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

# Putting Economics Back into Geoeconomics

Authors:   
Discussant: None  
Video: https://www.youtube.com/watch?v=3YzXcC2W46s&t=0s


## Talk (00:00:00 – 01:38:56)

[00:00:00] [music] Thank you for being here. Um, today we're going to cover geoeconomics which is largely work with my co-authors uh Chris and Jesse. And so we're going to  
[00:00:14] just jump into it. Um, you know, it's a topic that has attracted an enormous amount of attention over the last, you know, five years, but probably especially in the last two. And it used to be something, you know, that economists didn't really pay a lot of  
[00:00:28] attention to. And now there's a lot of research. So I'm kind of hopeful that for all of you this will become something exciting to do work in. Um what is the definition at least the definition that I'm going to use uh is going to be that aimmonic countries like  
[00:00:41] the US and China use the existing trade and finance relationships that come from their economies to exert power abroad.  
[00:00:49] And they might be exerting power for geopolitical reasons for economic reasons. They might be bullying a government or a private firm. we're going to sort of figure it out. Um there's just some set of interesting questions in this area that I think I  
[00:01:03] want you to to have in mind. One is you know first is this a thing like in what sense is this powerful where does it come from? If you have it how are you going to wield it? What does optimal policy if you're a country like the US  
[00:01:18] or China look like? Um, of course, we're immediately interested in welfare, particularly in the welfare consequences for the rest of the world. So, you can imagine that a lot of the colloquial version of this topic has a zero sum  
[00:01:32] logic. It's the logic of essentially mercantalism. There is a fixed set of resources and we're going to bargain over these resources. But as an economist, we've long ago dispelled this notion that that's how the world economy works. We think that the pie might be  
[00:01:47] increasing or shrinking as we exert power and that everybody could potentially be better off or worse off.  
[00:01:53] So we're going to think about that and then we're going to go a little deeper.  
[00:01:56] We're going to start thinking about okay what is national security? What is a national security externality? How do you measure it? uh and in particular is there a tradeoff between the gains from trade the arguments that you're all  
[00:02:09] familiar with that come from specializations that come from external economies of scale um versus economic dependency and if there is that trade-off then how does optimal policy shape it uh and of  
[00:02:23] course we're interested in measurement um can we do policy counterfactuals if the US will be cut off by red earths uh how much the US economy will go down and which sectors are going to suffer? Is  
[00:02:37] this going to cause a garden variety recession, no recession at all? Is this going to cause like a massive recession?  
[00:02:44] Like those quantity questions are really sort of interesting.  
[00:02:48] Okay. So the way I'm going to structure the lecture for today is I'm going to give you a bit of a background of the topic also away from economics in particular in political science mostly because I suspect that all all of you as  
[00:03:01] PhD students have no idea about that and to be honest neither did I like I had to go and learn it. Um and it's useful. Uh then we're going to go deep into theory and then we're going to do impedics uh in the second half and we're going to do  
[00:03:14] two different types of impedics. We're going to do something that is much more familiar to all of you particularly those that have taken a trade class uh which is sufficient statistics approach uh so mixing a model with the hard data  
[00:03:27] trying to get some quantitive answers and then we're going to look at something quite different potentially which is looking at unstructured text as a way to figure out who's exerting pressure on whom. Okay, good. So let me  
[00:03:41] start with the broader conceptual questions. Um, one complaint that I've had for a while that got me interested in this topic is that if you think about the modern econversion of power, we kind of reduce everything to mean market  
[00:03:55] power, the ability to sell a good markup and a lot of the basic classes that you taken in IO for example or uh in macro have that notion of market power. But when we say that a country is powerful  
[00:04:08] or that even a company like JP Morgan or Google might be powerful, we generally have in mind something broader than just the ability to sell a markup. And so there are some interesting concepts of power that come from political science  
[00:04:21] that are worth highlighting. The one that I'm going to use the most is the very top one comes from Robert Dal in 57 who says look A has power over B to the extent that he can get B to do something that he would otherwise not want to do.  
[00:04:36] Okay, you can see that a markup is one such thing that the guy wouldn't like to pay up. Um, but potentially is very general is what the political scientists call relational power. Why? Because they're thinking it's affecting the  
[00:04:50] relationship between A and B, but it's taking the entire environment as given.  
[00:04:54] Okay. Susan Strange um, she had a different notion called structural power. uh which is essentially the ability from a player to affect an entire environment uh she had in mind for example setting up the rules of the  
[00:05:08] game or setting up some norms you know today I'm going to give you our own rendition of this which has much more with influencing general equilibrium variables so being large in  
[00:05:20] that sense okay the political scientists themselves complain that this is a topic that within their field has remained very informal and chit cchaty and therefore the people don't know exactly what they mean. So part of what I'm  
[00:05:34] going to try to do today is to give you at least you know our own rendition or take on this to make it precise enough that you can do theory and bring it to the data. Okay. But I strongly strongly suggest reading some of these things. Uh in particular I would suggest reading  
[00:05:48] this book um which I did not put on the syllabus um but I would strongly suggest you read. Um it's a fun book um and it's also quite readable. Well, fun it's a strong word. Um, it's a book on a  
[00:06:01] particular topic which is Nazi Germany uh influencing the rest of Europe in the leadup to World War II and using his economy to exert that influence. So, Hman is an interesting character because  
[00:06:15] first he has a Netflix show uh it's called the transatantic is very good um but he's famous for many different reasons. Most economists know his later work uh in political science. Um that's  
[00:06:28] great. Some economists know the work in the middle of his career that is essentially the work that set up the input output matrixes and the backward and forward linkages like the work that for example Bakay and Far have been doing picks up on that. This is earlier  
[00:06:42] than that. This is him as a young economist. He goes to Berkeley. He's trying to get a job as a faculty member at Berkeley and writes this as essentially a trade book. And it's interesting because he has in mind two things. First, he had a very obvious  
[00:06:56] distaste uh for mercantalism. Uh the idea that we're simply bargaining over fixed resources. But at the same time, it was pretty dismissive of the English free traders of the idea that if we  
[00:07:08] simply had free trade, this will never generate any imbalances because everything will be evenly distributed with many partners. So he had in mind a view where uh trade leads to imbalances  
[00:07:21] and those imbalances lead to power imbalances and and it's a pretty interesting book and I strongly suggest reading it. Um now what's interesting is that as part  
[00:07:34] of the book uh you derive an index that you all know uh which is the Herendall index. Uh part of my complaint that everything has been reduced to market power is that you've all learned that index in the  
[00:07:48] context of measuring concentration of firms for markups.  
[00:07:53] But the original index was created to measure concentration of trading partners compared to another sovereign and therefore power imbalances in trade.  
[00:08:02] But it has totally lost um that connotation.  
[00:08:06] Now if you go um and pull out the citations of that book uh you can see that it used to be a big deal in economics then not so much in the middle a field open up in political science called international political economy  
[00:08:20] that has really focused a lot on these issues. Um so in the next slide I will have some citations of well the papers that I wrote and and I'm asking you to read for the class but also at the end a long reading list of papers that are  
[00:08:34] away from economics that if you're interested in this topic I strongly strongly suggest reading there is just a lot of good deep thinking uh in political science about these issues that economists tend not to be exposed to and if you're a young PhD student  
[00:08:49] there might be lots of papers hidden in there as those ideas spark some interest. Okay. So, I strongly suggest uh picking them up. Okay. So, for the purpose of um what I'm going to do um  
[00:09:01] for the modeling side, part of what I think we did the last few years has been building models or frameworks to understand these questions out of basic tools of economic theory.  
[00:09:14] You you will see that I'm going to highlight three. Um two of which are very familiar to you. Well, all of them are familiar to you, but some of them you're more familiar with than others.  
[00:09:23] The first is going to be very straightforward. I'm going to define power as essentially a gap between the inside option and the outside options and the participation constraints of the people you target. It's going to be a very nice way to keep track of these  
[00:09:38] notions of power. Okay? So, we're going to use that all over the place.  
[00:09:43] The second tool which is probably a little less familiar to you has been to think of um costly actions. Think of exerting power as in the D definition as asking somebody to do something that  
[00:09:56] they wouldn't want to do. We're going to make that formal by saying I'm asking you to take to put wedges in your first order conditions. They're deviations from what you would have privately optimally wanted to do. The reason why  
[00:10:10] they're so powerful is that 30 years of micro theory and public finance theory has figured out how to solve lots of optimal policy problems in that context.  
[00:10:20] Okay. And so we're going to review those.  
[00:10:23] Finally, we're going to try to give an incarnation of the structural power. The idea that part of what makes me powerful is that I can manipulate an entire equilibrium in my favor. um we're going to think of it as manipulating  
[00:10:38] externalities in general equilibrium. We're going to say that there are some equilibrium objects that I take derivatives over when I take my optimal policy actions that the agents take as given and that's going to give us a nice theoretical distinction between  
[00:10:52] relational stuff which I'm going to call micro and the macro stuff. Okay. So these are the papers that I've asked you to read but in and they're all with Chris and Jesse and for us this is a great area. We're doing lots of work. I hope you get involved. But even more  
[00:11:06] important to you is if you click on that button on the reading list, there's a long reading list in both economics and political science. Both of current work that other people are doing in this area. Steve Reading for example here with Bener working also on this and they  
[00:11:20] have great papers. But [clears throat] even more relevant in in terms of hard to access for you, there's a lot of reading in other fields that I suggest you do in general as an economist, but certainly particular if you're going to  
[00:11:33] get involved in this. Okay, so let's fix ideas. Okay, we're going to fix ideas and then we're going to go formal. So to fix ideas, start with an example, okay, which we're going to carry throughout.  
[00:11:44] Think of having ASML. So this is the Dutch lithography firm, okay? They produce advanced machines for semiconductors. They have US suppliers and they have Chinese customers.  
[00:11:56] And if you think of them as a private enterprise, think of zed as a vector of all the aggregates, all of the state variables. Prices, aggregate demand, the state of technology, all of the typical  
[00:12:09] stuff. Xar is the firm optimal decisions of input output. So who to source from, how much, who to sell to, how much if it's undisturbed. Okay. So it's essentially what comes out of first  
[00:12:23] conditions. Okay. Now think of the US government.  
[00:12:28] He wants to affect this firm. What he wants is for the firm not to sell these machines to China. Okay, let's forget for a second whether this is a good idea or a bad idea. That's what they're  
[00:12:41] trying to achieve. Well, the way we're going to think of it is we're going to think that he wants to put a wedge on a particular dimension of the actions of ASML that makes it more expensive for it  
[00:12:54] makes um ASML perceive that is more expensive to sell these machines to China or you can think of it as a quantity restriction. It might simply want to dictate that you can sell at most X machines. Okay, where X might be  
[00:13:07] zero. The fundamental issue is that ASML is not a domestic firm of the US. So the typical records that you have over domestic firms, you can dictate export controls, you can regulate them, you can  
[00:13:22] tax them. You don't have them. This is a Dutch firm. So you have to induce them to take the action.  
[00:13:28] What can you do? Well, you can do the usual things. First, you can provide positive inducement. You can pay. You can say, "Look, if you do what I say, I'm just going to give you money." Or you can provide negative inducements.  
[00:13:40] You can say do do what I say or else something bad might happen to you. In this particular case, um the US uh went for the negative inducement. They threatened ASML that if they were not to  
[00:13:53] comply with the US demands, they would invoke potentially or allegedly uh the foreign direct product rule. This is a rule that essentially tells ASML that uh if they don't comply, they're going to be added to the entity list, which is  
[00:14:06] the list uh of companies or entities that no US person can do business with.  
[00:14:12] So that essentially is a death sentence for ASM. You cannot have any relationship with any US person. Um that's not going to work. So if you think of it formally, what are you doing? the threat which we have with a  
[00:14:24] data in red. It's moving the value function of the firm way to the left if the threat is very powerful. It's telling you that potentially I can do something to you that while you're going to be able to react to it and reoptimize  
[00:14:37] might make the value of the firm a lot lower. So the participation constraint then becomes very natural. uh ASML is comparing what they can do at the outside option once they lose access to  
[00:14:51] being doing business with the US compared to at the inside option not having the threat so dealing with a full set of US persons but acquiesing to the demand of the US taking on the costly  
[00:15:06] action that's it I mean this is loosy goozy for now we're going to make it very formal in a second but a lot of The intuition of this work comes out of keeping track of these objects. You can immediately see that now it's all going to be about  
[00:15:21] how do you shift the inside option? How do you shift the outside option when you're collectively bargaining with many people? How is that shift in the zeds the aggregates that all of the firms take as given but that you might be able to manipulate. Okay, so a lot of it is  
[00:15:35] going to show up there. And just to warm up here are simple definitions. If you think of relational power like purely micro think of the firm the firm is taking as given the equilibrium aggregates and  
[00:15:48] it's simply checking what is the maximum cost of private actions is willing to comply with before is better off at the outside option. So we're going to define micro power to be precisely that  
[00:16:00] concept. Now you can see that later when we go to the data things like the las substitutions if I'm threatening you with not selling you something how easily can you find an alternative can you source it somewhere else if I'm  
[00:16:15] threatening with withdrawing technology can you create the technology yourself all of those objects are going to tell us by how much am I able to move the inside and the outside options so a strategic sector in this sense is going  
[00:16:28] to be one that makes you form very valuable threats so Let's make a simple example. Okay, consider oil.  
[00:16:35] Okay, oil, one variety of oil isn't strategic. Why? Because if I tell you that I'm not going to sell you this very particular variety of oil, many other types of oil are from other countries are pretty close substitutes. If I'm  
[00:16:49] OPC, particularly in the old days, and I control the entire supply, then oil might be very strategic because now the outside option is using something very inferior to oil. Okay. But you can see  
[00:17:02] immediately in this concept that you can go to the other one to micro power. In micro power the hedgeimon is exploiting precisely that gap is exploiting that when it bargains with individual entities they are taking they're  
[00:17:16] tracking their own private costs. But the benefits to the hedgeimon are going to acrue from changing the environment from the fact that as I tell multiple people not to do something I might be affecting the aggregate. So let's go do  
[00:17:30] a simple example. Suppose that I want to have a very powerful social media. Okay, social media has a strategic complementality. You want to be on it because everybody else is on it. So what am I going to do? I'm going to tell each  
[00:17:44] one of you that what I want you to do is to be on my social media and to close your other accounts on the competing social media. Why do I care about that?  
[00:17:54] because each one of you is going to take the equilibrium as given and is going to compute the private cost of shutting down their other social media and focus on mine. But as you do that, you're making my social media very attractive  
[00:18:07] for everybody else and the outside options very unattractive. So once I have you all hooked up on my social media threatening to kick out any one of you individually, it's very very powerful because now your outside option  
[00:18:20] is a social media and nobody else uses while the inside option is the one that I control and everybody's on it. So all I'm doing is a drug dealer model. I'm sort of getting you to coordinate on an addiction and make the outside option very very bad. That's a simple example  
[00:18:34] of this one. In this sense, strategic sectors are those that allow the hedgeimon to manipulate uh the economy in its favor.  
[00:18:46] You will see later when we do optimal policy that this also means that in this class of models the presumption for government intervention is much lower than say in micro policy in financial regulation. In financial regulation, the  
[00:19:01] problem is that we're all trading and we don't internalize that we're all employing the same strategies and that because there are some prices in in constraints, we're going to have five sales. But it's a coordination problem.  
[00:19:13] Here is very different. You're playing against a big player who's twisting the equilibrium in their favor. So the probability that you as a government, if you internalize more of the game, you wouldn't want to intervene is pretty  
[00:19:26] low. Okay? So it's just Here's another good example. Think about the western seabboard or the US. Each part that accepts a contact from China might take the value of national security as given  
[00:19:40] and say look China is asking is providing this bid. That's good or bad for me. But as China controls the entire western seabboard of the US and Latin America, you might think that that affects the national security of the country. But each of the private agent  
[00:19:55] isn't taking this into account and therefore the probability that you might want to have government screening for strategic sectors is pretty high. Okay.  
[00:20:04] So the the intellectual backing for it comes from a very different nature than a simple fire sale externality that you're familiar with in finance or market. Good.  
[00:20:17] So we're going to start the model and we're going to do it formally. Okay. And part of what I want to do is focus on the specifics of this topic, but it's also going to allow me to cover some tools in particularly the primal approach to solve optimal policy that  
[00:20:31] are of interest to you much more genetically than the specific problems.  
[00:20:35] Okay, so let's start. We have n countries, right? And there's a set of productive sectors and a set of local factors. So a sector here, it's an activity in one country. Think of Russian oil and US oil are two different  
[00:20:49] sectors. They might very well be highly substitutable in their output com when they're as inputs in other sectors. But an activity taking in in one place is a sector. Okay. The sectors have a unique mass of firms. They're pretty  
[00:21:03] straightforward. What do they do? They're buying intermediate inputs from other sectors. They're using local factors. And the only thing that is different is that I'm going to stick in their production function a set of aggregates. So this is a trick that has  
[00:21:18] been used a lot uh from greenwall stiglets on how to capture externalities very simply is just to think of a sector of aggregates that potentially affect the production function of the consumer.  
[00:21:29] Simple examples. Okay. Suppose that you wanted an external economy of scale.  
[00:21:33] Then zed is the aggregate output of that sector. The bigger the aggregate output, the more each of the firms is productive. If you want a strategic complimentarity, that might be the  
[00:21:46] sector usage of a across sectors of an input. The more other sectors are using this input, the more that input is productive for me. If everybody else is on on this payment system, me being on  
[00:22:00] the same payment system is productive. Okay? So, it's going to be very simple to play around. And then they're going to have a representative consumer in government.  
[00:22:09] The representative consumer is very straightforward. So the part that I'm going to focus on, it's simply utility over consumption. But I'm going to allow to capture some ulterior motives for geopolitics as saying that he also has a  
[00:22:24] direct utility over some aggregates. Just to make it simple, suppose that I truly don't like another country military sector being big. Okay, I'm gonna plug it in there. Or I truly don't  
[00:22:38] like uh I care about the difference between your output and mine. Okay, like a habit. I'm not going to play much with it because well, we have no reason. We have no foundations and so it's totally  
[00:22:51] arbitrary. Um you probably are familiar with those things for example in climate change where pollution might make the consumer unhappy. Um it's fine. I think what's useful in this setup is to see given the objective function how it  
[00:23:05] translates into the policy game. But the objective function is totally made up.  
[00:23:09] So I'm largely not going to use it and focus on consumption. Okay. The budget constraint is very straightforward. They have factor income. The factors are only used domestically. They have the profits of the firms and that has to finance the  
[00:23:24] consumption plan. So what did I sneak in? I sneaked in an assumption that uh each consumer owns a domestic firms.  
[00:23:33] There is nothing difficult in this set of papers allowing crossber ownership.  
[00:23:39] You can think of multiplying that matrix of profits by an exogenous ownership matrix. What's difficult and interesting is to solve for endogenous ownership.  
[00:23:48] And I will give you a sense when we solve the equilibrium of why you might want to have a shopping list for FDI.  
[00:23:54] you might want to to go abroad and buy stuff, but here for now I'm not going to allow it. That gives me wealth. That's essentially the right hand side of the constraint. Given wealth and prices, the first story condition will define a  
[00:24:07] Marshall and demand and direct utility. And that's going to be my objective function if I'm the government. Okay.  
[00:24:13] And then markets have to clear. So anything that gets produced get either used as an intermediate input somewhere else or gets absorbed by the consumer.  
[00:24:22] Okay. and the and the and the factors have to clear. That's it. Okay. So, think of it as we set up a large input output matrix globally and there's reasons to trade with each other and there's a very simple consumer prop where it's interesting as the policy  
[00:24:36] game. So everything we just described happens at the end.  
[00:24:42] In the middle I'm going to have the offense.  
[00:24:46] So that was part of our first paper is a theory of if I'm a hedgeimon what do I do where does my power come from and how I exert it I'm going to make the a country the hedgeimon exogenously I'm not going to think about  
[00:25:00] why is it the US going to say this country is a hedgeimon now what defines it being a hedgeimon is the ability to do the following in this paper is to threaten foreign entities that if they  
[00:25:14] don't comply with what the hedgeimon wants they're going to lose access to a set of input that the hedgeimon controls. So for the purpose of today I'm going to focus on threats not to sell you stuff. Okay? Then we can think  
[00:25:29] about threats not to buy from you other positive inducements. But for now let's focus on the threat being I don't sell you stuff. I don't sell you semiconductors. I don't sell you rare earths or oil.  
[00:25:42] As we discussed, what I'm going to ask you to do is to put a set of first of wedges in your first order conditions and potentially to give me transfers.  
[00:25:53] Okay, the wedges I'm going to make them revenue neutral. What does it actually means? It means that I rebate them to you, but each agent doesn't perceives the rebate as fixed.  
[00:26:05] Basically, they work like quantity restrictions. In fact, formally, we're going to use the primal approach, which means I'm just going to directly dictate the quantities, okay? But we're going to implement them with wedges, okay? And  
[00:26:18] you will see it in a second. And the rest of the world is going to simply accept or reject.  
[00:26:24] Okay? So, all of the bargaining power is with the hedgeimon who makes a take it or leave it offer. So, the theory of the offense is going to be what do I do with these things? What do I ask you? Why?  
[00:26:35] And how do I maximize my power? Good. So that's one paper.  
[00:26:41] Then we're going to go down here and we're going to say, okay, ahead of time at the beginning, every country is going to have a full set of wedges that they can apply on their domestic firms. And for simplicity, we're going to have no participation constraints. We're going  
[00:26:56] to assume that domestically you're entirely dictatorial. You can do whatever you want. Okay. The difference is that in the middle the hedgeimon is bullying foreign firms, foreign entities with whom it might have  
[00:27:10] an economic activity. Here you're doing domestic policy. So those wedges are going to be the typical things that you're familiar with.  
[00:27:18] Industrial policy, export controls, trade policy like tariffs, things that you do to your domestic firms. Now none of this has to have like a negative connotation. Some of that might well be uh solve for domestic provision of  
[00:27:32] public goods uh maximize the positive externalities of technology like all of that is going to be allowed. Okay. The part that we're going to be interested in is that the countries are anticipating  
[00:27:46] that the hedgeimon will come knocking in the middle. So they're thinking about how do I shape my own economy ahead of time to withstand the threats and coercion later. Okay. So we're going to  
[00:28:00] focus on the part of the policy that comes from thinking about how shaping your own economy is going to impact what the hedgeimon is going to do later. And so that's a big topic now in policy.  
[00:28:12] It's called anti-coercion policy or economic security policy. Is do I want to be for example very dependent on a foreign country? Do I have to tell my firms that they cannot import all of the materials from one dis from one source?  
[00:28:25] All of that policy is what we're going to focus on. Okay, that's going to be the game. Okay, so let's get started.  
[00:28:32] Okay, so I promise you the participation constraint was key. Now let's look at it formally. So each foreign entity in the middle, it's comparing the the inside option which is producing with a full  
[00:28:45] set JI of all of the inputs you can access to undisturbed but taking on some costly actions and making transfers versus an outside  
[00:28:57] option which is no costly actions, no transfers but you produce with a subset of the goods. Okay, J with a circle for outside option. So if you look at the value function on the inside option,  
[00:29:10] they're choosing the vector of intermediate inputs to use the vector of factors. They're maximizing profits, but they're perceiving some wedges. They're perceiving wedges that they have accepted from the  
[00:29:24] hedgeimon and wedges that their countries have imposed in the previous period.  
[00:29:30] And the stars here are reminding you that in equilibrium this gets rebaited. But each firm takes Xarat as given. So they proceed the cost. Okay, that's how wedges work out in these models. And the outside option  
[00:29:44] is very similar. I'm now producing with only a subset of the available inputs, but I'm only subject to domestic policy and I make I make no transfers. Okay,  
[00:29:58] that's it. The welfare functions I think they're pretty straightforward where I mentioned that for both the hedgeimon and the each country they're maximizing the indirect utility of the consumer. The only thing  
[00:30:12] I want to highlight is when you look at the wealth of the hedgeimon, it has one more component, right? It has the profits of its domestic firms, the factor payments of the domestic economy and the transfers he gets from the rest  
[00:30:25] of the world. The rest of the world has for the firms that don't get bullied the profits for the firms that get bullied because in equilibrium they're going to accept the contract is the value of the outside  
[00:30:39] option or is the value of the inside option minus the transfers since the constraints will always bind is really the value of the outside option and then you have factor income.  
[00:30:49] Okay. So let's start solving it. First let's solve an example together so that we see a close form and I'll give you a general for formulation.  
[00:30:57] First, I promised that we were going to get rid of this U and Node Z for now.  
[00:31:01] So, I just set it to zero. The other thing which is connected with what I was just replying to now is I'm going to make it even stronger. I'm going to assume that the consumer has linear utility, which means that prices in equilibrium are going to be constant.  
[00:31:14] Okay? So, the entire variation is going to come from the zeds from the technology. Okay? I'm going to have no terms of trade manipulation. All of the prices are going to be constant. And later, I will show you what happens when the prices are indogenous. It's going to  
[00:31:26] make us focus on the zeds. Okay. Good. Okay. So, let's specialize that economy to something that we can solve in close form to to get some ideas. Let's say that the hedgeimon is the US. It has a single country. Sorry, as a single  
[00:31:41] sector. It has uh bankers, so local labor factors that produce financial services linearly out of labor. Okay.  
[00:31:50] The rest of the world is a collection of identical small open economies. They only have two sectors. They have a domestic sector that is essentially the domestic alternative to the US. It's a sector of local bankers that use labor  
[00:32:04] to produce local financial services. Okay. In the middle, I have an intermediary, this blue eye, that takes the local variety and the global one coming from the US, puts them together  
[00:32:17] and creates a composite good. Okay, I'm going to do an example with payment services, but you could do information technology is totally generic. Okay, where are the externalities going to be  
[00:32:30] on the domestic side of age? I'm going to have an external economy of scale.  
[00:32:35] I'm going to assume that the bigger the domestic sector, the more productive it is to use it for the intermediary.  
[00:32:41] On the global one, I will have a strategic complementarity across the sectors. I'm going to assume that for each intermediary is more productive to use the global technology if all of the intermediaries in its country and in the  
[00:32:54] other countries are also using it. Okay. So you can think like this with the example we had in mind when we created the paper was the messaging system or the payment system. It's productive for me to use it because every other country is using it and so we can send messages  
[00:33:08] and payments to each other. I have a domestic alternative. If it's not scale up at all, it's going to be very inefficient to use it. As soon as I start scaling it, it's going to be at least more efficient domestically to use that alternative. Okay, good. We're  
[00:33:23] going to make functional form assumptions about the intermediary and it's one that you're going to be very familiar with and we're going to use it all over the place in the next two hours. So, just get familiar with it again. It's a CS aggregator.  
[00:33:37] It's combining the domestic alternative to the global alternative.  
[00:33:42] Where are the externalities that we just discussed hidden? They're hidden in the productivities of using the two alternatives. So you can see that for the global one, the productivity for intermediaries in countries J depends on  
[00:33:56] the average usage of that technology across all other countries with the strength of the externality being governed by Kai. Okay, the higher is SI the more as they use it this becomes  
[00:34:09] productive for me and if S is zero then there is no externalities the domestic economies of scale has the same functional form but only depends from what other intermediaries in my own countries are using so it's an external  
[00:34:24] economy of scale you have to make some assumptions to maintain the problems being convex uh that's it okay now let's solve it so this one I'm going to we're going to do as a warm-up. Okay, so before we solve the actual problem,  
[00:34:39] let's do two warm-ups together. The first is the global planner. It's a standard benchmark of international macro. I have a global planner that has all the instruments and is trying to maximize global welfare. So here is a planner that is maximizing the sum of  
[00:34:53] welfare of all the countries. Okay. And this is also where I'm going to take a little bit of a detour and go through the primal approach. And if you if you've seen it, it's pretty straightforward. If you never seen it,  
[00:35:06] this isn't quite the full general theory of it, but it's a useful introduction in an environment that is simple to visualize. Okay, so please stop me if you have questions about the technical aspect. Okay, so let's think about this first. Uh you just did a problem set  
[00:35:21] with me for a planning problem. Uh here's one. The since the countries are symmetric, the global planner optimals has symmetrical allocations. So I'm going to focus on what it does to a particular country genetically is  
[00:35:35] maximizing profits and here it is. It's the intermediary profits min sorry the intermediary output and revenue minus the cost.  
[00:35:48] So the first order conditions is pretty standard. This is the standard part that comes from uh taking us given the technology figuring out how much do you want to use of each of the two uh  
[00:36:01] instruments. This is the first order condition for the use of the hedgeimon u for country J uh inputs. [clears throat] But then because it's a planner and we assume that internalizes everything it's taking derivatives over the  
[00:36:16] externalities over the A's. So this is the production externality. It's taking into account that as I increase the usage of good J by intermediary I. So as I increase the usage of the hedgeimon  
[00:36:29] technology from each of the countries that increases the productivity from everybody else and then in particular I need to scale it so that I get the appropriate uh output elasticities.  
[00:36:41] Okay. So if I were not doing a planner problem if this was competitive with small countries this would be zero. They wouldn't be taking derivatives over this object, right? They would take the technology as given. They would think that I'm too small. Nothing that I do  
[00:36:53] affects the technology. Okay, good. Now do the same thing but for the each intermediary in the decentralized economy that faces wedges. So you can  
[00:37:06] see that here I didn't have the wedges. I simply said suppose that the planner had complete instruments and could directly choose how much each intermediary uses of each good. That's what we solved. Okay. Now go back and  
[00:37:20] say okay let's go back to the problem of each intermediary.  
[00:37:26] Let's figure out their first story conditions when they're facing wedges.  
[00:37:31] Okay, wedges on both the usage of the foreign inputs and the domestic alternative.  
[00:37:38] What do I get? I get something that is pretty similar to the previous one. Um I have the optimal usage of of J taking as given the technology. I have lost the part where I take derivatives over  
[00:37:52] the level of technology. But now I have a wedge on this side, right? I perceive the price of the technology not to be PJ but to be PJ plus EJ.  
[00:38:04] Okay. Okay. So that's the first order condition. Now what is the name of the game of doing uh the primal approach is that we're going to try to figure out a set of wedges  
[00:38:17] that induce the first order conditions of the private entities to be aligned with the planner solution given those wedges.  
[00:38:29] Basically what are we doing? We're comparing the global planner first story condition that's up here to the first story condition that just we just derive from each intermediary privately doing  
[00:38:43] its best given the wedges and then we're going to put them together and solve for the wedge. Now how do we do it? Well take this first object okay and then multiply through and bring it to the other side. Then the  
[00:38:57] left hand side is identical. On the right hand side in one I have PJ plus J.  
[00:39:03] In the other one I will have PJ plus this object times this.  
[00:39:09] Right? That's what I wrote here. So they equalize for this.  
[00:39:14] Now stare at it again and remember that from this condition I know that this part is simply PJ to AJ. This one is simply a functional form. If you work out the math of that derivative, turns  
[00:39:29] out to be J. That's why we chose that particular functional form. So now I solve and that's it. That's the solution.  
[00:39:37] That is the primal approach is you start with a planner directly dictating quantities. Then you look at the private solution of each entity in the game looking at the wedges and you're trying  
[00:39:51] to look for wedges that implement in that equilibrium the planner solution.  
[00:39:57] Okay, it's a very general tool and is very helpful for optimal policy. Here we're going to use it all over the place. So that's the outcome. The outcome is not surprising. The outcome is as a planner I understand that there  
[00:40:11] are two positive externalities. The domestic alternative gets more productive the more they use it and so does the global technology. I want to subsidize use of both in proportion to the strength of the externalities.  
[00:40:25] Okay, that's a very intuitive solution for the planning problem. So just to wrap this up um when can you use it? Um first on you're thinking about you know having convex optimization so that you  
[00:40:39] know that the intermediar is uh problems have a unique solution you have to have complete instruments you I assume that the planner had enough instruments that it could directly dictate the choices  
[00:40:54] okay the quantities in the allocation space under those conditions you can use it and we we're going to use it all over the place Okay, if you haven't seen this a lot, I suggest spending some time with it and getting familiar with it. It's a  
[00:41:09] very useful tool. Okay, so let's take this one here. We know what happens in this model if we go through a standard benchmark of a planner.  
[00:41:19] Just to take it for a spin, let's consider another benchmark. Before we go back to our game with the hedgeimon coercion, let's think of a world where there's no hedgeimon, okay? And every country is choosing  
[00:41:31] domestic policy in an ash game. So every country is choosing it as domestic wedges taking as given everybody's else policy choices.  
[00:41:41] Okay. And for simplicity let's take the limit where there are infinitely many small open economies. So everything is essentially taken as given by each country. Okay. Good. So let's look so we're going to do a a primal approach  
[00:41:55] again. So we're going to look at the problem of each intermediary.  
[00:42:01] Okay. And we're going to say look the same derivation as before are going to dictate that each intermediary uh sorry each government fully subsidizes the domestic technology in  
[00:42:14] exactly the same way that the planner would have done it. You can go through the math and I strongly encourage you that you do it and convince yourself but intuitively it makes sense. It's an externality that occurs entirely within  
[00:42:27] the border of that country. Therefore, the domestic government fully perceives it and fully internalizes it.  
[00:42:37] It's also intuitive that if you were to look at how much does each government decide to tax or subsidize the usage of the global technology, the answer is zero. It neither taxes nor subsidizes.  
[00:42:52] Where does that come from? It comes from the fact that each government perceives the global technology to be taken as given. It it perceives that this country is too small to affect it and therefore neither taxes nor subsidizes it. Okay,  
[00:43:06] this is special. It comes from the continuum result. What is generic in this planning problems which shouldn't be surprising to you is the fact that uh you normally undert tax or under  
[00:43:18] subsidize uh externalities that occur across countries when you're doing a Nash game because you don't see part of the externality. So if it's a positive externality you end up under subsidizing it. If it's a negative externality you  
[00:43:32] end up under taxing it. You underreact because you don't see part of what's going on. Okay? That's the only generic result. Good.  
[00:43:40] Okay. Uh so that is what I just said. Let's stick this one here. You can see that the way I'm building it, I'm keeping the right hand side constant and all of the variation is coming from here. Okay, it's going to make us easier  
[00:43:54] to compare the cases. Now let's go back to what we wanted to solve. So we figured out the tools, we figured out two of the basic benchmarks, we did the warm up. Now we can go back to the to the astral thing we want to solve and we're going to solve it in two stages.  
[00:44:08] First we're going to solve the offense what the hedgeimon asks for. What is the optimal choice of the tows that the edgeimon imposes on foreign entities and what are the optimal transfers. Then  
[00:44:21] we're going to do one step backward in the induction and figure out what are the optimal policies that each country pursues on its domestic economies and its domestic firms. Okay, let's start from here. So first let's do it  
[00:44:34] visually. Okay, so the think of the intermediary in country I we've assumed constant prices.  
[00:44:43] So the marginal revenue curve is flat. Doesn't matter how much it produces the price is constant.  
[00:44:50] The marginal cost curve it's increasing because we've assumed that the production function is decreasing with scale. So the value of the intermediary is this area is the color area in the  
[00:45:03] middle. Okay, this is like basic micro at the inside option.  
[00:45:10] This is the total value at the outside option because we're focusing on restrictions to the intermediary that they cannot use a foreign input. The marginal cost curve has to weekly shift to the left. How much? We don't know.  
[00:45:24] Oh, I mean it's it's easy to figure out that we use a CS aggregator. So if the inputs are perfect substitutes, it doesn't shift at all. As the inputs become cop Douglas, so a signal of zero, this thing blows up. Right? If you don't  
[00:45:39] have access to the foreign input, you can't produce at all. Okay, good.  
[00:45:45] So the firm let's assume for a second which will turn out to be true that the hedgeimon always makes the participation constraint bind. It's kind of intuitive because you can imagine that there is no  
[00:45:58] point here the way we design it to leave any slack to the target. So whatever luck you have you extracted with the transfer. Okay. So the firm or the target country retains the blue area the  
[00:46:09] outside option. the hedgeimon strikes the middle area.  
[00:46:16] That's going to be the problem. Remember, I started this by saying, look, we're kind of interested in figuring out is the pie shrinking, is the pie uh getting bigger, is it constant? The fundamental conflict in  
[00:46:28] this set of papers between the hedgeimon and the countries is a level versus differences.  
[00:46:35] Power, it's a gap between the inside option and the outside option. If I'm the hedgeimon, that's what I care about.  
[00:46:44] If you're the target, your welfare is a level is what you're left with.  
[00:46:49] That creates the issue. Why? Because imagine that the planner the sorry let me start this again. Think of the planner. The planner what he's done is that he has  
[00:47:03] made sure that he made that colored area as big as possible. He has shifted the inside option as far high as as he can.  
[00:47:12] Now imagine if the hedgeimon starts there. Okay. On the one end is good. You can see that given an outside option pushing the inside option higher. It's  
[00:47:23] good. So in these papers the hedgeimon has a feature that comes on the work of Kindleberger of being positive. It wants to provide public goods. wants to make the technology better all else equal  
[00:47:38] moving everybody else inside option up it's good because I can extract it but now suppose that it starts there and it checks the outside option okay what is it going to do it's going  
[00:47:51] to now check whether moving somewhere else is going to maximize the gap in particular it could be possible that I can p strategies that move the inside option down. So they're destroying value  
[00:48:06] on path, but they move the outside option down even far faster. So we're creating a gap. That's going to be what we discussed before the drug dealer model. I got you all to be just on my  
[00:48:18] system. You all are addicted to it. Your outside options are really bad. I might have wanted to do this inefficiently.  
[00:48:26] uh even if it would have been more efficient all else equal for you to have some alternative I just made sure that the equilibrium I prefer is one where you have no alternative and what it what that equilibrium really is which we're going to study in a second is one where  
[00:48:41] the inside option has come in but the outside option has move even faster down okay and that's a nasty aspect of the hedgeimon that in practice it might be willing to destroy value on path uh if  
[00:48:55] he helps his extract more power build up more power. Okay, so keep this one in mind. This intuition is very generic in these papers.  
[00:49:03] Okay, now let's do it formally. So let's start from um the offense. So we're going to do again the primal approach. We're going to say that the hedgeimon can directly dictate uh the  
[00:49:16] answers of what the intermediary should do and then solve for the wedges. So the wedges are going to be here.  
[00:49:24] There is one word of caution now which is this applies to the inside option but not to the outside option at the outside option the country that I was trying to bully said I'm sorry but I'm not going  
[00:49:37] to deal with you so I no longer have any way to influence them they're not accepting the wedges but on the inside option since they are accepting what I asked them to do and we set up the paper so that effectively you still have complete instruments we can still use  
[00:49:50] the primal on the on the inside option that's what we're going to So we're going to say look what am I trying to do? I'm trying to maximize transfers that I get from these countries because in the simple example that we have the profits of the  
[00:50:04] hedgeimmons are constant and the factor payment is constant because everything was produced linearly and the prices are constant. So it's all about extracting transfers subject to the intermediary participation constraints. I wrote them out extensively so that you can see the  
[00:50:19] inside option and the outside option. And you can see that at the outside option, we have to take into account that I cannot do the primal. They're going to independently choose what the what is good for them. Okay, on the inside option, instead these ones, we're  
[00:50:33] going to assume that the hedgeimon chooses them directly. Okay, that's the problem that you need to solve.  
[00:50:39] Okay, so what you do is first you figure out that indeed uh the transfers are always set so that the constraints bind. That means that you can substitute the transfers out with the participation constraint. Now you're  
[00:50:54] left with this the inside option and the outside options.  
[00:50:58] And then here's where it goes. Okay, I'm only going to do the intuition. So the hedgeimon is internalizing at the inside option that as I get more people to use my technology, the  
[00:51:12] technology is going to be attractive for everybody else. Okay? So then I'm going to fully internalize. Notice that the way we set up the model, this is not affecting the outside option because at the outside option, they don't retain  
[00:51:24] any access to the technology. Okay? And that's going to matter for the particular result that we're going to get. Okay?  
[00:51:33] Now look at what happens when the planner sorry the hedgeimon is choosing how much they should use their home alternative. there is going to internalize not only that as they use it  
[00:51:47] on path that is increasing the on path value for everybody else in that county to use it but also and more importantly that's going to affect their outside option. The planner, the hedgeimon  
[00:52:01] understands that as he tells each of the countries to scale down their usage of the domestic alternative, it's destroying some value at the inside option because it's making it less  
[00:52:14] productive, but it's also making the outside option of everybody else a lot worse. Okay, so that's going to be the game. So if you solve it, which it's again doing the same exact math, I just  
[00:52:28] did it fully for the planner because it was so simple that you could do it on slides. This one is just the same process, but you just have to look at the equations a little longer. If you do it, you end up with this. Okay, so what does he do?  
[00:52:41] The offense is I tell you that you have to increase how much you use on my own system.  
[00:52:49] And in fact I tell you to do it up to the anti-coercion wedges just as much as the planner would have. That's a sense in which the paper has an edgeon that all else SQL is pretty benevolent. It does exactly what the planner would have  
[00:53:03] done. It says look I want to this this global technology is efficient. I want you to use it.  
[00:53:09] The problem and where there is a difference with the planner is that I'm going to tell you not to use the alternative. I I'm going to tax it. I want you to have no outside option. That is inefficient  
[00:53:22] from the planner perspective. It's very efficient for the hedgeimon. It's optimal for the hedgeimon because of the gap mentality. Okay. And so I'm going to tax it. Okay. This one we're done. So that's the offense. Now let's do the  
[00:53:35] defense. Okay. So now we're taking a step back and we're saying, okay, each country understands that this is going to happen in the middle. They can choose domestic wedges to shape their economy  
[00:53:48] in a way that is going to affect the equilibrium in the middle.  
[00:53:53] Visually, what's going on is that anti-co engineer takes a very simple form for now, which is I'm trying to push the outside option of my firms up.  
[00:54:02] I know they're going to get bullied. I want to make sure that I'm not so dependent on the hedgeimon. I want to make sure that we have good outside option. You can see it's very practical.  
[00:54:11] Right now if you open a newspapers there's lots of news on we're trying to decrease our dependency on the US if you're Europe or on China if the the US everybody's thinking I have all these dependencies to foreign big countries and I'm worried that they're going to  
[00:54:25] squeeze me and so what I'm doing is I'm going to try to build up my alternative.  
[00:54:29] China thinks the same about the US and says look I want to have my own payment system because they might use it against me. So that's exactly the logic that we're going to exploit. Okay, so let's see it in this particular example. If  
[00:54:42] you work it out, um it's pretty straightforward and we you will see that I rigged it in a particular way. So the outcome is you ban dealing with the hedgeimon. You tell your firms you're  
[00:54:56] going to face an infinite tariff if you use any uh of the global technology and then you subsidize to the efficient level the domestic technology. Okay, the result is  
[00:55:09] entirely specific to the setup. It generates full fragmentation and you will see in a second why I want it. In general, what you're going to get is I'm simply going to want to walk away some of the dependency uh to make sure that I I maintain more of an outside option.  
[00:55:24] Okay, now let's see why it occurs. Thinking about the outside option, look at the optimal anti-coercion. So the key observation is that the count's  
[00:55:38] intermediate is outside option directly depends only on the productivity of the alternative. So what gives me the infinity the full bun is the fact that there is no advantage of the outside option from anything you do at the  
[00:55:50] inside option. Okay. As soon as we break the set up and allow that some of what you retain on the outside option depends on you participating on the inside option you move away from that result.  
[00:56:02] Okay? Uh but that's where it comes from. Okay, good.  
[00:56:08] This one I think is pretty straightforward. Let's just skip it.  
[00:56:11] Okay, so here we are. Okay, now we have the full solutions. So you can see that I I did the setup to have the home alternative always face the same policy no matter what. So that we compare what's happening on the other side on  
[00:56:26] the usage of the global technology. So if you think about it in the middle, the hedgeimon was hyper globalizing. The hedgeimon was trying to generate more usage for its own global technology compared  
[00:56:40] to the alternative than the planner would have liked. So one message is that a hedgeimon in this setup wants to globalize much more than a planner would as long as the globalization makes it  
[00:56:53] very central. Okay. The second part is that okay if you go one step backward and you think a country is anticipating it the message here is pretty stark it's telling you that the hedgeimon has all the power  
[00:57:07] I gave it exogenous powers to make these offers even has all the bargaining is making take it or leave it offers but we rigged it so that it was so coercive and was generating too little value for  
[00:57:22] the rest of the world so that exante every country moves way you get full fragmentation and in equilibrium the hedgeimon has no power in this model in this outcome in this particular incarnation the hedgeimon has no power  
[00:57:36] in equilibrium because he has no relationship with anybody and that goes to the question that one of you asked me at the beginning about commitment this isn't commitment to the threat to carrying out the threats but here you can start thinking okay maybe I would  
[00:57:50] have been better off if I were the hgeimon to commit to only limited bullying or to commit to find ways to leave you some surplus here. You're so scared of me that I end up vanishing my  
[00:58:02] entire power in an attempt to bully you all the time. Uh and so that brings up something that we can think about which um you know for those of you that take uh Kyle's um trade class at Stanford  
[00:58:16] should be familiar because Kyle and and and Bwell um sorry Kyle and Stiger Rob Bob Ster um did a lot of work on this with the WTO which is for a long time me personally I was convinced that if I was going to write a  
[00:58:30] paper about an international organization I was going to start with the planner problem and then think about the organization as a way to implement the planner. Now I kind of change my mind. I think the right way is closer to  
[00:58:44] what Kyle uh and and Robert did or close to what the political scientists have been pushing which is to think of the problem as I have a hedgeimon and he's trying to design a multilateral deal  
[00:58:58] that constrains his own actions but it's entirely self- serving. Uh and so I really come to think of the view of the IMF, the WTO. These are deals that the US offers that informally have the  
[00:59:11] following flavor. Uh you can be in my sphere of influence and yeah there will be some bullying uh but you're going to retain a lot of value and I'm constraining my hands to convince you that you should participate. Uh and you can think a lot of what has been  
[00:59:26] happening in the last few years has been a massive or a drastic rethink of that offer. Uh so let's just sketch it. Okay.  
[00:59:33] Suppose that you take the previous model and you say look what I'm going to do is commit to only extract a fraction of the inside option value as transfers. Okay.  
[00:59:44] So I'm not going to extract it all. I'm going to leave you some. It's like a bit of an equity contract. Okay.  
[00:59:50] Well, it's very clear that if I set the fraction to zero, then the hedgeimon doesn't care because it's getting nothing out of this. If I set it to one, we go back to it's too strong. You don't trust me. Somewhere in the middle, you  
[01:00:03] can prove that there is an equilibrium where they don't do full fragmentation.  
[01:00:08] They actually have a relationship with you. Hence, you have some power and you extract positive rents and you're better off. Okay. So I think it's really a view of commitment uh and multilateral organizations are sort of limiting  
[01:00:21] coercion and I think it matters and I suspect that you know in the next five years there'll be lots of papers in this space uh and if you're a student thinking of for example when I have two hedgeimons with the US and China what kind of deals that there should offer  
[01:00:36] what does commitment look like when they're competing all of that there is surprisingly little formal work or even like I would say good thinking like clean thinking So it's a good idea.  
[01:00:47] Okay. Um before I go to much more ugly formulas, I want to do some intuition and then I'll show you the full formulas. To me, the one that I really like are these first two.  
[01:00:59] So what did we really learn from these papers? We learned uh that there is a fundamental tradeoff between the gains from trade and economic security. Think of the standard argument of the gains  
[01:01:13] from globalization. They arise from specializations, external economies of scale. That's a Krugman argument. Uh it's me and Zoule.  
[01:01:21] We specialize in different things. We both get good at what we do. Uh that's very efficient and then we trade with each other and that leaves us better off. Inspect that argument again.  
[01:01:32] Hinder in that argument is that the markets exposed are not contestable.  
[01:01:38] There is no hold up. If I pay the price, I get the goods.  
[01:01:44] What if Zoule can hold me up and say, "Look, expost. I'm not going to sell you the things that I produce." Now, my problem is that I specialized the same exact thing that was generating  
[01:01:58] against from trade is also generating the fact that now I have a very poor outside option. The technology that I didn't specialize in, that she specialized in, I'm now not good at.  
[01:02:10] And that tells you that the trade-off is deep. It's not accidental. It's not as if there are two forces and somehow they go in opposite direction. No, no. It's the same force and it comes from thinking about the hold up problem.  
[01:02:24] So he's telling you that there is a deep tradeoff between the gains from trade and economic security and that part of what we're doing now globally is that we're rethinking about where we want to be on that surface and that we might want to be for example  
[01:02:38] in a place that is more secure at the cost of giving up some of the gains from trade because we perceive uh that the hold up problem is there. Okay, so that's really what these papers deliver.  
[01:02:50] It's just a different take on the basics of trade theory once you're thinking about a Germanic powers in the particular context of this. It also has a nasty feature because if you  
[01:03:04] solve the game remember that here anti-coercion policy is being done by each country taking as given the anti-coercion policy of everybody else.  
[01:03:13] So they're still doing Nash exante. Okay. What does it means? It means that you can show that there's a bit of a fragmentation doom loop here. Each country starts to say I don't I want to pick a different point on the frontier.  
[01:03:26] I want to be more secure and pulls out of the global technology a little bit.  
[01:03:30] But as that country pulls out, the technology has become less attractive for everybody else. So they pull out a little bit more. But as they pull out, the original country wants to pull out even more. So you get a bit of amplification. You get the fragmentation doom loop. So you can end up in a in  
[01:03:44] with an equilibrium where you're overly secure but you gave up too much of the gains from trade and it would have been better off doing some coordinated policy. Okay, good.  
[01:03:55] Um the last few slides before we switch to empirics um let me start with the general theory.  
[01:04:03] So if you remove all of the special assumptions that I introduced, all of the structure of CS, the fact that we switched off the terms of the utility function, the fact that we made prices constant by having linear utility, if you go back to the first five slides in  
[01:04:17] their full generality and you solve what does it look like? Well, first of all, no close forms, but also you have many more terms. Why is it useful? It's useful only because it it's helping you  
[01:04:30] capture what are the standard forces and what's new here. So what are the standard forces? So think of the hedgeimon in the middle. So this is what the hedgeimon is imposing as a wedge on  
[01:04:43] what I is buying from J. Okay. So as I do it the first thing that is happening is that as I for example if I subsidize it as that activity increases it's going to  
[01:04:56] feed to the entire input output matrix through the level of technology to the level of all the aggregate zs okay that's going to be like a gigantic uh sort of propagation matrix like a leontf inverse  
[01:05:11] why do I care about it well this is the standard terms I care about it because it ultimately affects the profits of my firms or I care about it because it directly  
[01:05:24] impacts a term in the utility function that I care about. So let's make an example. I'm looking at semiconductors and I'm imposing a tax that says I don't want you to export these semiconductors  
[01:05:36] to China. Why do I care? I care because the semiconductors are going to impact all of their downstream industries. all of the other countries and all of the input output matrix. But ultimately, I only care for two reasons. Either  
[01:05:51] because it's gonna make my own firms more profitable or because it's shrinking activities that I didn't want to occur. For example, you can think of like a a downstream military application. I want  
[01:06:05] that military sector to be small. I care to tax semiconductors exports to China because ultimately with the input output matrix I know that that shrinks the military application. Okay. And that's a sense in which I said look despite the  
[01:06:19] fact that these utility terms are totally made up. They're useful only in seeing how they would play out in the equilibrium. Okay. So these are pretty standard.  
[01:06:30] The other term that is even more standard, you've all taken a trade with Steve Reading, so I know that you know this, which is terms of trade manipulation. Now I have prices moving around. I care about manipulating an activity if it does the following. It  
[01:06:44] increases the prices of what I export and it decreases the prices of what I import. As you know, like for example, the optimal the classic formula for the optimal tariff is essentially one that aims to manipulate the terms of trade, right? and I'm trying to get the  
[01:06:58] foreigners to lower their prices for what they um what I import from them.  
[01:07:03] Okay, that's a standard term. Then I have some private distortions here that come from the fact that there were some pre-existing wedges. These you can ignore. Okay, what's new? What what are the terms that are not there in the  
[01:07:15] standard analysis is these terms here. That's what we've been focusing all along is building power. In fact, in the previous one, if you think about this, those terms were all zero. this by assumption because I make the utility  
[01:07:28] term not be there. This one because I made the linear technology being the only thing that the US has so there is no way to affect the profits of the US.  
[01:07:37] And this one because we simply said let's make the consumer linear that's not there. So in the previous mo in the in the incarnation I gave you everything was coming from this term. Now what are  
[01:07:50] they? as I impact the activity J that's going to impact what everybody else is doing through the propagation in equilibrium for all of the foreign firms that  
[01:08:04] reflects on the value of the inside option and the value of the outside option this is the gap why do I care about it that's my power in the previous example that's all I extracted that was the transfer so now you can see formally  
[01:08:18] that what I'm doing is I'm manipulating activity I A they themselves only care about their cost to them privately of that activity. I take into account how that one spills through the entire  
[01:08:31] equilibrium and how it affects the inside and outside options of everybody else and that's the entire previous example was characterized in this term building power. Now the same has to be true if I  
[01:08:45] go through prices. If I impact activity J through the equilibrium amplification that's going to impact the vector of general equilibrium prices given the change in prices that's going to impact  
[01:08:59] the inside option and the outside option of everybody else and it's entirely coming from the participation constraint. So now you can see that we've gone full circle. I promised you exanteed that there were three things to pay attention to the  
[01:09:13] participation constraint. You can see why it's all our definition of building power. The wedges as costly actions.  
[01:09:22] It's what I'm solving for. And the externalities and the equilibrium amplification. That's entirely where the effects are coming from. DZ DX DP DX.  
[01:09:32] That's where it's coming from. All of it.  
[01:09:36] And you can see that it's just a separate term from the rest. Now, they interact. So, here's an interesting one.  
[01:09:41] Uh think about the typical optimal tariff. What I want to do is I want the foreigners to lower prices of what they sell to me. But that might be in conflict for example with power building. I might want to live into a world where I sell my stuff very very  
[01:09:56] cheap. It gets everybody to use it to not build an alternative and then I extract power by threatening individually to kick you out. Okay. But  
[01:10:07] that's where they come from. questions. Okay, good. So, if you do anti-co, it's a pretty similar formula. So, we said look, there is lots and lots of  
[01:10:21] reasons uh to domestic policy in these economies. So, we said look there's suppose that we totally ignore the hedgeimon. We already know that there's plenty of reasons to do intervention. Uh you might for example want to sort out some domestic positive or negative  
[01:10:35] externalities. You might want to try to manipulate some prices. Those are here.  
[01:10:40] I care about intervening because it changes some activities that I control.  
[01:10:45] It feeds through the Leontf inverse and ultimately I only care if it's spilling back on my firms.  
[01:10:52] I said that we were going to focus on a distinct component which we call anti-coercion which is I care because I understand that by changing that activity I'm going to change the request  
[01:11:05] that the hedgeimon would make next period and those requests impact the profits of my firms particularly at the outside option. So that's the term that we characterize in the previous example is this object.  
[01:11:18] Okay. Okay. Good. So for the last part we're going to do two things. First we're going to take everything we just did and bring it to the data uh in a sufficient statistics approach. So if you're taking a trade class this is very  
[01:11:32] very familiar to you. We're going to use um structure from a model and the data to figure out counterfactuals. Uh you know a lot of what we focused on right now has been the gap between the inside and the outside option. the outside  
[01:11:47] option we don't see it in this paper for example in that theory is never realized so the question is what how can I use what I see about the inside option to infer what would have happened at the outside option to make it practical to  
[01:12:02] imagine that you think about what happens to me if I lose access to Chinese raers okay and it's a question that is very familiar to trade economist because one counter fascia that they're very  
[01:12:15] interested in is what happen if I shut down trade in its entirety. So what if what what is the cost of autoarchy?  
[01:12:21] Okay, here we're not going to go to autoarchy. We're going to go to losing a particular collection of inputs. Uh but a lot of the same technology is going to be helpful. So we're going to do a bunch of that and then we're going to switch gears and look at like what we're doing  
[01:12:34] with AI, okay? and thinking about like if if there are threats and there's a lot of uncertainty and maybe the asks are um you know difficult to prespecify can we use non-standard sources of data  
[01:12:47] to figure out who's doing what to whom. Okay, good. So let's start with this one.  
[01:12:53] So let's do a warm up to remind ourselves of what how this works. Okay, I'm sure you've seen it but let's just do it once in a simplified setting and then we're going to scale it up. So imagine that we go back to a simple  
[01:13:06] example of I have a production function that produces with decreasing returns to scale beta xn is a cs basket of intermediate inputs j  
[01:13:20] and here's my last substitution. So you can see that if I set sigma to one I'm getting to cop daglas in cop daglas if I lose any of the varieties j the whole basket blows up I can't produce anything. So anything more complimentary  
[01:13:33] than Cob Douglas will imply full loss of production if you lose any one of those inputs. Again to remind you of the analogy with trade, that's part of the reason why you get infinite cost of ordery if you're in a sort of cod dagla  
[01:13:46] setting with with foreign inputs that defines a price index that you're familiar with and it defines expenditure shares which are the fraction of the expenditure that fall  
[01:14:00] that falls on particular inputs. Okay, what we're going to play again and again with is this is that the log profits of this producer  
[01:14:13] are a constant plus the log of the price index. So it's helpful to think of uh everything that is going on as variation in the price index for you know before we phrase it as variation in the  
[01:14:27] marginal cost curve. Here we're going to keep keep track of that as a price index for the optimal set of inputs. So if I'm losing some inputs, for example, the price index will tend to go up. Okay. So  
[01:14:41] everything can be traced as this. Let's have a look at this. What are we really doing? We're looking at the log between the inside option producing with the full set of inputs versus the log of the  
[01:14:56] outside option producing with only a subset of inputs. It's exactly what we've been playing around all the time so far. Well, based the previous formula, that's going to be the log ratio of the two price indexes. And so  
[01:15:10] you can see that we're going to start thinking about how bad it is to lose uh some uh inputs in terms of the effective price index that I'm facing.  
[01:15:22] We write it out. We get our nice um expenditure shares. Uh it tells me uh that I tend to lose more if the inputs that I were using were a bigger fraction  
[01:15:34] of my expenditure on path. and of course with the elasticio substitution as the elastio substitution gets goes to co douglas and we're going to define this object which is the fraction of the  
[01:15:48] inputs that I'm losing is the fraction of the expenditure share in equilibrium that I was sourcing in some sense from the hedgemon okay it's what he can it's what he can  
[01:15:59] take out of me okay so this is familiar to all of you you've done this derivation before. Um, just breathe a second, let it sink in,  
[01:16:13] convince yourself that that's what's going on, and then we're just going to take it for a spin. What we're going to do is we're going to nest the CES. Okay?  
[01:16:22] We're going to do multiple nests, and we're going to start thinking about how that affects the ultimate outcome. Okay?  
[01:16:30] So, what are we going to do? Let's start from the bottom. The bottom are for each particular sector what am I importing of those inputs from all other  
[01:16:44] countries. So these elasticities are going to tell me how substitutable are two foreign sources of the input among each other.  
[01:16:58] So if you go back to our example of oil, we might think that oil has a very high sigma that Russian oil, Venezuelan oil and Canadian oil are a little different  
[01:17:11] uh but ultimately quite substitutable for me. So if I can if I get cut off from any one variety, the other varieties are going to be great substitutes. Okay, I'm going to move the price index a little bit against me as I  
[01:17:24] try to reallocate to those, but not a lot.  
[01:17:29] Then we're going to have a layer for each of these sectors that mixes the basket of the foreign varieties with the domestic variety.  
[01:17:42] The reason we're going to allow it is for two u separate but important reason.  
[01:17:48] One is that you might think genetically that because of selection transport cost because of things that we're not going to really model the domestic variety and the internationally traded ones are a little different. You know, I think of  
[01:18:01] it as for example trivially. Let's suppose that we're doing consulting services. Domestic services might include a lot of very trivial stuff. But if we're doing an international project and there's a lot  
[01:18:15] of services crossing borders, probably like some very high value added consultancy that is doing it and that might not be very substitutable. Okay, it's just a simple reason intuitively do it for us. There's a different reason to  
[01:18:27] also do it which is um if you're thinking about being cut off.  
[01:18:35] One thing that matters a lot is the size of your domestic economy. The you know the first thing that makes you secure is you can produce it yourself. So the home share is very important particularly if you are very large countries for a  
[01:18:49] continental size economy like the US, China, Western Europe, the domestic share is a very big deal and this is obvious to the trade uh community because again if you go to the calculation of the cost of autoarchy it  
[01:19:03] tends to be relatively small for countries that import very little that they have a huge domestic economy it tends to be very high if you're Singapore because so much of your GDP comes from the important the export but  
[01:19:15] that's going to be quite important and keep this in mind because I'm going to argue that this is the hardest part to measure and that it leads to a little bit of a  
[01:19:26] policy mistake uh because the generally at least for manufacturing goods um the foreign data is surprisingly much better. Why? It cuts on the simple fact that every country for the last 200 years is at a customs office because  
[01:19:41] we've had tariffs. So whenever a manufacturing good crosses border, it generally leaves at race. So the data for that is surprisingly good. That means that a lot of the time gets spent on computing how much do we import from  
[01:19:55] China. What fraction of this whole thing comes from China or China does the inverse with the US.  
[01:20:01] But suppose that I tell you that 90% of what we import in this particular category is from China. Well, that might still generate a tiny dependency if overall 90% of the total usage in that category is coming from the domestic  
[01:20:16] share. Okay? And but somehow because we don't have that data for many goods, that doesn't quite get as much attention as one would think. So keep that one in mind. Okay? So now we have a composite  
[01:20:29] basket that includes the domestic production.  
[01:20:33] Now we're going to go and aggregate across industries.  
[01:20:36] Okay. So we're mixing I don't know electronic equipment with cars with other things. And then what I'm going to do is I'm going to at the very end take this basket of all other goods and mix  
[01:20:50] it with finance. And you will see in a second why I'm keeping finance separately. I want to think of a world where those are basic financial services and where essentially little gets done uh if you don't have the ability to make payments where it's sort of cop almost  
[01:21:04] cop Douglas in everything else. Okay, you can nest in many different ways depending on how you want to set up the economy. But let's explore this one.  
[01:21:14] Okay, now take the previous formula and just redive it with many nexts. It's the same thing over and over again, right?  
[01:21:23] The the trick is precisely by thinking of them in price indexes. You're starting from the outermost and then you keep nesting and the formula just keeps going that way. It's one of the beauties of the nested CS. So the final log  
[01:21:38] differences are this. That's going to be our notion of power. It's the notion of how much of a in percentage loss of the value added of the target can I inflict if I restrict the inputs.  
[01:21:51] The red ones. Let's start from the big omegas. These are on path expenditure shares. For example, this omega here tells you what fraction of the expenditure share of the target are on  
[01:22:05] the basket of all other goods and finance. The second one here tells you for each of the baskets of manufacturing, what fraction of the expenditures are on equipment, cars, and so on.  
[01:22:19] This one here, it's the one that we were just discussing. is telling me what fraction within that basket it's coming from foreign varieties versus domestic.  
[01:22:29] So if omega this big omega is very close to zero it's telling you that almost everything is coming from domestic production and then I have this little omega this  
[01:22:42] tells me for that particular basket among the foreign varieties what fraction is being controlled by the hedgeimon so if you're thinking of the US as the hedgeimon is what fraction am  
[01:22:55] I importing from the US or you can flip it and make China your regiment and say what fraction of everything that I import is coming from China. So this little omega is what gets a lot of attention for example in the press or in  
[01:23:08] the policy circles when we say that 90% of process come from China or think statements like that or generally about this and this does matter and we're going to explore it in a second but I want to convince  
[01:23:23] you that that's not the only thing that matters and then in particularly trivially the domestic shares are quite important let alone of course how it aggregates up you will see in a second that I already preview that we're keeping finance separately because we were likely to shift the aggregates  
[01:23:38] quite a lot. Okay, good. Now this is the trick of the sufficient statistics uh in three. This is really what ACR um did um Kolakis Costino and  
[01:23:51] Rodriguez Clair. Um these shares are equilibrium shares.  
[01:23:58] These I can measure. They're expenditial shares in the data. What fraction of all things in this sectors are import from China?  
[01:24:07] The blue objects is the nasty part. Those are elasticities.  
[01:24:13] And since you have taken the trade class, you know that that's not good news because we tend not to have great estimates of elasticities. There's lots of estimates all over the place. Uh and particularly as we disagregate, as we go  
[01:24:27] to finer and finer things, it gets harder. Imagine trying to estimate what is the elasticity of substitution between a frontier semiconductor from Nvidia and one that is just below the frontier from Huawei.  
[01:24:42] That's hard. Okay. Um but it helps you to constrain your own view and debate of what do you have to believe for example to get big numbers or small numbers. And so I find it very  
[01:24:55] helpful even if there is substantial uncertainty about these objects. So to make it simple uh we're going to start with making the outer nest coagulas. So we're going to say that if you have no finance at all  
[01:25:10] if you have no ability to make payments you can't produce. Okay.  
[01:25:15] It's both because we want it to be important but also because of something that people refer to as the idiot law of elasticities which is if you don't know the number set it to one uh and do and do coples. So we're going to do that. Uh then for the rest we're going to take  
[01:25:29] from the handbook chapter from Arno um and and Rodigas clear we're going to take uh anio six for all other sectors.  
[01:25:39] We're going to assume uh and take it a little bit from the literature that finance is special. We're going to assume that it's very difficult to not use Zift. It's very difficult to not use the US financial system. So, we're going  
[01:25:52] to give it a very low elasticity of substitutions like in the 1.7 range. And then this is what I mentioned. We're going to borrow an assumption that the domestic variety is less substitutable with the foreign basket that each  
[01:26:07] foreign variety is with itself. Okay. And this comes from a paper from Finstra and Moriosel.  
[01:26:14] big data cities the way you like them run to the formula. Okay.  
[01:26:19] Okay. So first let's explore these nonlinearities. Okay. So let's focus on the inner bucket. This is the price index of movement of being cut off from the hedgeimon at the inner bucket level  
[01:26:33] at the lowest level of the baskets. So what we did is we picked all bilateral omegas across countries visav the US and China across sectors and these are the  
[01:26:46] keel muted densities. So first stare at this index it tells you that there are two very important nonlinearities.  
[01:26:55] The losses become infinite when little omega approaches one where the entire basket is being cut off. Why? because it's telling you that the way to get the firm not to use it at all is to make it  
[01:27:09] perceive that the the good cost infinity and we sell it already when the sigma gets very very close to one. Now this one which is a standard feature of CES I'm going to argue is very very important  
[01:27:24] for geoeconomics and has some pretty interesting implications for the nonlinearity of power. But first let's convince ourselves that it's somewhat relevant. For example, if you look at financial services, it's no surprise to  
[01:27:36] you that there are many bilaterals where 80 90% of all imported financial services come from the US and its allies. Essentially, you're most importing financial services from the US, Belgium, Luxembourg, London,  
[01:27:51] Germany, places like that. It's very very concentrated. When you say that China it's a very large exporter in a powerhouse in manufacturing is this one here that is kind of shocking or amazing  
[01:28:04] that for a single country the import dependencies in the fraction of goods that get imported from China is very often in the 20 or 30% from that single destination. Okay. But what's important  
[01:28:18] is that this you can see at least in this functional form is going to really kick in for very high u shares like in the 0.95. That's really when it kicks in.  
[01:28:31] Let me pause for a second on this because I want to mention two things. So why do I care so much about this? Well, first because it's telling you and I'm going to get back to this when we do AI.  
[01:28:40] It's giving you a definition of what for example a choke point is. what what dependencies do you care about? It's telling you that the ones that you have to be scared about are the ones where there is a very dominant player where there is almost no alternative and the  
[01:28:54] little omega is very close to 100%. Or places where the lasio substitution is very very low something where technologically is very hard to do without. So you can see that you know radars advanced semiconductors  
[01:29:09] finance are going to smell a little bit like that. there sectors that have uh some of those characteristics. And we might argue that we don't know exactly the elasticity for semiconductors or how well do you really control the entire supply chain. But it gives you a very  
[01:29:23] simple way that an economist will recognize in two seconds for what a dependency is. Flip side. It also tells you that most sectors generate no dependency whatsoever. This is very nonlinear. So as soon as you move away  
[01:29:37] from that corner, power dissipates very very quickly. Um and so if you go back to the motivation that we had at the beginning Hman you know this formula is nothing else that 50 or 70 years  
[01:29:50] later with a lot of help from trade theory and macro theory a better version of what does it mean to have a dependency on a foreign country. Hman had in mind summing the square of the  
[01:30:02] little omegas is that dependency was um do I have a lot of trade with very few partners. This one says well there's something quite intelligent about that which is the precisely the concentration  
[01:30:16] but it's making it more complicated and it's saying well no no it's really about that to the power of an elasticity because I need to know how easy it is to rearrange technologically if you cut me off and of course that's not the only thing that matters I need to know can I produce it domestically and I need to  
[01:30:31] know how much impacts other sectors in the economy all of that but it's giving you like a you know in some sense a fullyfledged theoretical origin that delivers on concentration is very important. Uh it's just a different way.  
[01:30:45] Uh it it also has the message that you don't have to go to equal share to to get safety. It's just enough to move away from the corners which I think it's an important part. It also has one more message which we're not going to explore  
[01:30:58] today um but that we're thinking about in a different paper with multiple hedgeimons which is that power by being nonlinear also means that is not additive.  
[01:31:08] What does it mean? It means that if I lose power it doesn't acrue to you not necessarily. So let me give you the following example. Think of the dominance of finance and let's consider Singapore. Okay. What is the value to  
[01:31:21] the US of controlling Singapore? The value to the US is entirely on the offense.  
[01:31:29] If you control Singapore, given that you control already almost all other financial services, that's going to really put those little omegas close to one and that's very important to you. So you acrew a lot of power by controlling  
[01:31:41] this relatively small financial ser center. Think of the value to China or controlling Singapore. It's not going to make China a financial superpower.  
[01:31:54] But it has potential a lot of power from a lot of interest for China in diluting US power.  
[01:32:02] [snorts] But as you dilute US power because they lose Singapore, Chinese power isn't going up by the same amount.  
[01:32:07] So power isn't a sum. Okay, that's quite important when you're thinking about uh these games.  
[01:32:16] Good. Okay. So if you crank this out and you know Kiara is here and she's cranked a lot of these calculations for us. Um what does it look like? Well first for both coalitions clearly small open  
[01:32:30] economies that trade a lot with these countries tend to have big losses. So Singapore because it has no domestic shares or very small domestic economy compared to to its imports and is very dependent on these countries is going to  
[01:32:43] show as having very large losses. Uh what's interesting is that for the US uh coalition a lot of the power comes from finance. They're essentially controlling  
[01:32:55] something that has a very small share of gross output. So it's not a big sector but it's important for all other sectors and you control almost all of it and that gives you a big kick. For China is  
[01:33:08] entirely coming out of manufacturing um and it's a lot more substitutable in general. A lot of the things that China sells tend to be things with very high lassio substitution like basic durable goods. Not all of them. Rare earths is a  
[01:33:22] good example. Okay. You can do a lot of these calculations in many different ways. You can disagregate. You can vary the elasticities. There's lots of stuff.  
[01:33:32] Um I'm going to highlight a few that might be um of interest to you if you do research. So first the level of aggregation does matter and we've already sort of stumbled a little bit on it. Let's discuss what's allowed  
[01:33:46] and not allowed. So these are calculations that I would describe as the medium run. So think about in even in the in the theory what are we allowing when I cut you off I fully allow you to reoptimize  
[01:33:59] who you source from and how much. So it's not a very short run. It's not like a lemon identification for the financial crisis that the shock happens and you have no time to do that. If if we wanted a model of that, we might have done a  
[01:34:13] model where you sent out your orders all your XIGs for sourcing inputs. Some of them get cut off and you can't change the other ones. That would generate much bigger numbers in terms of losses  
[01:34:26] because it will say you inefficiently order little from the alternative because you didn't think you were going to be cut off. Now you've been cut off and you're left with a very small order.  
[01:34:37] Here we're not doing that. We're allowing you to potentially increase that order a lot. But there are costs to that. Now they're not very long run calculations because for example we're not allowing for the equilibrium  
[01:34:50] aggregates and lots of other things to change. Um we're also calibrating to elasticities that are probably more like medium run in the literature. You might think that for example the ability to  
[01:35:02] substitute might depend on time time to rearrange your production function time to do other things. What are other things that are on my mind to tell you as you do this? One is that the level of disagregation matters. Um in particular  
[01:35:17] suppose that uh the choke points these nonlinearities were really at the very fine level like process rate or those particular semiconductors. Well, if you're doing these calculations with very aggregate data, you kind of smoother them out, okay, they're mixed  
[01:35:31] with all other equipment or they're mixed with all other metals, you're kind of losing some of that. The problem is that as you disagregate, two things tend to happen. First, the quality of the data deteriorates quite quickly. The noise in the data goes up a lot. That's  
[01:35:46] particularly true for the domestic share because most countries traditionally do not produce data for the domestic production out of micro data. the data for cross border is much better and for cross border for example services are  
[01:36:00] notoriously badly estimated why because traditionally we don't do tariffs on services maybe Trump will change that but not even Trump so far um and we just don't track them very well so if you think of for example the power of the US  
[01:36:14] or China they're competing over information technology AI all of that is going to show up in services >> like the access to Google none of that we're very good are measuring it. The other problem is that of course as you  
[01:36:27] disagregate the elasticities become harder and harder to measure and you have to be careful because you you you naturally think this is come again from the trade literature. You actually think that as you disagregate the elasticities  
[01:36:40] are probably becoming higher and higher, right? If you go super disagregated, it's very likely that this particular blue shirt I only import from one country. But that doesn't generate  
[01:36:53] infinite losses because that's a perfect substitute for a blue shirt with a particularly different color from another country. Uh so you think that the elasticities also go up a lot. um you just it's where we are okay there's  
[01:37:07] a lot of demand for better work in this uh in this dimension um some of you mentioned the input output tables um you know for example a lot of what China controls might be entering the US  
[01:37:21] indirectly uh and so if you're looking at the bilateral direct shares that might not be the right content you might be also having to look at the Chinese content in Mexican exports to the US or  
[01:37:34] in Vietnam export to the US. Uh and there are techniques to do that. Uh there's just a lot more to be done. If you're interested in this kind of quantitative empirical work, uh I think there's just going to be like a whole  
[01:37:46] long wave of of work in this dimension as the data gets better and better. In the particulars of what we did, what are other shortcomings? Well, we mentioned the time dimension. Uh what are we  
[01:38:00] allowing to change and not change? What I haven't discussed very much is that to keep the game theory clean at the very beginning I kind of explained that here you're cutting off uh small entities.  
[01:38:15] We're not recomputing prices. That's probably not a great idea if you're doing a calculation of cutting off all of China. You would think that if we cut off all of China, uh exchange rates, wages, interest rates, lots of stuff  
[01:38:29] will adjust. Now that means that you need to use other technique. There are conditions under which you might get away with something that looks like these formulas. But eventually you might have to go to just simply computing two  
[01:38:43] different equilibrium in a in a quantity model. Now that's expensive um computationally but you know again computations are getting easier. Uh there's a whole area there that needs to be explored that this is only giving you