PARTICIPANTS
Matthew Jackson, John Taylor, Michael Boskin, Doug Branch, Pedro Carvalho, John Cochrane, Steven Davis, Randi Dewitty, Matthew Elliott, David Fedor, Andy Filardo, Jared Franz, Paul Gregory, Robert Hall, Robert Hodrick, Ken Judd, Robert King, Evan Koenig, Roman Kraussl, Robert Oster, Radek Paluszynski, Alvin Rabushka, Valerie Ramey, Isaac Sorkin, Richard Sousa, Tom Stephenson, Jack Tatom, Yevgeniy Teryoshin
ISSUES DISCUSSED
Matthew O. Jackson, William D. Eberle Professor of Economics at Stanford University, and external faculty member at the Santa Fe Institute, discussed “Supply Chain Disruptions, the Structure of Production Networks, and the Impact of Globalization,” a paper with Matthew Elliott (University of Cambridge).
John Taylor, the Mary and Robert Raymond Professor of Economics at Stanford University and the George P. Shultz Senior Fellow in Economics at the Hoover Institution, was the moderator.
PAPER SUMMARY
We introduce a parsimonious multi-sector model of international production and use it to study how a disruption in the production of intermediate goods propagates through to final goods, and how that impact depends on the goods’ positions in, and overall structure of, the production network. We show that the short-run disruption can be dramatically larger than the long-run disruption. The short-run disruption depends on the value of all of the final goods whose supply chains involve a disrupted good, while by contrast the long-run disruption depends only on the cost of the disrupted goods. We use the model to show how increased complexity of supply chains leads to increased fragility in terms of the probability and expected short-run size of a disruption. We also show how decreased transportation costs can lead to increased specialization in production, with lower chances for disruption but larger impacts conditional upon disruption.
To read the paper click here
To read the slides, click here
WATCH THE SEMINAR
Topic: “Supply Chain Disruptions, the Structure of Production Networks, and the Impact of Globalization”
Start Time: May 15, 2024, 12:00 PM PT
Transcript
John Taylor: [00:00:00] So why don't we get started. We're very honored to have Matt Jackson speak to us today about amazing title supply chain disruptions. The structure of production networks and impact of globalization. So thank you very much. You're the William Everly, I believe, the professor of economics, BA from Princeton, 1964.
No, 1984. I just got younger. Older and younger. I just found out that you got the same prize I got at Princeton 100 years before you. But you're at the Santa Fe Institute, the Department of Economics here, a distinguished teacher. So it's a real honor to have you here. And you're also your co author, Matt Elliott, who's up there on the screen, all the way from Cambridge.
And a little bit hours difference, but, welcome. We appreciate what you have to say. It's a complicated paper, let me say. Go [00:01:00] ahead, you have a handout.
Matthew Jackson: Thanks, John. It's great to be here. And I guess we Can just have this as a conversation. So people should feel free to hop in and ask questions or make comments whenever you like.
as John said, this is joint work with Matt Elliott. And what we're looking at is supply chain disruptions. And I guess it's, pretty clear. That, things have been in the news, especially COVID caused a lot of disruptions, but there's all kinds of other types of disruptions. And the idea here is to try to just develop a simple model that we can begin to use to track supply chain disruptions.
And especially understand short run impacts of supply chain disruptions. And then how they contrast with long run disruptions. and then say some things about how they how disruptions depend on the network and the complexity of the network. and then [00:02:00] say, use that lens to say something about what's happening with globalization and the changes in supply chain structure and why we might be experiencing different.
Forms of disruptions in the future. just, to, give the punchline up front. If there's nothing else that you take away from this, I think there's one basic thing that pops out pretty clearly from, what we do, which is that there's a big contrast between short run disruptions and long run disruptions.
So if you, change the technology, you disrupted in the short run. Basically that propagates downstream to all the final goods that are using that technology. So if you suddenly have chips not available than anything that's using chips is disrupted, and that can cause even a good that doesn't cost much can have a major disruption in the supply chains.
In the longer run, if your technology changes a little bit, and you have some kind of, [00:03:00] Loss in productivity. That's something you can compensate for and reallocate. And, the long run disruption is going to be a lot less than the short run disruption. And in particular, it's going to be close, more closely tied to the cost of the input rather than the value of all the outputs, and we'll try and make that clear, but that's going to be a central theme here, and then we'll, we're looking at two, Different extremes, one extreme short run and the other, full long run, where there's full ability of an economy to accommodate and reallocate resources and redesign its, it's supply chains.
Okay, there's a lot of research, especially in the past three or four years. There's been a growing set of papers. I'm not going to go through all the papers, but there's a growing set of research. most of them look at long run. There's a few that look [00:04:00] at short run things mostly based on inventory and trying to understand how firms can combat difficulties, they might have the kinds of contrast that we're drawing here are going to be a little different than what's out there in the literature so that'll be the main contribution.
I'm just going to stop starting to the model and then take you through some pictures that will illustrate how we think about this, but it's going to be really simple model. And the idea here is the to try and make it as tractable and simple as possible. So we have some set of countries there.
There's a set of intermediate goods. and a set of final goods and each country has labor. There's just going to be one type of labor. So we're not going to distinguish between high skilled and low skilled labor. There's just labor, which is used in production. And, we're going to abstract away from capital.
So capital is going to be absent from this model. It's just a labor based model. And we have a finite set of [00:05:00] technologies that's available for each country. So each country has got some set of recipes that can use to produce goods. It's got people that can produce them, will allow for trade between countries, and that's going to be the basic building blocks of the model.
John Cochrane: You're also disallowing inventory.
Matthew Jackson: yeah. so here, exactly. It's going to be very important that we're going to completely disallow inventories. And we're going to be looking at a, really extreme version of the supply chain disruption in the short run. And, then we can talk about, how this model could begin to be used for, inventory management and other kinds of things.
but that's something we're pushing, to simplify and we've taken away inventories completely. So here's an example. And these kinds of pictures will be central to what we're doing here. this is just going to be a typical supply chain. And in the picture here, [00:06:00] we've got labor up in the top.
So there's an endowment. Think of this is just one country. there's a labor endowment of 10 units. So there's 10 units of labor available. The consumer, the laborers are going to be the consumers. They're going to consume the final good. They're going to provide all the labor in elastically. So they just supply this labor.
And there's two Intermediate goods, good one and good two and then one final good in this economy and good one is used both in the production of the final good and in the intermediate good. So you could think of good one, for instance, as electricity and good. Two is some physical good that then is used, to produce the final good.
Okay, and labor is used in all of the we're going to assume that labor is used in each one of these technologies We don't need that but to get an existence of equilibrium You do need some labor to be used in each supply chain Okay, and so then the way technologies are going to be represented Is just in a simple [00:07:00] form, which is just a list of, what you're putting in.
Those will be in, in negatives and what you're getting out. So for instance, the first technology takes one unit of labor in and produces one unit of output. The second technology takes seven units of labor in produces one unit. sorry. And one unit of good one and produces one unit of good too. And then the final good.
Good. Takes in one unit of labor, one unit of good one, one unit of good two, and so forth. And so all of these can be just described in these flow charts, right? So we can have really simple flow diagrams that describe where everything's going in the economy.
Question: Now, this is what we call a directed acyclic graph.
Exactly, yeah. Whereas, I think if you look at a modern kind of description of production, that there are cycles that, basically like the old fashioned input output tables, everything was both a final and an intermediate. Good. Yeah.
Matthew Jackson: And [00:08:00] we'll have cycles here too. This is just a simple example for illustration, but we'll allow for cycles, and cycles complicate our lives a bit.
but, they're there. yeah, they're in the real economy. They're in the, model and, they. It's a way to, create complications. So this is just an example. So you could think of many goods, different inputs, right? exactly. let me just skip ahead a little bit.
So here's like another example that has, and here we've taken labor out of the picture, right? So now it's just the technologies of the intermediate goods. The final goods are always at the bottom in our pictures. So the four green ones are final goods. And then there's a series of different, intermediate goods and so forth.
And, as you can, as Ken was saying, you can see a cycle here, right? So for instance, tau 8, the 8th technology goes back into the 9th technology, which feeds into the 6th technology, which feeds into the 8th technology. So you've got [00:09:00] cycles in this graph, right? So you could have chips that are used in making computers, that are used in making chips and so forth.
So you get these cycles in the graphs as well. So the general structure of this is going to be that there, there's an economy. Each one of these technologies is that a vector of what it needs to take in and what it produces outwards. And then you can work out, effectively given this, what would be the most production you could get in terms of the final, The final goods.
Question: No time.
Matthew Jackson: Yeah. So here we're going to be extracting away from time and we're not going to be tracking. You know who's shipping what? When? so it's going to be a snapshot equilibrium of an economy, and we'll just be talking about what the flows are, but abstracting away from time. And that's another important simplification.
Okay, so just in terms of the technologies, then we've got these, [00:10:00] vectors. You can think of those if you had something like a Cobb Douglas production function, you can begin to approximate that by just a list of different recipes, right? So you can think of a technology is say, we've picked a particular plant, which is going to be some recipe for what we're doing in terms of labor and other inputs and producing something there might be different technologies that a firm could use.
And so there's a whole set of them that are available. They might be just using one in an equilibrium. Okay. Great. Okay, so that we've got these laborers and who are also the consumers, they're supplying their labor. Then they have some preference. We're going to make our life really easy. We're going to have homothetic preferences.
So homothetic preferences will mean all the consumers will basically be consuming different scaling's of the same bundle. So there's not gonna be heterogeneous consumption here. So we're ruling out heterogeneity [00:11:00] that'll just make our life easy so we can concentrate on the supply side of the world and, not have to worry about the, different, consumers could be impacted differently by different goods.
So here all the consumers are gonna see the same shortages of goods. producers are gonna maximize profits. It's going to be competitive market. So this is just the Arrow Debreu equilibrium. And this fits exactly into the Arrow Debreu world. you can fit this exactly into Arrow Debreu. Existence is, fairly standard.
there's, no real tricks here, in terms of getting, existence. And so then we've got these kind of. Structures that will be using as the diagrams that describe what's going on. Okay. So any questions just about the basic model. So no time, no inventories, simple recipes for production. And then, generally, there's going to be for generic set of [00:12:00] technologies, there's going to be unique equilibrium in terms of the Structure of the production.
it could be that there's different wage wages to different laborers in the world if you've got different countries, but up to that, these are basically just, the equilibrium will be generically unique.
John Taylor: So your path goes from back to town. I what is that's unusual picture.
Matthew Jackson: Yeah. So the fact that you've got these things could be said, think of tally like a computer.
And then computers are used in the production of chips, and then chips are used in the production of computers. So you've got these possibilities that you're using goods that, are used in their own production. And there's going to be possibilities. And as Ken was saying that the economy these days is full of these kinds of cycles.
So they are important to admit in the model. Think of gasoline. Some of it's a final good,
Question: some of it's
Matthew Jackson: not.
John Taylor: Yeah, [00:13:00]
Question: in production
John Taylor: process. Yeah, We only have one.
Matthew Jackson: Yeah. We had one for simplicity of the picture, the, actual drawing out the actual, supply chain there, there are companies that do it now.
So I actually advised one of 'em, as
Question: an amus aside on the old Soviet Union. They used steel to make ball bearings and melted the ball. Melted the ball bearings back down to make horses.
Matthew Jackson: I didn't hear that. Sorry. The
Question: old Soviet Union, they used steel to make ball bearings. And they didn't need all the ball bearings.
They make a Soviet production quotas. So they melted the ball bearings back down to make more steel. Make steel, yes. Yes. Exciting. So what's the allocation of technologies to countries? Is this exogenous?
Matthew Jackson: Yeah, so we'll, assume that, there's going to be So here, for instance, you could imagine that tau two is in country two, tau one is in country one.
So for most of the talk, I'm not going to be explicit about countries. So there'll just be some world supply chain. And then at the end, we'll come back [00:14:00] and we can say that these belong to different countries. And the set of technologies that are available to the different countries would, is going to be an endowment.
And, originally, what we were interested in was understanding innovation as well. And, so we want to use this model to study innovation and, but that's, for, future. Okay. So let's now start looking at short run versus long run impacts. And the first thing we're going to do is, just think of a.
Technology shock. So take one of these towers that's used in an equilibrium. So we've got one of these goods that's being produced and we're going to shock. It's technology. And generally, we're going to think of these shocks. We're going to think of reducing the output of the technology by some fraction.
So normally, when we look at this vector, it's going to have You know, a unit of production for how much inputs. Now we're just going to drop that by some amount. So I say a 10 percent drop [00:15:00] in the output for the same set of inputs. So take a technology vector. You've got the same inputs, but now you're just getting a lower amount of the output for the same amount of inputs.
John Cochrane: if they were Cobb Douglas. Are you thinking of a linear technology now?
Matthew Jackson: Yeah, so all of these were what we've done is. we're going to be looking at a particular point on a Cobb Douglas function and then saying, okay, look, you've committed in the short run to this prediction, particular production technology, which is a certain recipe of and so forth.
John Cochrane: What do you mean by short run? Is this like the ONTF? Yeah,
Matthew Jackson: it's going to look like the ONTF in the short run. Exactly. And then in the long run, you could use, you could begin to shift away. But in the short run, think of this. And then even in the long run, when we're doing small changes. you can think of it as more or less Leon tf.
Okay. So we're just gonna shock this input, output by some amount. And then in the long run there's gonna be a new equilibrium. And in [00:16:00] the short run you just are stuck with whatever the production is. you're gonna lose some production and then that's gonna propagate down this supply chain or are shock symmetric positive and negative shock symmetric in this.
So what will be true is positive and negative shocks will be symmetric in the long run, in the short run. Positive shocks don't help you at all. Because nobody else can absorb, the other producers are already set. So suddenly if I have a bunch of extra chips, nobody can, the economy can't adjust to use them tomorrow.
Six months from now, they could begin to adjust. So in the, short run, shortages are going to matter. Surpluses won't, but in the long run those will look more symmetric. So there'll be a difference between short and long run in terms of the asymmetries. Okay, so first The output is actually not
John Cochrane: linear, it's a max operator,
Matthew Jackson: right?
Exactly, it's going to look like, precisely, it's going to look like a max of all your inputs, or you have it. [00:17:00] Yes, exactly. And we'll get to that in
Question: This is very different from conventional macro views. Difference in short and long. The difference is just raising capital, I'm sure, and moving
Matthew Jackson: everything else.
Yeah, exactly. So when we mean short, super short run, and then what we have is in the end, we have what we call the medium run, which is closer to what, where you can begin to make some adjustments, but not every not everything. So you can adjust readjust inputs and so forth. And yeah, think of this as.
Extreme short run, which is everything's fixed your recipe. You're committed. So think of this. I think a good example is actually the rolling, blackouts. We had in power in California a couple of years ago. So it's suddenly, all suddenly a bunch of plants have. four hours a day where they're, they've lost power and they just can't produce.
And that's gonna be a max operator. [00:18:00] 'cause they're basically just gonna shut down. And even if they've got all the parts sitting there, they, can't do any, it's the weakest link that's gonna hurt them. And they're just gonna be reduced to, four hours a day of lower production. So we're thinking of that as the kind of shock that's hitting in the short run.
Yeah. Yeah. Yeah. And exactly. And in the long run, what can happen is you can compensate for this by building more plants or adjusting your hours or adjusting by generators and so forth. And so in the long run that you overcome this to a large extent, but in the short run, it's just going to be a pure disruption
John Cochrane: here in your super short runner.
Prices clearing markets
Matthew Jackson: competitively. No. So, what we can think of is in the super short run, we all have contracts and I suddenly just don't deliver the goods that I'm supposed to deliver to you. And then suddenly you can't deliver the goods.
John Cochrane: Several parties demanding my output. how is the output allocated across them?
Matthew Jackson: Yeah, so what we're going to think of is we're going to think of the extreme short run as I'm just turning off power, I turn off the lights to all the [00:19:00] different firms, four hours a day in the medium run, what we'll call the medium run. Is that now some of them can say, look, I'm going to pay extra. And I want you to keep my lights on all day long.
And some will turn off the lights. And so in the medium run, you'll start shifting the things towards the, ones that are more productive and more, valuable and supply chains. And then in the long run, you can completely compensate for this. We didn't so product market. Do prices get to change in the product market?
So, in the short run, there's going to be no price adjustments whatsoever. And in the medium run, then prices adjust and we'll have an equilibrium to clear the markets given the available inputs. And in the long run, you've got a completely new equilibrium.
John Cochrane: Your friends at the factory, they shut down for hours a day.
Taco truck keeps going because it doesn't use any power. So then on the product market, if I then use tacos and cars to produce something else, what happens to me?
Matthew Jackson: Yeah, so here we've assumed that essentially that all the contracts are [00:20:00] already in place. So the supply chain is already contracted. I've already paid this.
The I've already bought my tacos. I've already bought everything and then there's just some delivery and I just get shortages.
Question: There's a
Matthew Jackson: max operator production and there's another or a min operative is a min operator on the prices, and you'll see that in the algorithm.
Question: Yeah, Okay. Inventory would help a lot.
Matthew Jackson: Yes, exactly. So, inventories will help a lot. And we can talk about, how inventories will work and where the inventory should be at the and it's
Question: a completely surprise shock, like an M. I. T. Shock. It's not a, That is not at all anticipated.
Matthew Jackson: Exactly. So this is literally, suddenly there's wildfires.
We didn't anticipate that the powers out or, where we've got the Suez Canal that's suddenly not working. And we assumed it would. Yeah,
John Cochrane: we think of
Matthew Jackson: inventory is just another firm. You can think of, yeah, inventories. You can think of that as a. In the long run, building up [00:21:00] extra sources of supply. I guess it's a little different.
So if you look at the way Starbucks operates, for instance, they source from 25 different countries. And part of that's because their supply chain is constantly being disrupted. Different countries are, the countries that produce coffee tend to be very unstable in their production levels.
John Cochrane: There's a time element that's missing from the model. Think of it as another firm. If what that other firm did was buy. wait a year and then seven and then sell out time.
Matthew Jackson: yeah, Okay. So, first proposition is just to say that Houlton's theorem shows holds here and I'm not sure how many people are familiar with Houlton's theorem, but let me just talk you through what Houlton's theorem is.
it's actually a pretty powerful theorem, and it says if you want to look in the long run, so you shock some technology and you ask what's its impact on the economy. So, we, have suppose we put in. Some new emission controls on electricity. it's [00:22:00] 10 percent less productive than it used to be.
what's the impact on the economy? And what Houlton's theorem says is you look at the cost of the technology in question, and you look at how much you're spending on it, As a function of GDP, and that's the proportional multiplier. So if this, if the electricity is 10% of what we spend in the economy and we have a 10% shock to it, then it's one 100th of the, in total of, GDP that we're gonna lose.
So you just look at how much it costs in total, and that's all you need to know.
John Cochrane: Remind me, this is resting on flexible prices.
Matthew Jackson: this is a long one, and this is a complete equilibrium adjustment. So this says, take a shock to a technology, allow the economy to completely reequilibrate, and what's going to be the impact.
And so this is the, the sort of usual Thing that people look at for supply chains is [00:23:00] say, this is an estimate. You look at how expensive the inputs are, and you say that's more or less the value. And the intuition behind Holton's theorem is pretty straightforward. It says, Look, if we have a shock to something, we can just substitute by building an extra plant.
If we have a 10 percent shock, just build an extra plant or two, and that will compensate for it. That's just going to be A little more expensive, but it's as expensive by the amount that we were spending on that originally. This looks a lot like the envelope theorem. It is. It's the envelope theorem. So you use Euler's homogeneous function theorem, the envelope theorem, and this is that's what the serum.
So it's got Holden's name on it, but it's the envelope here. yes. Yeah. Yeah. Yeah. basically, it's it's the envelope theorem applied in this particular case.
Question: Yeah.
Matthew Jackson: okay. So, You know, sufficient statistic [00:24:00] is just a spending. So the nice part about Houlton's theorem is it says you don't actually have to know what the economy looks like in terms of the network.
All you just look at is how expensive is the thing that's being shocked. so it's very powerful in that way. But of course, how much, You're spending on electricity depends on where that sits in the supply chain. So the supply chain ultimately determines those costs, but that's hidden into equilibrium.
So a sufficient statistic is just that cost. Okay. just in terms of a example, let's go back to our really simple example, and then shock this technology of good one. Let's suppose that we shock it by 10%. So we had You know, two units of labor going into it originally seven units on good two, and then one unit on the final good of labor.
Now, when we shock it, what happens? it says. The total cost of this good in equilibrium was 1 5th of GDP. So if we shock up by 10 [00:25:00] percent that should be 1 50th, right? So who serum says it should be 1 50th of GDP. If you work out the new equilibrium, what happens is effectively you just reallocate labor to you.
this good is producing less. You just have to put more labor and more so you basically are building up more plants of it to compensate. So you're putting more labor there and you're cutting labor in the other parts of the economy. So you just shift, shifting things and that eventually leads to a two percent disruption.
So Houlton's theorem is very accurate in this particular model for small marginal changes. That aren't changing the technological structure, dramatically. So you're not bringing in new technologies.
John Cochrane: Is your original allocation optimal in some sense?
Matthew Jackson: Yeah, So actually these in this model, you can solve for the equilibrium just by a simple planner problem because everybody has the same homothetic preferences.
So [00:26:00] you're basically just trying to figure out, the equilibrium really easy to solve for by just, representative agent. It's a representative agent model. You just wanna maximize things and you try and get the maximum flow through this network.
Question: Does it matter? I'm sorry? Does it matter whether the total labor supply goes downstate, constant, or goes up?
Does that have anything to do with this?
Matthew Jackson: We've got a constant labor supply. Okay. Constant. So, labor is completely constant. You can also do shocks to labor in this world. I was thinking
Question: namely of robotics and artificial intelligence. In the long run, dramatically changing how we think about labor.
Matthew Jackson: Yeah. Yeah. So what those would do is actually what they would do is just change the technologies here by saying, you're getting more production with, fewer units of labor. So those would be like an increase in the, production. Can we go until. The labor
Question: disappears.
Matthew Jackson: Yeah. in this world, the, the labor with inelastic labor supply, it would just get reallocated.
But eventually, [00:27:00] that's an interesting question. 2200
Question: no people, no labor.
Matthew Jackson: Okay. Yeah. We're all just using AI.
Question: I was just reminding myself of a cam far he's beyond Holton paper. And, they emphasize the non linearities. But that's because the shocks are large. Is that why it's different from this case?
Matthew Jackson: So, here, the shocks, it's two things. One is in, in, their world, they allow for non competitive markets and large shocks, and either of those can move you away from Houlton's serum. Yes,
John Cochrane: you do. Becker also had this thing where really, big shocks. It all comes down to one single.
Matthew Jackson: so the reference on that
John Cochrane: one, you do Becker Northwestern. Yeah,
Matthew Jackson: you don't know this
John Cochrane: paper. Yeah. Working for a sort of a quadratic approximation around, and he said, no, look at the really big
Question: limits
John Cochrane: and then. As you can imagine, it all comes [00:28:00] down to exactly one. One.
Question: Yeah. The
John Cochrane: only important one.
Question: Yeah.
Then we found that, that energy was really big, didn't it? Yeah. Cool paper.
Matthew Jackson: Okay. here's a quote from Larry Summers that says that there'd be a set of economists who'd sit around explaining the electricity was only 4 percent of the economy. So if you lost 80 percent of electricity, you couldn't have lost more than 3 percent of the economy.
But we understand that if there wasn't any electricity, there wasn't really going to be much economy. So obviously the quote that he's making is, is, a contrast between the long run and the short run. So in the long run, Colton's theorem says, you adjust, but in the short run, if you do lose 80 percent electricity, you're going to lose a lot of the economy.
And so what we want to do next is try and go to this contrast. And, so we're going to this opposite benchmark where you can't adjust the technologies. you can't [00:29:00] source additional units. so you're just prices are not adjusting and you're just disrupting the stuff that's coming in.
Question: Can't you have In the business world, people whose job it is to anticipate shocks and plan for them as best they can so we don't get killed off by a disease, for example, and in so doing, there will be some slack in the system from this planning to prepare for it. That's not in the model. And
Matthew Jackson: I think it's, actually, it's interesting because, I'll talk about this.
I'm happy to talk about it offline too, but there's a company that I advise called Altana AI, and they've been building the world supply chain map. And the reason that the map is so useful now is because for firms, it's possible that I can look at my direct suppliers. And see where there might be a snag, but these supply chains go, eight, 10 deep, [00:30:00] and in some cases, hundreds deep, and I have to be anticipating potential shortages, which come way down the line.
And so actually assessing that risk and assessing where it is that I have to worry is non trivial problem. So just building up inventories of all my, is a pretty crude way to operate.
Question: Get better and better with, GPT four and then five and then six and then seven.
Matthew Jackson: Yeah, no, exactly.
we can be doing better and better with additional data, additional mapping, additional, potential risk assessments of each node, and then you can begin to try and plan for this. And the crude way to do it is just build up inventories of all your goods and sit on them. But that's an expensive way to deal with it.
Question: You have to trade off some inventory versus some risk. Yeah, Now, this is not really an Arrow Debreu economy because these shocks are not states.
Matthew Jackson: Yes, yeah. So it's not, so, here we've done like a static spot, Arrow Debreu economy, but people are not, are, not [00:31:00] buying insurance against risk, which is causing knockouts of, so there's inefficiencies in this, world due to the fact that we've got incomplete markets.
Yeah, there's missing markets here.
John Cochrane: Your, as I understand it, you're super short on what these rigidities is actually not telling us the maximal possible disruption associated with these shocks, because you're just allocating the reduced output. Uniformly across the downstream sectors, but you could imagine asking a different question.
What's the worst possible outcome from the shock? Yeah. Yeah. Yeah. And I'm pushing this and Trying to come back and forth often in response to these kind of crises. There's also Various factors that inhibit the operation of the price system and the allocation. I realize you've taken a particular stand You're not relying on the price system to do the allocation.
You're super short run But you've taken a particular stand on how the allocation happens and it could [00:32:00] be worse. Yeah, that's a good point
Matthew Jackson: Yeah, yeah. so so I guess With the medium run is the best case scenario of the disruption. And then you're saying it could be the worst, which is, I actually, if I'm so, one example that we think of, for instance, suppose there's a chip shortage and ships are used in making toys and cars, are a lot more valuable, right?
So if I'm a, if I'm the ship manufacturer, the price should adjust. So that the car manufacturers are getting the chips and the toy manufacturers go out of business for a short run or just, close their plants up. The worst case scenario is I ship all my chips to the toy manufacturers and none to the car manufacturers.
Yeah, so that's a possibility where, you end up with a much worse outcome. yeah. So here we're just looking at, like 10 percent less shipping to each of the, the downstream firms. And so then when you begin to look at the short run disruption, if this good [00:33:00] here is knocked down by 10 percent instead of just a 2 percent change in the economy, there's no readjustment of labor and so forth.
So this just propagates down, right? So they ship less in this case, equally less to each of their downstream buyers. Then that means that in the final good, the final good goes down by 10%. So here the 10% shock is propagates all the way through downstream, and then it's it, and this is because of this operator.
The way that we're with these preference, production functions, we get a full propagation of that shock downstream. So questions about how that works. So the
John Cochrane: more you could get more if
Matthew Jackson: you know how you allocate
John Cochrane: more
Matthew Jackson: depending on how you reallocate stuff. And so here, this is in this particular example, you couldn't get more, but in some examples, you could
John Cochrane: outcome.
Matthew Jackson: [00:34:00] Yeah. Yeah, it depends on the example. But this Yeah. And so here you've got the long run versus the short run. You've got basically a five times higher disruption in this particular example in the short run than the long run. And it's because of this reallocation that It's not there. in general, if you want to propagate, if you want to figure out how this works in cases where, as Ken was pointing out, there's cycles and so forth.
What you begin to do is exactly the obvious thing. So we have the shop shock propagation algorithm. So suppose we hit, one of these goods with the shortage, so it's got 10 percent of its production down. It's gonna, ship 10 percent less to each of its Downstream buyers. So then those are each hit with 10 percent disruption of some input.
Now what that does to the to these downstream suppliers depends on where they were sourcing. So for instance, the [00:35:00] one on the left here actually was sourcing from two different producers of the same good. So it's not hit as badly as the one on the right, which was just, sourcing from the one, supplier.
So then. You calculate how much loss it has given the constraints it has on its inputs. Those propagate down in terms of what shipped further downstream. And then you just keep following these disruptions. So now we've got new amounts being shipped. That is going to affect these firms. Now we've got the cycle, right?
So, tau six goes back up to tau seven. So it's only shipping 4. 76 units up to tau seven where it used to be shipping five. And so then you can begin to map this out. And in this case, you're going to get a site, that cycle is going to keep propagating where tau seven then goes back to tau four. And then, so this thing would go infinitely.
John. Did you have a question? [00:36:00]
John Cochrane: so it's usually the network structure is fixed, right? And then prices allocated across the network structure, but you don't think, Hey, if I bought better,
Matthew Jackson: right? So the long run is going to be allows you to do that reallocation and a short run. We're just we're treating this thing is fixed and just propagating it downwards.
John Cochrane: Could Mr. Brought second row, right? Could he call up the top row left and say, Hey, let's make an arrow.
Matthew Jackson: Yeah, no, we can't. So that's the short run. There's, you're basically, in the long run you can. Yes. Yeah. So the long run allows for, reallocations. but the short run doesn't.
Question: That's not such a simple model then because you've got an integer choice.
Matthew Jackson: Yeah, so the, I, yeah, so, basically generically it's simple for local changes, and then once you hit these possibilities where I wanna reallocate, then things get, yeah. Yes. And, you can [00:37:00] follow this algorithm to convergence and then see, what the disruptions are,
Question: convergence.
Matthew Jackson: the, nice thing about this algorithm is that it just keeps cutting down. So basically, you never. Nobody ever gets a, bigger shipment than they expected. So the, disruptions only keep moving things down. So you get convergence just by the genicity. And so there's a, nice properties to the algorithm.
Everything is just being disrupted downward. And then you, necessarily converge here. So convergence is easy to get. So in, in this case, it actually, that cycle cycles infinitely, but it converges to, the numbers that are there. Yeah. And so one of the propositions is that, effectively this will converge to a 10 percent loss in this case on each of the final,
Question: if you had a negative shock somewhere in a positive shock somewhere else,
Matthew Jackson: that would be, then you wouldn't [00:38:00] necessarily.
Yeah. here we've not allowing for positive. Extra output to, and it wouldn't help in the Antiaf case, given these production things that extra can't be used. So you'd still get convergence, but if you do longer run, you could, yeah, you could have problems there. Okay, so you can think of, instead of the algorithm that we just talked about in terms of just tracing this down, you could instead do a planner's problem.
And a planner's problem would be, let's maximize the total value of production subject to a bunch of constraints. And the constraints are that anybody who's shocked can't produce more than their shock amount. And then technology constraints that you can't produce more than whatever you're constrained in terms of your technology and the minimal amount of input you have.
So if one of my inputs is down 10%, another one's down 5%, I'm down 10 percent because I, with these Leontief technologies, that's what you can [00:39:00] do. Proportional rationing. So we're not allowing the prices to readjust and then any inactive technologies stay inactive. So that's the short run. And if you do it.
this minimum disruption problem that will give you the same solution as the algorithm. So that's a proposition. Basically, these two give you the same solution.
John Cochrane: The solution to this problem also tell you the sensitivity of final good output for each of the individual shocks. They'll give you some insight.
yeah, exactly. You might want to have strategic stockpiles.
Matthew Jackson: so either one of those will give you the same kind of thing. They'll trace it through. This is easier to see, because then you can see which constraints are hitting and binding, and that'll allow you to see where the, and we actually have a definition of pricing chain centrality to
John Cochrane: the real economy, which are the critical sectors that are most vulnerable.
Yes, yeah, They will have the biggest negative impact on output for the economy as a whole if you shop.
Matthew Jackson: Yeah, [00:40:00] so here, the easiest way to think of that is here's a simple version of this. This is an upper bound, but the bound is going to hold in a bunch of cases. And the upper bound would be take any good and then look downstream of it and look at all the final goods that it is used in.
And the value of those final goods in total production compared to GDP would be the amount that you could expect a disruption of, right? So you just and then goods that are on more supply chains and hitting more final goods and more valuable final goods are going to be the ones that you're most worried about, and this holds as long as You've got, either, the technologies can't be substituted for.
So if you've got multiple producers of the same good, then, I have to worry less about any particular producer of it.
Question: this reminds me of something I read in graduate school, a book by Barrow and Grossman. [00:41:00] Which I don't think is talked in.
John Cochrane: that was my first year graduate. Yes.
Question: Okay, so, the non equilibrium. Yeah, Non equilibrium. The other theme was, okay, prices can't change. yeah. But something's happened. Yeah. Okay.
Matthew Jackson: Yeah, so think of this as a
Question: market disrupts the good market.
Matthew Jackson: Yeah,
Question: Yeah, I think he would have liked this.
Yeah,
It's
John Cochrane: brilliant. It's have to make sense. It came to an economic. Yeah. And the
Question: Soviet system, because I remember you teaching. Here's the Soviet regime and here's the
Matthew Jackson: So so so here we have a bunch. There's sufficient conditions for when that bound will be tight. And basically, if single [00:42:00] sourcing is sufficient.
Low shipping costs in the world means that you tend to specialize. And so then, for instance, all of the high end supply, ships now is in Taiwan. So if you have problems shipping out of Taiwan, you're going to have disruption. we basically have the graph cut. You can give sufficient conditions, which are based on a graph cut.
I won't go through the details of that, but the simple idea is that if you've, if you're disrupting something and it produces a cut in the graph in terms of, between it and between the inputs and the outputs, then that's, that's going to lead to major disruptions. So you get conditions that are fairly intuitive.
Okay. So, the thing to emphasize then is. you've got this extreme, both extremes, Holton, where you've got full equilibrium adjustment and ours, where we've got no inventories, full, proportional rationing and so forth. [00:43:00] And what you end up with is in one case, things are just proportional to the cost of the inputs.
And the other case is proportional to the value of all the outputs downstream. And so you end up with a very different, potential, disruption calculation. Okay. when one thing that's interesting about this is when you look at different configurations of supply chains. So if we do the long run of so imagine you've got this, we've got two parallel supply chains of final goods.
You shock one of them by 10%. it's value instead of 2%. Now it's actually 1 percent because there's two final goods in this example. So it's like the other example. But now we've got two final goods. You hit it. It's 1%. If we had a different configuration of who is buying for whom in the long run, that wouldn't have mattered.
[00:44:00] In the long run, the equilibrium adjustment would have been the same. So no matter how it happens, You would end up with the same total disruption. and so here you know you get 1 percent disruption and it's either disrupts both final goods or just disrupts one, but it still ends up with the, the same overall impact, in the long run in the short run.
That's it's that's not true. So in the short run. If you have things all, you've got more compartmentalized supply chain so that, the supply of resource one is just used in one of the supply chains, then it's only hitting that one supply chain by 10%. In the other case, it hits both of them by 10%.
So in this short run. Having something that's on multiple supply chains is much more costly. In the long run, that doesn't matter. It's just the value of that input. And so you get different, the network [00:45:00] matters a lot more in the short run in terms of determining what the overall costs are. So you can begin to use this kind of algorithm and, looking at these supply chains to see how that works.
Okay, so short run network position matters, disrupting all the final goods downstream, long run, much cheaper, relative cost of input matters, network matters, but only to the extent that it determined the original prices
Question: for the median run is not a feature of your exposition.
Matthew Jackson: Yeah.
Question: There, but
Matthew Jackson: it's not, and lurks there.
And then I'll show you the medium run. I can show you a picture of the medium run and at the end, they give you some of the intuition. The medium run's a lot harder to deal with Yeah. Than, as you can imagine. here we've got full rationing. It's easy to, and then, Full adjustment, you just do equilibrium calculations.
The medium run, you're reallocating resources subject to a bunch of constraints. And it's trickier. Okay, [00:46:00] so a little bit on globalization and disruption probabilities. let's take, let's suppose that we're in worlds where that upper bound bites. So we have conditions for that. And then let's just think of a simple thought experiment.
We're just going to, with, some probability pi, we're going to disrupt any particular technology. So we'll make our life really easy. Any good technology can be hit with some probability, say 1 percent chance that it's going to be disrupted. and then what we're going to do is we're going to take that IID across the supply chains and see how do those disruptions affect final good production, as a function of the basic structure of the network.
And the things we're going to track are How many inputs are used to produce a final good? So that's S. And then what's the cost of a random input compared to a cost of the final good? So you've got some ratio of cost of inputs to cost of final goods, and then you've got [00:47:00] the average number of final goods downstream from a random input.
Those are going to be the key variables that we'll have to track, okay? So how big are the supply chains in terms of numbers of things? That's going to make it easier to disrupt a supply chain, and then the long run thing is going to depend on Cost of random input compared to cost of final goods.
And then we've got to think about how many final goods are downstream from an input that's going to affect things as well. So then, if we think about, say, here's a simple example with a horizontal kind of supply chain, we've got four different goods that are all combined into a final good, but they're independently produced.
Okay. short run. Any one of those things gets disrupted, you disrupt the final good. So your chance here of disruption is just four times whatever the, you're only gonna get 90 percent of your production. so four times 10 percent [00:48:00] times whatever the probability of hitting those are. here you've just being scaled 0.
4. long run impact is much lower. It's proportional to the cost, and each one of those inputs in this example has equal cost. So you're just scaling by one fifth of the total economy. and then you see here. The long run impact is just point eight times this probability. Okay? You could begin to look at the contrast.
So this horizontal supply chain has high short run low long run. Suppose we did a vertical supply chain. Instead, in a vertical supply chain, what happens is Still, the disruption in the short run doesn't matter. Hit any one of those goods and you disrupt the final good, right? So all you need to do is disrupt any one of its inputs and it hits, it goes down.
In the long run, the, closer you're getting to the final good, the more expensive the good is. So if you've got a [00:49:00] vertical supply chain, costs are accumulating along that supply chain, and now if you disrupt that fourth good in the long run, that's a more expensive good to disrupt. And so here that long run disruption is much higher than it was in this horizontal case, because there's a chance that you're disrupting a good, which is really close to the final good, which is fairly expensive in terms of overall cost.
So here, long run begins to look more like the short run, depending on, that now you've got more costly inputs as you go further down the supply chain. And that's because there's total accumulation of labor that's been used in the production process is accumulating. So you've got more. valuable goods.
It's if the if you're getting the plane fuselage is being disrupted. That's a major cost disruption. and not easy to overcome. Whereas if it's just a small input, then it's a difference
Question: in a sense. You mentioned international trade, international production, [00:50:00] the mirror image of mellitus results are supply chains.
Greatly increasing the gains from trades. You have to balance these things.
Matthew Jackson: Yeah, here you've got exactly So this is the flip side of. Of the fact that we're getting more and more complex supply chains and so forth. on the one hand, it saves costs. If you're using the most efficient production technologies around the world, you could be lowering your costs, but then you can also increase the possibility and the cost of disruption.
And so there's a trade off in those exactly. And then here, you can also do independent supply chains, where now in the short run, you're disrupting less of the final goods. But in the long run, it looks just like the left hand one, because it's the cost of the inputs that matters, not the way they how they plan.
So So basically, you can get different contrasts depending on the structure of these networks. [00:51:00] and, the short run, you can figure out both of these end up more the bigger the S is, which is this, the complexity of supply chain, the more disruptive they are both in the short and the long run in the long run.
You have this extra adjustment, which is more or less what's the cost per technology compared to the final goods. And then how many final goods this is involved in. In the short run, you're just getting, work increasingly complexity.
Question: I don't know the history of that. Technology, but at some point we went from inventories to just in time.
And are there studies by economic historians or economists who've looked at how much benefit you get or how much profit or how much cost savings you get from the switch from. Traditional time. So there's some sense of what you're
Matthew Jackson: Yeah, I don't know of any good. I'd love to hear if somebody knows of one a good, empirical study of how that's changed.
But [00:52:00] I do know from talking to some of these firms that do the consulting on this, that what they advise firms now is something that's somewhere between the Lots of inventory in the just in time, which is there's certain things that you need to be careful of and multi source and carry inventories of and there's other ones that you worry much less about and they still, there's some value of the just in time is forcing production, innovation and so forth.
But once your production system is settled down, then you know, they're trying to be much more judicious now in terms of how they do it. some of them are very sophisticated and others are quite naive. It seems in terms of how they do it. Could you explain how the Q over M fits in? Yeah. So the Q over M is capturing.
So when you're disrupting a particular part of your technology, then, or part of your supply chain, what really matters is how expensive that is in the long run. And so the Q is [00:53:00] capturing sort of the cost of that good compared to the final good. And then compared to how many final goods it's responsible for producing.
And so it's, basically capturing that cost element, of the, yeah. You get measures. Yes. yeah, So I think all of these, part of the reason we built this model to be fairly simple was that then you end up with. Fairly direct formulas that you could begin, it's heroic in terms of its assumptions that there's no inventory and extreme short run and so forth, but you can quantify, S is how many goods per, per final good, Q is cost average costs compared to final good.
M is, number of final goods that tend to be downstream of a particular. So all of these are easily quantified variables and you could begin to look at these. empirically. Yeah. So I guess, [00:54:00] short run and long run. One thing that then this comes back, to these trade costs and globalization.
in the model, what we do in terms of international parts is we could put an iceberg transportation costs. So just have a simple variation of the model where if you want to ship something from technology tau to technology tau prime, you have to ship some extra number of units to get a certain number there.
So we just have a simple way of putting costs in, and those costs, some theta tau, tau prime. Says how many units of output you have to ship from technology tau to get one unit into tau prime. And so that could account for geography. It could account for, maybe it's, displaced labor. So if I'm getting sourcing labor from some other country, the hours that they're working and so forth, it involves some costs and those costs [00:55:00] are captured through these.
And then what you can begin to do is say, how does the equilibrium change as you change these costs? And as you change these costs, as you reduce the costs, where you're going to tend to see is that then the technologies that are going to be used in equilibrium are going to be the cheapest technologies overall, right?
So once the shipping costs go to one where effectively you've got no shipping costs at all, then effectively you just want to source things from the cheapest supplier and you get full specialization in the equilibrium. And so the interesting thing about here is in our model, once you get full specialization, then you're fully reliant on single technologies, right?
You just have one producer that produces all the chips, all the high end chips in the world. and so then you become very dependent on that producer. and if, they're shocked, that's a shock to everything downstream of that.
Question: Buying the video stock two years ago. Yes,
Matthew Jackson: exactly. yeah.
[00:56:00] So in the video is yes, the right place. Yeah. So yeah, here. So 90 percent of the most advanced computer chips are assembled in Taiwan. and actually the materials across the border 70 times now before final assembly. When you look at all the materials that are used in the supply chain for that. it's amazing how international that supply chain is.
So they're sourcing different. Pieces from all over the place. And then there's things that are going back and forth
Question: to TSMC or whatever the company is there and look at what they do for inventory. do we have an empirical record of what their inventory is to make sure they don't get stuck somewhere?
Matthew Jackson: Yeah, I don't know. But there's also a problem. Even if you if they have an inventory. if China blockades Taiwan at some point, but, if you, imagine that Taiwan's shipping was, disrupted that could cause major issues. [00:57:00] And so then anybody else who's sourcing those chips.
Has to make sure that they're maintaining the inventories. So this is a bottleneck, right? This is a classic example of a bottleneck. And, the kind that could even take a
Question: fiscal blockade. They just have a massive cyber attack. Yeah.
Matthew Jackson: Yeah. So anything that disrupted that supply, would have the same impact.
Yes. Yeah. And so then you can have a proposition. Basically, as you consolidate these things, you get more disruption, or sorry. So as you consolidate. supply chain. So imagine you take two chains where you combine some technologies that used to be, competing technologies. And now there's just one producer of them.
So take to change, combine them. Then what's going to happen is if these technologies have independent chances of being hit, you're going to have lower probabilities of being hit. But larger disruption conditional on being hit, right? So [00:58:00] you get supply chains that have fewer producers. So you've got fewer opportunities for disruption, but they're going to be larger determinants of the total overall production.
And so you get expected short run disruption. Conditional on disruption is higher, but lower probabilities. So you've got that trade off between lower probability and higher, overall cost. So more specialized production, fewer, larger producers, larger shocks, but fewer producers. And so less frequent, you could be facing more political risk as you are shipping these things across borders more often as well.
a summary of what we looked at short and long run differ. short run depends on all these long downstreams. Long run depends only on the cost of the inputs. and then There's a series of things that we talked about increasingly complex change and more [00:59:00] vulnerable, in certain ways, and globalization specialization could lead to less likely the bigger shocks.
And I think, getting back to the where's the Where are the missing markets here? The missing markets are you're not able to contract over all these states of the world that are potentially disrupting things. And so you end up with, externalities where somebody has lost production, translates all the way down the network and ends up having, potentially large disruptions.
Question: Or following up Alvin and other people's comments, classic examples of firms that actually stock and trade. That's exactly what they do, and oil and other things are subjects.
Matthew Jackson: I'm sorry, I didn't hear the
Question: There are, in the real world, examples of markets created by firms who specialize in managing inventories and buying and selling.
Oil is a classic example.
Matthew Jackson: Yeah, and actually, it seems that there's an emerging market [01:00:00] now on supply chain disruption insurance, where you can actually buy insurance against certain forms of supply chain disruption and firms are trying to price that out. So there are growing markets in this. Yes.
John Cochrane: Build parallel chains. One is, I object a little bit to verbs without subjects. We're
going to
John Cochrane: do this building and favoring. Yeah. But, building parallel chains is, an interesting question. Do you want to build them and use them or build them and mothball them? Because what we're doing right now is we're building a parallel chain at tremendous expense to produce chips about the same way the Jones Act produces Merchant Marine.
Now. That doesn't mean necessarily you want to operate it. You might want to have it mothballed, ready to go when TSMC shuts down. We seem to be in the direction of we're going to build it and operate it. Yeah. [01:01:00] It's two different questions, really.
Matthew Jackson: Yeah, I know. how you build inventories, substitutes, parallel chains, and exactly whether you want to operate them.
I guess it depends on how easy it is to start. how do you jumpstart a, a supply chain that wasn't there and suddenly is needed? I
John Cochrane: think it's something the machinery could easily do. How do we do, let's add a technology, which is a separate one, not subject to the shock, but costs twice as much, it's an inefficient technology.
And then, do you want to use, I guess I have zero use, but yeah, then when you shut this one down, you might, yeah, you're taking your
Matthew Jackson: Yeah, I think that it depends on the timing, which is absent from this model in terms of how quick it would be to source the inputs and get the labor and everything else.
But presumably, yeah, having that [01:02:00] mothballed. And I think that the electric power industry operates somewhat that way, where depending on usage, you suddenly turn on and off plants effectively. And, when you get peak usage, then you're suddenly using all these plants and a lot of times they're just, you have minimum, you don't mothball them completely, but you have, you rev them down as much as you can and you save a lot of costs on them.
John Cochrane: Hundreds of Liberty ships parked in, Vallejo and the Bay there. Yeah. just in case, Germany and Japan come back again.
Question: the difference with semiconductors is that they're such a dynamic industry. Mothballing means that a few years later it's obsolete. It seems.
John Cochrane: I think the stuff we're gonna be build is gonna be obsolete.
Question: Yeah. But there's dynamic learning within the fact.
John Cochrane: there's the mothballing, the parallel production lines, and then there's the stockpiling, which also runs into obsolescence questions too. One of. Things I find so unsatisfying about the Discussions of our [01:03:00] strategic risks to our supply chain is nobody's trying to quantify the alternatives, if you, accept that there's a, that there's a problem where some kind of governmental solution will then mothballing production lines, operating production, new production lines, stockpiling, critical inventories, they all seem like alternatives.
And you'd like to go with the least cost alternative.
Question: Yeah, stuff isn't very useful today. everything is digital. So if you were in the analog business, how much, whatever you did was just a waste of money.
Matthew Jackson: Yeah, I think part of the difficulty is that you're chasing a constantly evolving, technology and so forth.
So it becomes increasingly difficult. And if that's, If innovation is dependent on operating the tech, then mothballing becomes even more difficult because then you're not innovating and so
Question: government into this discussion because government is in every discussion. [01:04:00] And when you have a crisis like this, I was trying to think of, I don't know, let's say transportation.
So I'm either the chair of the House and Senate committee on transportation, or I'm. Secretary of Transportation. I've got this paper and for it to be useful to me, there has to be like in the old days, staples in front of 300 word explanation in English, telling me what I should do. So what's the 300 words I can take from your paper as a government official to make things better, not worse.
If I can't make them better. Okay, fine. But is there anything I can do in the short run to make things better, or is that just.
Matthew Jackson: Yeah, I think, in the short run, it's targeting central technology. So one possibility is here, I think, looking at the supply chains and identifying goods which are used downstream and many final goods, and then seeing situations where you've got single suppliers or possibilities of shocks to [01:05:00] ones that end up being used.
Going downstream, just that knowledge and identification is step one. And then how you want to deal with that, whether you want to build alternative suppliers, whether you want to subsidize, alternative supplies of that or build inventories or, whatever is, but then you
Question: end up the government directing companies to go do that, or you subsidize company.
I know I'm interested. I want to do something right now. Yeah. Yeah. So
John Cochrane: you didn't emphasize it, but I think your framework answers the question. How valuable is letting the price mechanism allocate in response to these shocks? One answer is, as I understand your framework, you could use it to calculate and answer the question.
If we let the price mechanism work, how much less of a disruption will we get? Yeah,
Question: Except for contracts that Yeah, unless you allow something, you know what the English do with property? You just come in and raise a price.
Matthew Jackson: [01:06:00] And actually you can quantify that to some extent because in the example of like toys and cars, then you've got an alternative where you can re, re.
Allocate things towards cars. The problem with the Taiwan example where these are high end chips aren't used in toys, right? So pretty much every use of high end chips is a fairly expensive, valuable technology. And so you're going to end up then, in that case, prices aren't going to help you much, but prices help you a lot.
And
John Cochrane: then it's often with the price mechanism. Is there any
Question: case at all for government subsidizing something critical? I suppose if it has military implications? World War II bombing studies. Yeah, Yes, yeah.
John Taylor: Why don't you talk about the policy implications for the government?
Matthew Jackson: Yeah, here, yeah, targeting, it would be exactly.
Targeting these finding the central technologies in the short run, and then, either building inventories of [01:07:00] substitutes or parallel chains or, subsidizing. So in particular, the externality here is that companies don't take into account that if they're sourcing from just a few that they're going to cause problems downstream
Question: when your supply from Malaysia's cut off.
Matthew Jackson: And so you want to make sure that you're, and Starbucks can internalize it because they can source You know, coffee from 25 different producers. I'm not raising the
Question: question of if they can do it efficiently. I'm just trying to find out if there's a logic that says they should do certain things.
Matthew Jackson: yeah. And I think, the first point of attack would be just the information. So getting better information about what the long run supply chain or short run supply chain looks like, is first order problem. Big picture question,
Question: COVID. Supply shock. I remember walking along El Camino, no traffic.
So from my observation, I would have expected a huge supply. [01:08:00] I've not looked at the GDP figures, but during that period of time, how much did we lose?
Matthew Jackson: Yeah,
Question: I think, yeah, by the resilience or what?
John Cochrane: He isn't
Question: in April and April. First of all, there, there have been attempts to deal with precisely this issue, what the government should do, and they haven't all turned out so great.
when John and I were negotiating with the Japanese over trying to stop a trade war, we proposed to them stockpiling rice, rice banks, so they wouldn't be so protectionist. there are some things that make sense that economists came up, but often it gets perverted on the CHIPS Act. DSMC has no plans whatsoever to build.
to produce their high end chips in the United States They're you might call them obsolete or soon to be obsolete previous generation chips that they're building So that's something that's probably worth paying attention to because [01:09:00] you're so focused on innovation And lastly, I don't know if darryl buffy's work on the propagation of risk in the financial system That's very much the flavor of yours these networks and nobody knew who was relying on somebody else to pay them Who's relying on someone else to pay them and he has some proposals for how the fed could You Basically, they're basically injecting a dye and seeing where it goes, and it's very much in the spirit of yours, so it might be something you could take a look at.
John Taylor: Yeah,
Matthew Jackson: good point, Mike.
John Taylor: So you, your list of policy implications speaks out for what you think should be
Matthew Jackson: done. I think, it's always difficult to do this on a large scale. So it's, these are just the, kinds of taking the MO model at face value. These are things that should be done, and I think that the information is lacking, so it's very hard for, companies to know for sure in long run.
short run you. So I think the [01:10:00] main two main messages. One is short run. You really need to know where things sit in the supply chain and where the final goods downstream are, long run. It's easier. You're just looking for expensive technologies. And,
John Taylor: where is the market?
Matthew Jackson: Yeah. So I guess the question. So the missing market. So the alternative is to build better insurance systems for this. So that allow it for richer contracting, abilities for firms to be, ensuring themselves against some of the shocks that are happening, figuring out how to better inventory and, overcome these things.
partly there's missing markets here and the alternative is to try and figure out how to get those markets
Question: in,
Matthew Jackson: draw it more forward.
John Cochrane: Markets won't have generate incentives to sufficiently stockpile. And I alluded earlier, just don't [01:11:00] let the prices work. Then obviously you're not going to get the correct inventory response, but there may be other reasons as well.
I'm just. Restrictions on the price allocation.
Matthew Jackson: Yeah, I think here's one thought experiment, which is, imagine you're a company who can really understand the supply chain risks and so forth, and you're selling I don't know some consumer good. And what you do is you go ahead and you spend all this money investing to ensure that you've got.
a robust production system and that costs you an extra 10%. So your good is gonna be 10% more expensive. So the way in which you have to regain that 10% expense is by having the consumers be willing to pay for you, which could happen if you're the only one producing when there's a shock. Or it could happen because then you're going to.
contract long term with consumers. So it suggests in industries where you could have long term [01:12:00] production, contracts, then I could contract with somebody say, Look, I'm going to be a robust producer. And if there's somebody rating me and they can be sure that I'm going to be a more light.
Likely to still be producing when they need me then I should be able to get that and internalize it but to the extent that they can't see that and they can't and I can't contract long term and Then it's harder for me to do it
Question: make everything for christmas
Matthew Jackson: Yeah
Question: Has to be made six months in advance.
Yeah. Yeah. Yeah. No, it's
Matthew Jackson: but that's expensive as well, right? Yeah,
Question: clearly we'd have non zero interest rates again
Matthew Jackson: Yeah, yes.
John Taylor: What would you come away with one or two policy recommendations based on what you
Matthew Jackson: know? the key one, I think, is just, information, better information about the supply chain, and I think governments are moving in that direction, and the data is [01:13:00] more, better available, and certainly that's happened in the financial system too, central banks are more aware of the, Financial network that's out there.
They're starting to map that this means do the same thing for production networks. and then that allows both firms to be able to better take advantage of this and government and, subsidizing whole alternative supply chains is a bit drastic. So I'm not sure we recommend that. But, it's,
John Cochrane: there is a, classic public good argument here for the government to collect statistics.
Vulnerable points in the supply chain. Yes. And make those
Matthew Jackson: widely available.
John Cochrane: Yes. And that's, we have government statistical agencies that are basically founded on that in other domains. like a strategic petroleum reserve, but you just have more of them for different things.
Question: Yeah.
John Cochrane: The point I'm making is about providing information that no private party can easily gather, but that [01:14:00] has high public value.
Yeah. If they are aware of that information and operated on it in their own private interest.
Question: Yes, there is a there are two things. Three, three quick comments on the information. One is John and I will recall that we were negotiating with the Japanese. They had much better information on input output tables for United States than we did.
So it's available. People want to invest in it can do it. And they were very sensitive on very minute detail. You don't have a free trade agreement by saying there shall be free trade. It's thousands of pages of details on specifics. secondly, when you think about, the supply chains and providing information, there's already an agency, a group within the Commerce Department that's looking at inter industry flows.
Now, I don't know if they're up to the task you're suggesting, but it's actually, there's an inter, inter, inter industry, and intra industry [01:15:00] data source in the Commerce Department in the Bureau of Economic Analysis. That might be the first place to look and might be the first place to staff up or, for you to be the, guru of overseeing a consultant or something that actually does exist.
they actually publish data. John keeps asking about the policy implications. When I look at these, short run and long run, the short run always, to me, look like they're, business and market things. In the long run, it looks like. Where the government enters into this when I look at, your list of six items here, except for maybe that central target central technology, but the other two are basically on the business side or below.
They're all, I think, more government oriented.
John Cochrane: Yes,
Matthew Jackson: supporting diverse
John Cochrane: technologies. Those could all be profitable. It's not obvious, which of these are not profitable for private. You do. [01:16:00] If you're allowed to raise prices in a crisis, which you're usually not to, if you're allowed to raise prices in a crisis, stockpiling stuff is tremendously, but usually they say anti price gouging stockpile, but, policy number one,
Question: you're an 83.
And 84, 5 cents would have cleared the market, but they couldn't be raised at all.
John Cochrane: There were toilet
Question: paper shortages two years ago.
do you want to add anything? The only other thing that I would, have on the policy side is you can turn this on its head and ask how effective sanctions would be in the short run versus the long run and where you might want to, to target.
but otherwise, I think the discussion has been excellent.
John Cochrane: The short one is really there's, building resilience for the shock and then responding to the shock. There's really two kind of different,
Matthew Jackson: Yes, exactly.
John Cochrane: Yeah. Short run
Matthew Jackson: things. [01:17:00] Resilience.
