirical papers that try to actually estimate uh  
[03:05:56] how big these things are and I'm taking their approach and feeding it into this input in in my model. But of course there's also a huge literature that I couldn't possibly get to everything on today about you know external balances and saving and aging and all these  
[03:06:11] things. Uh and then this quantitative trade literature which is where the kind of structure of the model uh builds from. Okay. So just to preview the findings really quickly in case I don't manage my time well uh you know large  
[03:06:24] changes in volumes uh coming from aging and these are mostly going to be driven as I already kind of alluded by labor force size. So the fact that you just have more of this endowment this sectoral almost type of mechanism uh  
[03:06:36] that is going to affect uh how much countries trade with each other. And so if you look at exports to the US from these major trading partners, the model's going to predict a big decline from countries like China, Japan, Korea,  
[03:06:50] and a big increase in countries where they've got this relative increase in endowment in youth, India and Mexico, at least relative to the rest of the world.  
[03:06:58] And then what I think is interesting the composition of trade, you know, skill mix is going to be it's going to have a non-trivial, what I would say is kind of small but really, you know, non-trivial effect on the volumes, 2% of these pretty big things. uh is going to come  
[03:07:12] from this change in composition skill mix. But where it's really going to have a huge effect is on the composition. So what's traded. So tilting the sectors themselves in terms of what's going in and out of these uh trades uh trade relationships. I'm going to show you  
[03:07:26] things about a little bit about external balances and some welfare effects uh and hopefully caveat those uh when I need to. Uh the welfare effects ones especially, they're small but kind of meaningful. They're, you know, on the same kind of scale of the ones that you  
[03:07:38] might see in uh, you know, trade kind of uh, you know, NAFTA type policy effects.  
[03:07:45] Uh, but they're really small compared to the model also spits out welfare numbers that are just huge and those are mostly coming from domestic margins. So, of course, if your labor force is shrinking a lot, that has maybe much larger welfare effects on a domestic margin than this kind of trade balance effect.  
[03:07:59] So, I'll talk a little bit about those. I think there's some interesting future work to try to really understand those better. Um, so getting into the mechanisms and the model itself. So, uh, I'm not going to spend too much time on the model. There's a lot of kind of  
[03:08:13] machinery there, but hopefully give you a picture of what's there and what I change so you can understand how that maps into the results that I'm going to show you. As I mentioned, there's n economies. So, n countries, 29 plus arrested world, 20 tradable goods, one  
[03:08:26] non-tradable, three age dependent task inputs, which I'm going to tell you a little bit more about on the next slide.  
[03:08:31] But they're appreciating cognitive, depreciating cognitive, and physical.  
[03:08:36] The labor supplied inelastically uh and the age distribution sets kind of how much of each of these a country has. Uh firms are going to have a continuum of varieties within each sector. uh there's going to be Cobb Douglas uh production  
[03:08:49] of these three labor inputs uh as well as intermediaries from the other sector for SHA productivity draws iceberg trade costs all the standard stuff in these quantitative trade uh literature and here services can't be traded I run some extensions where I allow services to be  
[03:09:03] partially tradable um uh as well and I'll show you those uh households be just one representative Cobb Douglas uh which has preferences over all of these sectors uh with fixed expenditure shares  
[03:09:15] uh uh income from factor payments. Uh and then sequence of static equilibria.  
[03:09:20] So this is a the stat these trade models are static. I'll talk a little bit about kind of the trade-off that that makes um looking at counterfactuals. Uh so what are these what's the change that I'm doing relative to the quantitative trade  
[03:09:35] literature uh that existed before. So uh and this is really coming from this Kai Stov paper this GIE paper uh which looked at this cognitive aging literature uh trying to understand uh  
[03:09:48] the how these different skills uh vary with age and then how that's going to then map into changing endowment of skills in the economy. So essentially I do what they do is I take these ONET  
[03:10:01] ability uh scores that they use uh which map into occupations uh build these profiles. Uh I initially set out with these three different sets because you can't really distinguish very well on the data between a  
[03:10:14] depreciating cognitive and physical. I end up giving those the same which is more or less the same as collapsing it to two uh and reweing everything I suppose but I leave it as three uh because I do run some counterfactuals where I kind of let those vary a little bit such that we can see a difference  
[03:10:29] between them in the data uh but there's not a lot there uh and of course if you just kind of shut this down and make all three profiles the same then of course demographics ceases to have uh an important kind of margin here in terms of affecting this kind of skill level.  
[03:10:42] It'll still of course have if you've got a bigger or smaller population some some aggregate effect. So those on net occupation scores get uh interacted with sort of the share of those occupations in these sectors uh and I end up with these cost shares which are what  
[03:10:55] actually get fed into the model which are you know normalized to one across these uh sectors uh and then just uh show kind of the share of appreciating skill within a particular sector. Right?  
[03:11:09] So you've got computers and electronics with really high skills and things like wood, mining, uh basic metals with a very low share of these age appreciating skills. Um so what really matters is a ranking of these intensities which is  
[03:11:24] what the data is pinning down in my calibration. Um the cardinal spread matters a lot for thinking about kind of welfare and magnitudes but all the effects kind of go through if you believe in this ranking so to speak or at least directionally uh they go  
[03:11:36] through. Um, okay. So, just briefly on sort of the framework. I mean, uh, this is, uh, fairly standard kind of version.  
[03:11:46] It's really just taking Kellyo uh, straight off the shelf and changing this one, uh, bit where production instead of having a single unit of labor in this first equation is now going to have, uh, these three skill dependent labors. So  
[03:12:01] it's got that alpha J K which is a you know the uh previous slide oops uh these were the cost shares these alpha JK so these sector J's share of a particular  
[03:12:12] skill K uh their Cobb Douglas weight in this production function and so you're going to have this unit cost measure which is going to build up from those endowments of age specific skills and then of course from intermediate inputs  
[03:12:27] which is the second P term and then they get a Feet productivity draw uh that row term in that second equation is something that uh I don't have a great mapping to in the data but it turns out to not matter very much uh but this is essentially a tilting that allows that  
[03:12:41] if you have sort of an advantage in a particular skill you can become better at those skills and draw more productive draws over time and so I run some robustness on that uh for my main result it doesn't really matter very much and then the reason you want to use one of  
[03:12:55] these kind of eaten court style models is that it spits out a gravity equation which is that essentially the third equation we get there. Um and so this is what I'm building the model from. And  
[03:13:08] then I take this exact tat algebra uh that allows you to sort of simply calibrate this model off changes. Uh you would need a lot more information to be able to estimate this on levels but a lot of the those things difference out uh in this exact hat algebra which  
[03:13:22] allows you to do these counterfactuals in a really clean way. Uh and so that's what I'm going to be estimating to get these trade flows uh uh and these kind of general equilibrium changes. So again, static equilibria uh I'm going to  
[03:13:35] compare 2014 to 2050 in I think everything I show you. Uh and there's two margins of the shock essentially. So you've got um all three inputs are going to move by their geometric mean. And I'm going to call that kind of the scale shock, which is how much of your  
[03:13:50] endowment actually moves. Of course, they're all moving slightly differently.  
[03:13:54] uh and that's kind of be the pure labor force share of aging and then back out from that this uh compositional skill effect uh for uh how big the compositional mix matters when it comes  
[03:14:06] to things like uh trade flows um calibration I won't spend too much time on this because I don't want to go over time but pulling this input output database most of the sort of trade stuff just rests on kind of the standard literature stuff the onet stuff I told  
[03:14:21] you about which all comes from uh and and this this larger literature uh that that uh the Kai Stov paper uh built to sort of bring that towards trade uh stuff. So where does that shock that  
[03:14:36] I've been telling you about the demographic shock that I'm feeding in actually look like? Well, if we look at the size of the workforce, uh uh there's going to be some countries that are having kind of big increases in their size of their workforce and many that are going to be shrinking. Uh, and then  
[03:14:50] the skill mix is going to be the thing that, uh, backs out from there. And I'll show some interesting things. Part part of what matters with the skill mix, too, is how you're aging relative to the rest of the world. So, um, some countries that are very, very old but aged a lot  
[03:15:04] faster and have now kind of stopped moving are not seeing their skill mix move by very much. Uh, you know, so like Japan and Italy. uh whereas some are that are more rapidly aging are going to see uh it move a lot more and some are  
[03:15:18] going to become relatively much younger. Um so uh I think that's the sort of what I'm doing and maybe it's a good point to stop. I haven't been looking at the chat at all so I don't know if there's chat questions but uh if anyone has any  
[03:15:31] questions on sort of understanding what I'm doing I can take them or I can move right into the results. Um okay well jump in if uh if uh if you if you want. I'm monitoring the chat for you, Jose. So, >> okay. Thank you. I appreciate that. Uh,  
[03:15:46] okay. So, the first thing I kind of wanted to look at, um, and this this comes a little bit from some gravity equation, pure kind of reduced form stuff that I had done in an earlier paper was trying to understand how much of this aging can actually affect the  
[03:15:59] China US trade relationship. Because if you just plot uh the top graph there which is the relative working age population share. So kind of if you look at the share of working age in China versus the working share in the US you  
[03:16:13] see this pattern that if you're thinking about trade feels like oh well it's China's becoming much more relatively young right at the point where you know it's really starting to come out and and become this big uh dominant uh force in world trade. uh and then of course you  
[03:16:28] know that peaks and moves around a bit uh and then is is going to turn from tailwind to headwind. So the top one is just that it's these relative work forces. I could probably also uh just plot a few different things that would give the same flavor of an idea. And  
[03:16:41] then the bottom one shows the model both the historical kind of predictions of the model as well as the ones that use projected uh data in terms of how the model would predict that trade relationship to move. uh and then the gray line is just this actual uh trade  
[03:16:55] movement uh which you know the the model predicts very big uh effects of demographics but of course we know that the change in trade between China and the United States was even was huge I mean even bigger than that so I mean one  
[03:17:09] thing I should say is I'm not trying to fit the history but I want to understand how much of not only future trade uh maybe was affected by will be affected by demographic tailwinds or headwinds but also maybe how much past trade uh  
[03:17:23] has benefited from demographic uh tailwinds behind it for countries like the US and China or corridors like the US and China I should say. Okay. Uh so you can run this kind of for a bunch of countries. I'm I often show things relative to the United States. So these  
[03:17:38] exports to the United States uh and uh just because these are some of the largest I'll also show some some import things and some other corridors of trade. uh but essentially what the model predicts then is that similarly you  
[03:17:51] would have expected demographics to be supporting a whole bunch of countries increased exports to the US uh over the period from 1950 uh to more or less today uh but for a lot of countries for that to to start to to reverse. So the  
[03:18:06] first headline results and this is a little bit of just using the model to do a very kind of sophisticated accounting exercise says that demographics uh has been a tailwind for trade and that that tailwind is going to maybe turn into a  
[03:18:18] headwind uh over the next 25 and more uh or so years. Um if we think about kind of getting below that into the skill composition I mean first thinking about volumes itself. Um  
[03:18:32] the the left is this change in bilateral manufacturing exports by US for some major corridors. So Ind India to US, Mexico to US, Indonesia to China. Um I mentioned these two things right the scale effect which is how much is coming  
[03:18:45] from just the workforce itself changing size and then this composition effect which is all about this skill composition comparative advantage. uh and the scale completely dominates. I mean the composition effects again, you know, these would be reasonably big numbers to to think about. Uh but  
[03:19:00] they're dwarfed by the ones that come out from uh from the composition side.  
[03:19:04] Uh so composition accounts for roughly 2% and it's never more than sort of a percentage point uh factor in terms of shifting these bilateral trade relationships. So the size is kind of the headline result and the skill mix is about what's inside it, which is what  
[03:19:17] comes next. And I won't do it, I think, for time, but I have this button. was curious if this could have accounted for any of these changes. You know, there's this there's this discussion of China plus one which is mostly about trade policy, right? Rerouting of things from policy. Uh you can't explain any of that  
[03:19:31] rerouting in a meaningful way from demographics. Uh you know, in terms of maybe that's part of why you see things going through some kind of third party intermediary uh uh instead of directly between say the US and China.  
[03:19:44] >> So it seems like composition just to I guess to clarify on this on this result of scale versus composition. So >> is is the if you take the version of the model where you have these flat kind of skill by age profiles  
[03:19:59] u where I guess then it no longer there's no longer this shifting comparative advantage it's just about total size then then you'd get zero composition effect and it's all scale and is that basically  
[03:20:13] >> how do I back out say it again yeah >> is I guess I'm asking is the scale basically just the data in the sense that if I just assume that how much I export is proportional to my working age population >> and then I feed in productions in  
[03:20:27] working age population. >> Yeah. Yeah.  
[03:20:30] >> Then that changes the mix of what I'm importing and that's is is that and is that what you mean by scale and and >> Yeah, more or less. It's not exactly the same. Um it's not that I I flatten those estimate the model and figure out what  
[03:20:43] that is. I probably should report that somewhere to see how different it is. My sense is that the way I back out composition is is similar to that in that I I take the uh workforce size uh  
[03:20:57] from these three different kind of inputs, you know, cognitive, appreciating, cognitive depreciating and physical and I kind of take the geometric mean of those and then and then back out how much of the trade comes from that and then the residual is the composition. I think that's actually  
[03:21:11] quite close to being just feeding in flat profiles and I should I should probably do that. Uh but yeah, and that's why I kind of say it's a it's a fancy accounting exercise in some way because I'm really just feeding those in. I think I think what gets interesting then is where composition does kind of matter which is underneath  
[03:21:26] the flow itself this changing uh distribution of stuff I guess and that's where the model I think uh adds I think some insight that wouldn't be immediately obvious from there but yeah it's a good question. Um so if you if  
[03:21:39] you kind of dig into a particular quarter so this is China to the US inside of that of course we've got all these different sectors um and at any given point in time they've got this kind of relationship for uh the intensity well you've got the  
[03:21:53] appreciating skill intensity alpha and these these fixed relationships with with how those work. Um and if you plot those different uh changes in these sectors, the the the circle size is the  
[03:22:06] size of the the sector itself, relative size. Um you can kind of see how these sectors are moving. And so again, I can kind of these things would move even if you held um if you held the skills fixed. If you put those flat profiles  
[03:22:21] through, there would be movement in these sectors, but they wouldn't be systematically related to these appreciating skill intensities, right?  
[03:22:28] They wouldn't have this kind of component that's related to your demographic kind of skills uh which always match pretty well uh our squared of I say.9 in the in the bullet there.  
[03:22:40] Whereas if you just kind of try to get these uh skill mix otherwise uh it would be kind of much more sort of scatter plotty. So uh again looking at these kind of shares and the change in the full shock. So you know all of these things if we look at China to the US  
[03:22:55] everything's kind of declining right the full shock tells us that there's going to be just a big decrease across the board here. Uh but it's going to be offset partially in some sectors where you've got a big boost. So electronics are going to have an appreciating skill  
[03:23:08] intensity. So that decline is going to be much less heavily felt in electronics relative to uh textiles say where that'll actually the skill effect is reinforcing the size effect shock. Uh and so you get much bigger movement when  
[03:23:22] you dig in on these kind of second order things. There's still kind of sometimes a fraction of the total effect if you've got a huge decline in trade full stop.  
[03:23:31] uh that offset uh with an age appreciating skill that's growing or declining by less I should say uh is maybe not going to fully offset it but uh makes uh makes some difference and then that comparative advantage you know  
[03:23:46] this is a snapshot uh you know this this one tells us about the China US from 2014 to 2050 if we kind of take kind of means across time you can kind of see you know from the ' 50s to the to the early 90s or the late ' 80s I should say  
[03:24:00] you know Japan was aging much faster than the rest of the world. So this kind of tilt towards these highskll intensive sectors was very very large. So you would have seen a very steep slope with respect to how much those were growing.  
[03:24:13] Right? So the you know electronics growing at a very very high pace relative to other sectors for Japan in that period. Uh by the 80s they had already started to age at least relative to the rest of the world at a pretty  
[03:24:26] fast pace. And so that advantage within that sector starts to deteriorate. And so Japan from 1980 to 2016 is closer to zero. And actually uh you should note that the zero shifts over. It's essentially just zero. Uh from 2014 to  
[03:24:41] 2050 I should maybe normalize all those axes. Uh well they are normalized but I somehow it just looks unnormalized to me. Anyway uh so from 2014 20150 it is essentially just below zero in terms of  
[03:24:53] uh kind of its relative advantage in skills. So it's it's sort of uh how countries are moving along relative to the rest of the world because of course this is a story of comparative advantage uh when we think about your skill  
[03:25:06] relative to your trading partners. Um so what we have is this kind of tilt of this relationship that I showed you for the one corridor that's kind of happening over time with respect to my skill endowment relative to yours and  
[03:25:20] how that affects my comparative advantage in a particular sector uh over time.  
[03:25:26] And you can kind of see in uh a slightly kind of hard to parse at first but I think clean way uh how this looks. So this is kind of change in sectoral exports by 2050. And so each dot I've actually written it there. Each dot is  
[03:25:39] is a sector at its 2050 uh export change sized by employment. Uh so kind of the as a share of overall employment uh and the bar is kind of the change facing the average goods worker. And so you know  
[03:25:53] for a country like India you know you've got the average uh change and then this distribution of sectors underneath it right and I think um this is where a lot  
[03:26:07] of what the paper says that the model isn't able to tell us sort of specific answers about but tells us that maybe these things are going to be really important. uh if you think about kind of goods workers in Japan uh 97% of them are in sectors whose exports are going  
[03:26:22] to fall by more than a tenth right which is which is a pretty big amount now in in the model I I don't have the tools here I think it's a very important next step to really price that and sort of understand those frictions and how much that matters but this is a large part part of the workforce that's going to  
[03:26:37] see a huge decline in its uh demand from abroad uh and so I think this kind of shows you how there's kind of a huge disparity in terms of who's going to be hit by these kinds of trade shocks that are going to come from aging depending  
[03:26:51] on which country you're in and which sector you're in, right? You know, some some countries that are very old have sectors, you know, like in Italy, I should know what this is. I guess it's uh uh but uh you know, there's there's some sectors that might be spared and  
[03:27:03] some that will hurt very badly. Um >> uh can I ask a quick question? Um >> yeah, of course. Do you have some underlying productivity TFP catch-up growth or relative decline compared to  
[03:27:17] the US in your model? I mean if you think about China uh obviously composition is going to change of of its workforce, the absolute amount of its workforce, but also the productivity. They're catching up over  
[03:27:31] time uh in terms of you know relative average wages. So I was just wondering if that's part of your model. Yeah, it's a good question. So, you know, in the sort of historical data, I'm going to be calibrating these things and I guess  
[03:27:45] implicitly there might be a little bit of that that baked into these relative kind of um how the model gets calibrated, but going forward I I I don't have anything like that, right?  
[03:27:55] So, I'm not assuming any changes with respect to productivity. I I think it would be interesting, but it would be it would be hard to do in a way that would sensibly not just muddy the main story, uh so to speak, right? So I kind of I want to know how much demographics are  
[03:28:10] kind of capable of moving these uh I do think you know we've seen a lot today actually just in the last few hours about productivity and and and demographics how it's going to affect growth.  
[03:28:22] >> Let me let me put it a different way. Does your model, if you look at the historical, what your model's been has predicted for the past and what's we've seen in the data, does it really do a  
[03:28:34] great job picking up these shares of uh you know uh by by sector by industry, the share of whatever textiles that Korea is doing uh producing in the world  
[03:28:48] as a share of the world even forgetting exactly where it goes, but just as a share of world production.  
[03:28:54] Do you >> Yeah.  
[03:28:56] Yeah. If your model is really good capturing the data, then we would be less concerned that it might be missing something.  
[03:29:03] >> I see. Yes. Yeah. Yeah. Yeah. I could I could I could run that counterfactual and see uh I I suspect that, you know, it's it's not trying to nothing's going to under these underlying productivity draws are not going to change. So at at a fundamental level, we wouldn't be able  
[03:29:18] to capture those those types of things. Uh so I I can say with some confidence that there would be a lot unexplained by the model in that sense. Um but uh I I could do kind of a historical counterfactual and and and try to see  
[03:29:33] what fraction of it it gets. I mean, I think when I think of why you would set up a model like this, I guess I guess what would matter most is if I could think of a mechanism where that's going to interact with demographics, and I I  
[03:29:47] can, but uh I think the it would be it would be hard for me to think about how to do that in a way that that uh that that is credible, I suppose. Yeah. But but it's a good question. I should do maybe some some  
[03:30:02] historical counterfactuals and I know there there's a long history of these models that uh I should be able to tell you uh that have done similar things to what you're asking I think. Um but uh yeah I should I should maybe have some in my opinion.  
[03:30:14] >> So maybe we can uh >> yeah sorry time >> we can try to wrap up relatively soon and then move some of that discussion to the >> Yeah. Yeah. Yeah. So there's there's there's not too much left. Let me just say two quick things. The first is about welfare. Um the welfare implications are  
[03:30:29] huge from a domestic level. Uh small but like big by trade standards if you think about sort of um the international trade margin but there are offsetting forces.  
[03:30:38] What essentially happens in the model is that the goods sector ends up getting relatively more expensive and that's bad for welfare calculations and services ends up becoming from a demographic kind of trade perspective cheaper. Now they become much more expensive domestically  
[03:30:52] because of falling workforces and that's bad. But thinking about kind of more services being brought home because of these declines in trade that actually has a positive effect if you if you uh strip out the domestic side. Uh and so that's why we get these kind of net  
[03:31:05] positive numbers from trade. But there's a lot of reasons why I wouldn't really want to put a fine sort of point on these kinds of welfare numbers and it's because you're not doing the transition with frictionless kind of reallocation of labor. uh you know services here are fully non-tradable and the welfare numbers do shift around if you if you  
[03:31:20] allow uh some of the tradable sector you get much more sort of negative effects if more of those services are being traded because you don't get that that uh boost uh but uh and then you know there's there's a lot of other things uh the last thing other thing I'll say is  
[03:31:34] if you if you bring external balances in and allow for like a stylized version of external balances that just are essentially composition shift shifts the composition kind of portion of of Adrian's paper essentially um you can get external balances and  
[03:31:48] that matters a lot for like who's lending and who's uh who's lending and who's a the credit or debtor status uh allows countries to run kind of deficits uh and surpluses but it doesn't really move any of the trade uh changes that are sort of the interesting results of  
[03:32:02] the paper I think uh and there's a lot of robustness and nothing really changes any of the headline stories so I'll leave it there uh since I'm past time now uh but I think there's a lot of interesting extensions really what I wanted to do here was kind of have a sort of what are the first order effects  
[03:32:15] of demographics on these trade margins if we use kind of a quantitative trade model. Uh, and they're they're pretty big, I think. So, >> great. Um, thanks so much, Joe, for a  

## Q&A (03:32:28 – 03:40:30)

[03:32:28] great presentation. So, the floor is open for um questions. Um, and we'll have the next paper in about eight minutes.  
[03:32:44] you have yeah um a great paper and I'm wondering like whe whether you've looked at more like inequality um because I think um it's kind of natural that these  
[03:32:57] models like quantities are mostly driven by just scale um so but I think there's much more action on the relative prices of different uh  
[03:33:10] skills and uh I forgot what what you call Yeah. So, I don't know about inequality, but the the one big thing I think that that I think I'd like to do, but is um requires I think a lot of care in  
[03:33:23] calibrating properly is thinking about the demand side of this model, which is that there could be sort of and then there's a lot out there on this um sort of uh the kinds of goods you consume as an older society shift a lot and that affects the demand for these goods and  
[03:33:38] that I think maybe would feed in a little bit more to an inequality story as well. thinking about uh you know if you wanted to have something like that I don't know how tractable it would be to have distributional effects on that side um you know we just have a represent I have a representative household here um  
[03:33:53] but I do think that would be kind of the next logical step and at least getting in some kind of demand side where you could credibly try to estimate similar kinds of profiles for the age distribution on the household side  
[03:34:08] domestically in terms of what stuff they're going to want to buy. you know, you're going to need a lot more of certain services perhaps. Um, that might reinforce a lot of a lot of the story.  
[03:34:18] Uh, maybe in some sectors there's a lot of high demand for manufactured imports as well, uh, for an older society. Uh, and then yeah, so I I suppose that's not even close to an answer to your question, but uh, I think I think would  
[03:34:32] be the first step in getting there. Uh I guess unless your question was more about country level inequality in which case uh yeah I'm not sure.  
[03:34:42] I I have a related question about uh just the focus you have on exports as opposed to imports because I guess I was thinking okay as workforces age we we'd expect to  
[03:34:55] see a tilt in the composition of imports towards say the countries that have bigger war workforces. But if I think about exports, it seems like it's more about demand and and so I guess in your model you  
[03:35:09] don't have this distinction, right? But um >> yeah, these are symmetric. Uh >> it's symmetric but right but if you thought about well I might expect my exports to tilt towards the countries that have bigger populations say right or my imports are tilting towards the  
[03:35:24] countries that have bigger workforces. So I guess if there's a distinction between sort of population and workforce um then perhaps the you know like imports are a more natural outcome to look at um I don't know if you if you've  
[03:35:37] kind of thought about that um or there's a particular reason why your results are presented in terms of these exports as opposed to the imports.  
[03:35:46] >> Yeah. Um no it's a good question. I I suppose I I started with exports because it's sort of the logical kind of uh I think imports kind of almost demands this kind of why are countries buying the stuff they buy argument whereas everything's coming from the why countries produce what they produce in  
[03:36:01] terms of the way the model's set up. Uh because it's very much about the endowment of skill. Um I think there would be some interesting things I could probably present here  
[03:36:13] thinking about um the mix and where things are going right because of course even in the balanced case right so in the stuff that doesn't have the external balances uh everything has to net out uh in terms of exports imports because everybody's running  
[03:36:28] balanced trade uh but even in that one uh I could probably show a lot of things because of course uh the concentration of where these exports are going is going to be is going to be interesting.  
[03:36:39] And I haven't I haven't really done that, but I think it maybe could tell an interesting story because of course you see from a lot of these, you know, most of what I show are just based on what the biggest flows are, right? In terms of the biggest corridors of trade, uh, which is why I choose the country corridors I do, you know, they're just  
[03:36:54] the ones that matter the most. Um, but as you can see in like the the the pictures, you know, just everybody except for a few are trading less with the US. this kind of massive corridor is kind of like seeing a lot of uh movement  
[03:37:08] away from it from a lot of its historically largest trading partners. I mean Mexico and India are big trading partners but um and so I could probably try to map that out in a way that that says something that that shows a different side of the model. Um, but I do think that when I think about  
[03:37:22] imports, it even though here it'd be the same thing, it makes me think more about the demand side, I suppose, which I think is interesting, but I think just had a much less credible calibration target, right? There was kind of some clear science already done essentially on kind of the labor market side of  
[03:37:37] things. Uh, whereas my reading of the demand side of things is a little little foggier in terms of clean calibration targets. Um, >> but another way to pick maybe or think about it is is to see whether there's  
[03:37:51] any empirical counterpart just like in the raw data to some of your regressions where you were looking at like the tilt towards sectors that have >> that employ the old people effectively, right?  
[03:38:04] >> Yeah. Yeah. >> And then also the tilt towards you know countries and it's like is there are are the regressions from your model let's say 1950 to today you know are they >> predictive of the actual changes in the composition of either exports or  
[03:38:18] imports. um I see that we actually saw right and and then you know perhaps perhaps one is just empirically like a better >> fit than the other right if you just >> just plotted model prediction versus data um in in  
[03:38:32] >> yeah I did do a little bit of this I've got it in appendix in the paper um it it in some ways it matches really well and in some sense that's because this kai paper that I sort of took this  
[03:38:45] this skill tilt thing from >> uh because I'm calibrating to match the things that they did and they run these kinds of regressions. It's an empirical paper. Um so I I kind of match that very well but kind of by construction because  
[03:38:59] of course I'm feeding into the the model these these kind of outcomes >> but not from the way to the >> Yeah. When you're saying it matches really well you mean you mean even the export prediction that's not something that came from them right?  
[03:39:13] >> No. Yeah. So so in terms of that Yeah. If if I run a gravity equation with with these kind of things in uh I it does match reasonably well. There there's there's two papers, one I wrote and one that some other uh Bman and some co-authors wrote that that did this kind of like what if we just shove  
[03:39:27] demographics into gravity and and and do an empirical thing. And it you know I I I should do sort of a version of okay take my model simulated data and and just see see the correlation between that and the demographic component of those gravity equations. Yeah, I I I  
[03:39:42] should just do that, I suppose, because that's a simpler exercise than some other things I've tried to do and might show.  
[03:39:48] >> Very nice. >> Thank you. It's a good suggestion.  
[03:39:52] >> Great. So, um >> we have another 30 seconds to the next presenter. So, is Larry, are you are you presenting the next uh paper?  
[03:40:02] >> Uh I am. I'll uh do I want to share my screen right now?  
[03:40:07] Uh yes, you can stop sharing your screen and then we'll just wait one minute so everybody's synced up and then we'll uh we'll get going.  
[03:40:16] >> Awesome. >> Great. So we're right on time. So um our