Research Session 1: Global Supply Chains: The Looming "Great Reallocation"
Laura Alfaro, Warren Alpert Professor of Business Administration at the Harvard Business School, presented research on a shift in countries supplying US firms with intermediate goods and how a world of fragmented trade might affect the economy. Discussant Julian di Giovanni, department head of the Climate Risk Studies, Research and Statistics Group at the New York Fed, and moderator Camelia Minoiu, an Atlanta Fed research economist and adviser, joined Alfaro for questions and further discussion.
Transcript
Camelia Minoiu: It is my honor and privilege to welcome you to our first research session. I'm going to introduce our speakers. Today we'll be talking about global supply chains. This session will essentially provide a deep dive into the topic. We all know this topic has been top of mind for quite a few years now, with notable changes in US trade policy starting in early 2018, and then significant disruptions during the COVID pandemic.
Many have argued that we are witnessing a retreat from globalization and as a consequence we should think very hard and try to understand what is happening to trade flows. What's happening to financial flows? How are US importers and exporters, and those globally as well, adjusting to the new reality? Importantly, taking a step back, how do trade and finance come together to determine the new geographic footprint of global trade and global economic activity?
To help us think through these issues, let me turn it over to the paper presenter. We are excited to have Laura Alfaro, Warren Alpert Professor of Business Administration at Harvard Business School, present two of her research projects, "Global Supply Chains: The Looming Great Reallocation" and "Bank Financing of Global Supply Chains." Laura, the floor is yours.
Laura Alfaro: Thanks to the organizers for bringing me to a place with a beautiful sun. When I left Boston, it was miserable, it was raining, so it's always appreciated to come a little bit closer to the tropics. I'm going to present, as Camelia said, two papers: one was a paper that we were commissioned for the Jackson Hole conference, and I thank the organizers because they suggested the paper.
Some of you may have seen it. Apologies, I don't have new jokes. I'm going to recycle old jokes because I know the president of the Costa Rican Central Bank here is here and that does put on a lot of pressure. This is coauthored work with Davin Chor. I'm also going to present a new work that we're building on that previous presentation—that is, with Camelia, Mariya, and Andrea—that is trying to understand how these patterns came to be.
I usually start this presentation by mentioning that all of my coauthors and myself. We're children of globalization: we were born in one country; we came to the US. We were part of this optimism of the '90s and 2000s, me being probably the oldest of the group, where we saw the changes associated with reductions of tariffs, policy barriers, and technological change that allowed one firm to go to another country, hire labor, source inputs, and sell to another one.
In the case of Costa Rica, Intel is the example of this process. This supply chains revolution, and the search for efficiency and profit, was about also bringing growth and development to many countries. As I said, Costa Rica did this with some of these changes. My coauthor is also from Singapore, so again, in many ways, we're part of this optimism.
Fast forward to now and we're in this world where there are these concerns of natural disasters, this rethinking about the costs of globalization, in particular supply chains. There is this concern that they're actually exposing firms and countries to disruption risk.
What is interesting is that this is also on the back of a backlash, mostly in developed countries, against globalization. This is work that I have been doing with Davin and Maggie, where we have been conducting representative samples, surveys in the US, since 2018. We have documented that not only on average are people now against trade, but also there's a trend to go that has been increasing.
What is perhaps for us a little bit more disheartening is that, even when we mentioned the benefits of trade or we mentioned the cost of the tariffs, we still find people are willing to impose tariffs, in part because they cannot disassociate "trade" with trade with China and loss of jobs. I do mention that this is mostly in developed countries. It has also been documented in Europe and in others. In developing countries, even though there's a trend on average, there's still support for globalization.
We see this backlash, and in this backlash we have seen rich countries reimpose tariffs and this new view that perhaps trading with friends will reduce the risks. In the paper with Davin, what we did is we decided to take a broad view. We went back to the '90s. If you're as old as me, the '90s is not that long ago. If you're like my students, it's decades ago, although I keep telling them we had better music. Maybe only Taylor Swift would survive this era, but that is fine.
We go back and we look at partners, products, modes, value change position, and then we zoom in on the last five or six years and look for hints of reshoring and nearshoring, and we advanced two big points. The first one is—again, this is also being said by other economists, like Pol Antràs, and Penny Goldberg, and Richard Baldwin—we are not seeing deglobalization yet. I do stress the "yet," but we're not seeing it. Even in the US, the last five or six years have been years of growth in trade.
One does need to be careful—and again, many authors have said that—not to extrapolate the growth rates that we observed in the '90s and the growth rates we observe now, as "we're deglobalizing." Of course, after the '80s and '90s, we were going to see very high growth rates. The growth rates when you go from 0 to 1 to 2 are very different from the growth rates from 100 to 101 and 102.
Significant barriers fell, so we saw those high growth rates. We won't see them now. The other thing to be careful is that China reached levels of trade-to-GDP that one could argue were excessively high for a country that size, and perhaps they were not even optimal. China has also tried to rebalance their growth after the global financial crisis, trying to do more local consumption, more investment, and thus their trade-to-GDP has gone down.
Again, one may argue that it could be optimal, but a lot of the changes are actually driven by China. Rich countries have actually maintained or increased their trade levels to GDP. We're not seeing deglobalization, but we are seeing what we call "the great reallocation of global value chains." We're seeing, from the point of view of the US, less trade with China as a market share. It used to be close to 22 percent imports. It went down to 16 percent in 2022, and now it is close to 13 percent.
Who has taken that market share? It has been friends like Vietnam, and it has been friends that are close by, like Mexico. We look at these effects, and what is interesting is that these effects happened very quickly. There is little appreciation of how quickly this reallocation has taken place. We usually tend to think that there are fixed costs, these relations take time to find, and thus they tend to stay for a while, and there was a very quick switch. In this new paper, we're trying to understand how this happened, and we explore the role of banks in allowing that very quick reallocation to happen.
We also raised two points of caution. The first is, this will be costly, as tariffs were costly, and there is enough evidence to show that the tariffs were paid by US consumers. The consumers are also probably going to pay the reallocation. Going away from China, going to Vietnam and Mexico, has higher unit costs, and it may likely not reduce the total dependence from a firm that is owned by China.
Again, the government is not trying to eliminate all trade with China; it's impossible. China is very big. It's supposed to be just from strategic products, but still it is likely that this won't happen.
As I said in the paper, we start going back, and we look at the period in the '90s. The '90s already saw this big change, where the US switched from high-income partners to lower-income partners. This is the time where the US changed from Japan to China, and from Canada to Mexico. This is the gray line to the dashed line, so we see this change.
What is also interesting about the US: we tend to talk a lot about global value chains; they are actually more regional than global. In this period, even though there were substantial shocks, the US has maintained relatively constant market shares with Europe, close to 20 percent—and yes, I'm including the UK there. Sorry, UK. Asia is close to 40 percent; NAFTA is close to 30 percent. Even in this time where Japan lost, Asia wins. China won. NAFTA lost; NAFTA wins, Canada was going to Mexico. You have maintained these regional patterns. Germany and France lost; Eastern Europe won. Very stable, these global patterns.
The other thing that we stress in the paper is there are different modes in which a firm can cater to a country. They can do it through imports directly or they can do it through FDI [direct foreign investment]. This was the way that Japan dealt, and Japanese companies dealt, with a lot of protectionism, but also higher costs in Japan.
All the stories we hear now are about a country that has better cars, that they help with the environment, there's also security concerns with semiconductors and some other things, were part of the '70s, '80s, and '90s. What did Japan do? Japan started to open plants in the US. If you look at the direct involvement of a Japanese-owned firm via imports, or FDI, the share of Japan is actually higher than China. That is the blue line. That is also true for other rich countries.
Again, we like to stress the role of FDI because it has always been a way in which countries deal with barriers, but also increasing costs in their own country. I'll come back to this point. China will try to do this. It's unlikely that they're going to be able to do it in the US but will try to engage also in FDI to go around barriers.
The other thing we do is we summarize US exports and imports through a measure that was developed by my coauthor, Davin, that is called the "upstreamness of exports-imports trade." A product is very "downstream" if it's very close to the final product. Think textiles, think toys. Those directly go to the final product. A product is very "upstream" if it has to go through certain stages before it reaches the final consumer. Think semiconductors, think chemicals, think oil.
You can see from the figure that, on average, US exports have always been higher upstream than the imports. What is also interesting is that even though imports were relatively downstream, around 2010 there was a big dive and they became even more downstream. This is associated with US energy independence. It's not only that in US shale and oil there was a production boom, the US has become more energy efficient. This has meant that the US turned from a net importer of energy to a net exporter. That means that the imports are now more downstream.
We also like to stress this because countries are trying to deal with climate change. A lot of the energy for climate change as of right now is very non-tradable. This probably will look like deglobalization because countries are now not going to buy oil, but it's not really deglobalization the way we think about it. It's just the outcome of, as of right now, a lot of these energies are very non-tradable. We also like to stress that the fight of climate may look like deglobalization, but it's not really what we had in mind as deglobalization.
We zoom in to the last five years. This is a period where the US real imports grew by 6.7 percent. In fact, if you go to 2023, they grew in real terms by 10 percent. This is a time also where imports from China grew up to 2022. In 2023 they went down because there's a price effect, but overall the US has been out there trading.
What is clear is this reallocation. You see the loss of market share by China, close to 5 percentage points. If China and Asia lose, Asia wins. Vietnam has won, what we call "high-income Asia"—Korea, Singapore, and Taiwan—has won, what is middle-income Asia—that is India, Thailand and Malaysia—has won.
Who else has won? NAFTA. In fact, that managed to reverse some of the losses of Canada and Mexico has also increased. At that time last year, we stressed a lot the role of Mexico, because Mexico already had a very high and deep integration with the US, and they're gaining more. It turns out that right now the biggest market share win, if you are 2023, has been Mexico. China has lost 8 percentage points, which is close to one-third of the share with the US.
We also plot the four, if you want, "buckets" of products, or sectors, that have dominated policy debates: autos, auto parts, semiconductors, and electronics. If you do a little bit of eyeball econometrics, you can see some of the changes in semiconductors are evident. You do see China losing market share, you see high-income Asia winning market share, but you see it also in electronics, you see it in cars, especially with Mexico, and auto parts, also with Vietnam.
In the paper we do a little bit of econometrics, which I know probably doesn't fit well with this glorious location in Florida, but it's nice to do a little bit of econometrics every morning, before you go and enjoy the sun and the waves. The reason why we like to stress econometrics is that what we do is: Let me find a four-digit product that used to be traded in China. Can I find it in Vietnam? Can I find it in Mexico? Can I find it in high-income Asia, middle-income Asia, or Europe?
It turns out these countries pretty much span the changes in the US market share. So, it's not mechanical that this has to add to one, but you see that the rest of the world is not significant. This pretty much accounts for a lot of the change. Ireland and Switzerland, a lot of it was also related to COVID.
The other thing that we do is plot... I don't know if you can see Vietnam and Mexico. Here, what I want to stress: there is this movement to try to get kids out of Tik Tok. If you want kids to get out of Tik Tok, give them trade data. Give them trade data at four digits, give them trade data at six digits. You're going to see them go into the rabbit hole, and you start looking, "85? What is...85 is electronics, and this is transport, and this is metals." They may even pick up a periodic table, to find out why there's so many metals being traded. "And what are the ceramics? And what are the glass?"
What I want you to see—and you will see that you will go into the rabbit hole of trade data—is Mexico has a lot of four digits. It is very dense. I have always known the relationship between Mexico and the US was important. Quantitatively, I was not aware how deep it was. Every four-digit product is there. It's not the same as Vietnam. In fact, that's why we have to do four digits. If we went six, I would find them in Mexico, I wouldn't find them in Vietnam. It is a very deep relation in products. It was a very deep relation in quantities, and it is increasing.
Again, we go into the rabbit hole of the four digits and we try to get a sense of what was going on. Vietnam is very downstream. It is textiles, it is wood, plastic, wood floor coverings. Everything that you go into a house and you think is wood, is not. It's plastic, and it comes from Vietnam. Already a lot of that was coming, but Vietnam is going a little bit upstream—electronics, microphones—but it is still very labor-intensive.
Mexico, on the other hand, is everything. Everything you can think of is there, from electronics, to cars, to everything. It does seem to suggest that more than labor share is associated with the fact that they're close and they already have very deep supply networks.
So, there's the movement to Mexico. What is clear from all of the regressions is that tariffs played a role. Some firms were moving, but tariffs made firms move faster. That is caution number one: this is likely to have costs, so we look at unique costs. Unique cost is a proxy of the value. It could be a proxy of quality, but in this case we argue it's more a proxy of cost. Products that used to be in China that are now being brought from Vietnam are close to 10 percent more expensive. Products that used to be in China that are now being brought from Mexico are close to 3 or 4 percent more expensive.
With this data, I cannot tell you if it's supply or demand because, as I told you, all the Trump tariffs were passed to the consumers. If I'm a firm now doing business in Vietnam, I have that I can pass it to consumers. From talking, the speed this has happened at does seem to be associated with increasing costs of production, so it is likely to be a mix of supply and demand.
This that I just told you, it's all at the trade level. We try to find and aggregate it at the firm level to find intent. We use data from Tarek Hassan, from earnings calls, and we go and look: I was in China, am I trying to open in Vietnam? Am I trying to open in Mexico? You do see a spike in the data where it is a strategy of firms trying to do China plus one at the very least. We also look at greenfield investments in China, and you do see that there is a trend. Going down the trend actually precedes the tariffs so you see it's not being driven by the tariffs. As everyone knows there were changes in the regime making it harder for firms to conduct business, and you do see that in the data.
Most of the analysis so far has been at the trade level, but not at the firm level, so we want to explore a little bit more these changes at the firm level. The amount of fixed costs that exists to be able to do trade abroad is unappreciated, to the point that the average firm or the medium firm actually only has one supplier in another country. So, it is not that firms have many suppliers in many countries. Typically it's just one because it takes time. It is complicated too find the right firm with the right specifications of the product you have. It also takes time; you need to go in a boat, find the product. Again, there is a fixed cost, and typically you only do one. This change was extremely fast.
In the paper with Camelia and other coauthors, we're trying to understand what enabled this extremely fast transformation of the US supply chains abroad. What we do is we get this data set that is from Panjiva that allows us to map firms to suppliers and products. It is shipping, so we are limited to things that happen in Asia relative to Mexico. A lot of the trade with Mexico happens with trucks, so we focus on Asia and we first document at the firm level because this data is at the firm level. And indeed, these patterns that we saw in the trade data are at the firm level. It is associated with firms changing their supplier relations. They're doing less from China, and more from Asia.
We wanted to understand a little bit more what enabled this. Again, this is just trying to do that same regression that we did at the trade level at the firm level, and we do find that they are now doing less suppliers in China, they're entering less in China, they're entering more in Asia, and the net is now higher in Asia. As I said, we want to find out why this was able to happen so fast if that's what the literature tells us, and what has been documented to date is that these relations tend to be very sticky.
Of course, we look into the role of banks. Why do we look at the role of banks? Because banks, specialized banks, that is what they do. They're supposed to provide information to clients. If you're a specialized bank, you have better knowledge, you can help your firms do that. Also, relationship banking is about understanding your customer better; it's about understanding the firm that you lend to better. This was a moment that we argue that relationship banking should be able to allow the firms to exploit the knowledge and the relation to try to enter into Asia and some other countries.
That is the hypothesis that we tested: banks enable firms to do this reallocation very quickly. Indeed, looking in general we do find that there was an increase in the use of credit lines, there was an increase in the use of credit, there was also an increase in the cost, and in the loan amount. There was clearly an increase in demand. Firms wanted to borrow to be able to cover these fixed costs: trying to find a new firm in another country, set up the shop, so on and so forth, increasing demand.
Then we try to find out if the banks that had better knowledge, helped their clients, and we do find evidence of that. It turns out that if you were in a special banking relationship, you were charged lower interest rates, so there was an advantage to this process, and we're trying to understand what has enabled this change. Indeed, banks were doing the job at the right time to the right firms.
Then, we go and also present caution number two: as I said, the way Japan dealt with a lot of trade costs—internally. Japan was developing; developing is paying higher wages. It was Japan opening shop somewhere else. This is also what China is doing. It has always been an explicit policy of China not to set FDI in Mexico. When you go and look at the FDI data in Mexico, it's overwhelmingly US. China always considered Mexico as a competitor. That is changing.
Mexico has very good data. You can find at four-digit, six-digit of trade, ownership of the foreign firm, and location. If you look at the aggregate, it is still US. If you look at two sectors, auto and auto parts, 75 percent of all new FDI is China. Again, it is going into some other sectors. It used to be a competitor; now they're using Mexico as a potential way to deal with all these tariffs. China has always been in Vietnam. Vietnam doesn't have the great data that Mexico has, but Nina Pavcnik from Dartmouth has studied the case of Vietnam, and indeed there are different sources from US, other Asian countries, but China has always had a presence in Vietnam, and it's also increasing. It's interesting that it increased during COVID.
One way they're going to get around is FDI, and the other is when we look at the US's main trade partners. In every single one of them, China is increasing market share, except Japan. If you look at G20 countries, in every single one of the G20 countries China is either number one or number two. I was in Brazil last week. Electric cars, guess where they're coming from, and every other Uber had an electric car. So, penetration.
Again, the policy is not to eliminate dependence of China. It cannot be done, but it is likely that it will be very difficult to end dependence from a firm that is owned by China because of the different modes of entry. This is just, again, to underscore: Mexico has become the main trade partner and the main changes, but it was always there. What is interesting is Mexico has never seen the level of imports of capital goods, and the share of investment, in 20 years.
I mention this because it's very hard to see FDI data at the macro level in Mexico, and growth and productivity, but when you go to the micro data you see it there. There is a transformation in Mexico, even with the deep integration they already had with the US.
Is the US near-shoring? Remember the "upstreamness" measure? When you go at the end of the period, you do see a little bit of an increase in US imports. One could argue that those things that were very downstream that used to be imported, now they're produced in the US.
It's very hard to tell a lot of these changes are happening. A lot of the investment announcements, as of right now, are announcements. They haven't happened. We look at employment in autos, auto parts, electronics, and semiconductors. In auto, there's no doubt there has been an increase but it's also hard to argue that it was caused by the tariffs. It does seem that it precedes policies after the Global Financial Crisis. Auto parts also has an upward trend after the Global Financial Crisis. They were particularly hit by COVID. There's a little bit of a recovery but it's not there yet. Electronics, it's very hard to argue there is action; a lot of the announcements are still announcements. If you look at the Mexican data there is a lot going on in Mexico. What is interesting in semiconductors is that despite there also being announcements you do see an increase in semiconductors. It is possible that a lot of the change will happen in semiconductors.
To conclude, we are seeing the great reallocation of global value change. From the point of view of the US, friends are gaining market share, countries that are close by are earning market share, and perhaps some evidence of reshoring. Again, it's very early to talk about that. But caution, this will be costly. This reallocation is happening faster than probably firms would have done it, and it will have implications in terms of unit cost. Also, if China is also reallocating, it is yet to be seen what the final effects will be.
Having said that, we do find that banks played a big role in allowing this reallocation. For that, sound banks were helpful in allowing the firms and relationship banking to undertake a lot of these changes. Having said that, we do conclude the paper by saying that for a lot of this policy, it would be nice if there's some cost evaluation, social evaluation, of what is the impact and what are the benefits once the data is available. Thanks.
Minoiu: Thank you, Laura. Our discussant for the paper is Julian di Giovanni. He's department head of Climate Risk Studies in the Research and Statistics Group at the New York Fed. Thanks for being willing to discuss these papers. The floor is yours.
Julian di Giovanni: Thank you very much, Camelia. Thank you for inviting me, and to the organizers as well, for setting up a great conference so far. I get to discuss two papers, which is very nice. It's going to be a bit of a broad discussion, but to help fix ideas first, what is a global supply chain? I thought it would be useful to put up the global supply chain of our favorite mobile device, the Apple iPhone, just to give an idea what's going on around the world. Ideas come from Silicon Valley, but then Apple sources different parts of the iPhone from all over the world before final assembly and production takes place in China, and the good is shipped to customers all over the place.
Recently, you might have seen in the news that Foxconn, the main major assembler of the iPhone in China, has actually made a huge investment in Karnataka, outside Bangalore in India, to diversify where they are producing the final good. This is not necessarily an indicator of things to come. Apple still very much seems to be very entrenched in China, but it does give you an idea that not just looking at specifically US imports, as Laura and her coauthors do, but just thinking about the whole supply chain of a large US firm. Shifts are happening, and what this might mean for the future is definitely an interesting question to think about and something that Laura's papers have really made me think about over the last couple of weeks quite a bit.
What are we doing here? Two papers that use a lot of data at different levels of granularity. So, just to give a brief summary, paper one with Davin Chor thinks that focus on US firms' changing sources of intermediate inputs over a five-year period, including COVID, should take into account the exploit-level dat, sector-level measures supply chain positions, as well as earnings calls, FDI data. There's a lot going on here, but the three I would say punch lines that Laura elucidated already are the notion that US are shifting away, at least at the product level, from Chinese suppliers to friends: Mexico, Vietnam, et cetera.
There might be a potential cost for this, namely, increase in unit value of products. In truth, they might not be diversified so much when you start looking at FDI data and seeing that maybe some of these Mexican firms are all of a sudden being bought up by Chinese firms or represent Chinese affiliates.
Paper two goes into even more granular data, and they take this Panjiva supplier, basically what US firm Panjiva sources from US customs, where every firm in the US economy sources their goods from, at least for the first stage, and think about this, and merge it with credit registry data for the US. It's a huge data set, provides a lot of granular information. Then they can start thinking about more facts on supplier chain linkages as well as how these changes are financed.
What do they find? First, like paper one, they find that the tariffs tended for some substitution away from Chinese-direct suppliers towards other friendly countries. Along the way, US firms of US importers specifically demanded more credit for these changes. There's a discussion that this might be also indicated in some type of financial constraints. The authors note, for example, that depending on the banking relationship a given US importer has might lead to easier access to credit, or a larger import share, which they take as easier access to credit or lower credit costs.
There's a lot of facts here, so I'm going to zoom out. I'm an international macroeconomist, although I work with microdata as well. I thought it'd be nice to first think about what globalization has looked like in a few figures to date over the last 200 years, then think a bit more about these micro facts of what they might mean for macro risk. In particular, what I would think about is supply chain resilience. I'll leverage some work I've done on this. Finally, I'm going to hopefully get some specific comments that are useful on the two papers along the way.
First, let's look at how world trade has evolved over time. We can think about it really in two waves of globalization over the last 200 years: pre-World War II, where we're plotting world trade over total world GDP, growth in trade. World War I and World War II, I should say, kind of stopped that for quite a while. It was the same, a similar picture, in fact, for capital flows. Then, particularly since the 1980s, trade has boomed around the world. It's exponential growth until the Global Financial Crisis, where it has slowed down. As Laura mentioned in her discussion, this is actually to be expected on some level, given the high growth rates we saw in the previous couple of decades.
What might have been driving such changes? First, both shipping and communication costs have fallen quite a lot over time. Indeed, there's a beautiful paper by Kei-Mu Yi, that he wrote actually when he was at the New York Fed, which shows that even small changes in trade costs can have huge changes on trade flows around the world once you start actually factoring in these supply chains. As the chain gets longer and longer and more trade flows through it, since it's in gross terms, it's just going to get larger in actual numbers.
Two: obviously, China. Not just China, but I would say the world trade regime started to liberalize a lot more, but you do have what I call the "China shock." When they entered the WTO, especially in the 2000s, that really expanded the availability of cheap labor for firms to try to outsource and produce new goods.
All this, arguably, has also led to very complex global production networks to develop. Here, I'm giving a couple of pretty complex-looking pictures of what we can embed in our macro models to think about how shocks might spill over across countries as well as across industries. The blue ones, and this is from a paper from the coauthors, capture just nodes that are linking countries with their two largest trading partners. At the country level, the network is actually even more complex than what this looks like.
Then, we scale it. It's scaled by size of the country, by the actual flow. That's the blue, and on the right side, on the red, it's across industry. This is using sector-level data, not firm-level data. Just to make the point, you put these two together and you have a very, very complex kind of interweaving of transactions around the world in international trade, both across manufacturing sectors, how manufacturing sectors spill into services, et cetera.
All this matters when we want to think about, in particular, shock propagation, what we saw during COVID, inflation spillovers, et cetera. Going forward with all this "friendshoring" and reshoring, et cetera, is the US in particular facing fewer or greater risks? Perhaps the globalization party, as we know it, is over? Given COVID, geopolitics, climate risk, there's a lot of reasons to think that, yes, the world is definitely going to be changing.
Then the question arises: Will friendshoring or reshoring even make the US economy more resilient, at least on the production side? I would say, it applies to all countries around the world a priori, at least to my thinking of the problem, not so obvious. Why? I just showed you a production network that looks pretty complex, and as Laura was explaining, for example, even Mexican firms might be held by Chinese companies. As long as we have these production networks going, the world isn't in isolation.
To give a more specific example—a little self-advertisement; apologies—in some joint work with Ṣebnem Kalemli-Özcan, Alvaro Silva, and Muhammed Yildirim, we construct a large model of global production linkages to think about how shocks transmitted throughout the world, during the COVID year in particular. What such models show is that there's a slew of things we have to think about when thinking about spillovers in the production networks around the world.
For example, both domestic and import demand shocks will impact inflation. We understand this from basic macro models, but then you drill down and start realizing what's going on at the sector level matters as well in terms of productivity shocks, in terms of factory supplies. You close one factory in China, workers can't go to work. This is going to spill over everywhere.
Moreover, global imbalances; we've always thought about this in terms of capital flows, or savings and investment. It also matters in terms of demand, and in particular sectors as well, where these kinds of bottlenecks can be created. In thinking about resilience, we have to go look beyond just where firms are sourcing and think about where the whole chain, the global production network, as well as the source and types of shocks that might be hitting the economy going forward, and in that lens think about this in terms of resilience.
Getting into the papers a bit more. Paper number one: thinking about the micro facts, or as I've already mentioned, thinking about the global network is paramount. Even in what I would call a partial equilibrium analysis, just focusing on what's going on in the US specifically, one question that comes out is how many Vietnams are needed to replace one China. Laura was very upfront about, we're not getting rid of China altogether, but for me at least, old calculations given the regression results.
Turning to the rising unit costs, the big question was "What are they capturing? Is it more expensive factor costs in sourcing countries?" Indeed, in China the real wages have been going up quite a lot over time already, so we might have already expected some of this, with firms searching for cheaper locations.
Larger transport costs: are these set up for switching costs, are they transitory, or are they longer run? To think about what might go on with unit costs going forward or inflation in particular: Is it a one-time change, or is it something that we have to worry about reshuffling going on all the time? Of course, the general equilibrium effect of things like tariffs, obviously, are going to spill over and impact costs for goods coming from other countries.
Paper two, which is definitely more of a work in progress: the rest of my time, I'll just spend on comments here. First, it provides some super-interesting, complementary evidence on friendshoring, using these more granular data. I'm in the world of macro networks. I've fortunately never had to work with firm-level supplier data, but it's huge. Everyone should be applauding just for them working on these data. It took a lot of time, I have no doubt.
What they find is, just like with the product level, that US firms increased the number of non-Chinese suppliers and decreased the Chinese ones over the time period during the first trade war. Again, I came back to ask: Will this necessarily dampen the potential risks to production change and overall macro risks? The answer is it depends.
A key question is how diversified firms are. Laura presented something about how the average firm has only one importer. Indeed, I've worked a lot with French data, not US data, but you see the similar patterns across countries. In particular, firms tend to not diversify very much, and even when they do these large conglomerates, even within the product level, tend to be very specialized.
It's not obvious, first, that firms really want to diversify so much, but even when they do they might still skew, like have one super-important supplier, one super-important export market. In that case, idiosyncratic shocks, whether it be a climate event, something going on in the country specifically, might still have a large effect. Even if firms might look a bit diversified, it's not necessarily the case.
Secondly, of course, it depends on the ease of shifting production within a firm, across firms, and across countries. While it does look like...it's interesting, the US firms did adapt quite quickly, given the 2017-18 events going forward; it still takes time. We definitely saw it during COVID. Of course, it was a world shock, but firms could not nimbly jump around so easily. In general, in economics what we think about elasticities of substitution matter a lot. In general, we think they're quite low in the short run, versus being higher in the long run. Short run, think business cycle; long run, 5-10 years out. All this is important to think about, taking these results in context going forward.
Some more specific questions. Laura mentioned fixed costs, and thinking about what firms have to overcome to find new suppliers. They use the language in the paper very much about search costs. It naturally led to the question: What are the costs faced by US importers in getting the new suppliers? They only have indirect evidence, but for sure there's other "doing business" costs to find new suppliers that need upfront financing, so it would be useful to see if also these firms in their sample are accessing trade credit. Maybe some of this information is available at the bank or the firm levels, for example, looking at accounts receivables, et cetera.
A second point, which is—and this paper focuses more on exports, but it's something that I think is often overlooked—that there's a natural network effect that domestic firms might take advantage of when entering into foreign markets. There's a very nice paper by Thomas Chaney in the AER, where he makes the point that the more US firms that are in Vietnam, the easier it might be for a new US firm to access it, to access Vietnam, and at a lower cost. I think this is actually something you could probably control for using your data, and it would just be interesting to see if we see such patterns in the data.
My second question, really, is: What are firms' financial constraints? This is a very reduced form, and I know it's a very hard question to get in general. The authors try to control for it using what I call non-time varying, fixed effects, et cetera, or predetermined variables. More generally, we could think that as tariffs were being put in place, et cetera, that firms might become more risky. Those who are already in China might be viewed as more risky by banks, et cetera, so there might be just some interrelated relationships that they should think about in the data.
More generally, if firms appear to be getting the necessary credit for finances so that they can actually create these new supply relationships, what are the constraints? I got stuck a bit asking myself: What is the counterfactual? Is it that the firm should have gotten even more access to more loans so they could develop even more relationships, or is it simply that they're paying a higher cost of credit?
To me, my cross is where higher demand usually kind of increases price. You would think that perhaps they would be actually getting a higher rate. It wasn't obvious to me that there's a particular financial constraint going on in their analysis.
My final question was really, how to interpret the results on the banking relationships, which I think stood in a bit for an easing of financial constraints. First, banks specializing in foreign markets may have followed other domestic firms that are already there, so these regressions bank specializations might be picking up this firm network effect that I mentioned from the Chaney paper.
To be quite clear, I'm not a banking person. I do work, though, with a lot of financial people, including Kristian Blickle, who has a very nice paper on bank specialization. Their story in this paper is essentially that specialized banks are able to cherry-pick better firms, so maybe it's not so surprising that they're also offering better terms to their loans.
More generally, firms that can borrow from many banks: Are they better than firms that can borrow only from one bank, or are they just spreading risk? When you're borrowing from many banks, in general you're just spreading your risk across. Again, banks are willing to offer you better terms on your loan. In an econometric parlance, I was worried a bit about selection in your results.
Finally, some concluding remarks. I enjoyed reading both papers. They gave me a lot to think about. Globalization as we know it might be over and we're heading towards a more fragmented trade regime, so this has implications for both real economy and of course financial markets, what it might mean for capital flows, if we have any time to talk about that today. There's a ton of work going on at multilateral institutions thinking about such questions, particularly the IMF, and the research group, are working on thinking about polarization, what this might mean overall, but that's more the macro level.
Documenting all these micro facts in these two papers: super interesting, very complementary. We have to understand the micro mechanisms to understand how the macro models are working, which is just a concluding point for both researchers like myself, I am now at the New York Fed, and policymakers as well. Thank you very much.
Minoiu: Thank you, Julian. Laura, any comments, any responses?
Alfaro: First, let me thank Julian because one paper is published and the other one is a work in progress, and I do appreciate that that does impose a task. All your suggestions, all well taken. Some of the things you mentioned we have done, and we need to update the paper, but let me just give some broad thoughts.
What we do think is very interesting about this episode is that it did happen very quickly, so it is hard to use existing measures of substitution because they're not going to map well. It did happen very quickly. I also think of financial constraints not in general, but I think financial constraints are a "T." If I can borrow to buy a house, I can self-save and after 40 years buy the house. Alternatively, a bank can look at me and say, "Yes, she has a stable job, she's tenured, she's not going to get kicked out, so I can lend it to her instead of self-saving."
Why I do think financial constraints have a "T" is "Will someone give me the money that I need now?" I think that's this case: the firms needed to react quickly, because there were the tariffs and banks were able to say, "Yes, at this point I'm going to—I know you don't have it, you could self-save, but you need the money now"—and that's what happened.
We do need to be clearer on that. The point on the network is well taken, but again there is this fact that we tend to forget, and in part because we're always thinking about Apple, but the typical firm has just one supplier. The folks I found in Mexico when I went there, they are sourcing one thing from Mexico. It is not this complex "Apple" world. In that world that you only source from one place, there's not a lot of externalities in terms of network. There are externalities that are more complex, that are in terms of logistics and knowing someone who knows someone who can tell me about that place.
We do need to control for that. The other thing is I agree it's always better to have a framework. It turns out we have the census. We do know what span the change in China, and that was the regressions of table one. It was those countries that accounted for that change in China. I can tell you, in the last six years, the countries that accounted for the change in China.
It's a little bit more complex if we want to forecast. I agree that for that we do need a model. But right now, we're trying to be more positive than normative. In the world of normative, there's a lot of work that is coming out on what is optimal. Hellman from the economic department is producing a lot of these great papers. In general, there's always this: that diversification is the best strategy, much more than near-shoring, reshoring, just have different suppliers. But it does have a cost, which was the reason why firms didn't do it.
The bigger question: Is this building resilience? This is the question that we do need a theoretical and empirical work to try to understand if indeed we're making the world more resilient. I say that because, again, I am a globalization child. I was born in one country, studied in another, my husband is from a third. These changes that all presuppose that the world with less trade is better make me nervous. Also, I think we were very quick to judge trade as the reason we had problems in COVID, and it's actually the opposite. Trade helped the US in COVID. There was no world in which the US could consume the amount it consumed if the US had been self-sufficient. We need to keep in mind that there might be advantages of world trade.
The last thing I want to say is this work that we have done is in goods. There's an obsession about goods, but the US is a service economy. The US actually has a trade surplus in services, and services have continued to grow. The four sectors that I put represent 20 percent of manufacturing, which represents 9 percent of US employment. All of the employment in the US is in services, and perhaps we also should think about that. Thanks.
Minoiu: Thank you so much. Let me just say, many thanks for all the questions that are coming here to the iPad, and please keep up-voting them. I am actually going to start with a question that's a little bit more technical, and then go to the big policy questions.
Alfaro: Which might be weird, because you may be able to answer them better.
Minoiu: I just want to put to rest some concerns about data. When we work with so much data, we want to make sure that it's representative of the activities we're looking at. A question from Ian: The US data on imports from China are $130 billion lower, almost 25 percent, than Chinese estimates of their exports to the US. This gap only opened up with the expansion of tariffs from 2020. Might we just be seeing US companies under-recording shipments?
Alfaro: I forget the name, but if you look at the US trade data there is an "everything else" clause, and that's the one that has been increasing. Many people have argued that some of these things that pay tariffs are going there. Actually, in the work we did, we control a little bit for that, but it is what it is. If it's nonclassified—I forgot the name, but it has a name, and that one has increased, and it is possible that it's going there. But, yes, if it's not classified, it is very hard for us to fix it, but point well taken.
Minoiu: The next question is for the both of you. This is stepping back from this entire body of work. I have two questions that speak to the same topic, which is the impact of supply chain disruptions on inflation. So, let me quickly run through the questions. First one, to Laura: Is the right interpretation of your research on the great reallocation that it produces relative price shifts that are not inflationary in the sense that they are one-time shifts in costs?
Alfaro: Our work is about levels. Indeed, these are reallocation costs. They may show up in the data's inflation, because they may not necessarily pass in one year. If they pass over several years, they will be counted as inflation. But inflation, as we all know, everywhere from Latin America, is always about printing, so we're picking up relative changes.
The one thing is that these tariffs keep coming, and there are threats of more tariffs, so it is also possible that firms may be anticipating and passing before the changes. But yes, these are relative changes.
Minoiu: Perfect. To add on, and then I'll also ask Julian for his thoughts: What are the implications of the reallocation in global supply chain for US consumer price inflation? Over what horizon should we think about them? And feel free to draw more broadly on your own research in tackling that question.
di Giovanni: On the reallocation, again, as Laura said, it's not obvious. If these are just one-time costs, then it's not going to feed directly into inflation. In the short run, you can imagine that there will be price hikes so it'll show up in inflationary, but long run it's hard to say. At the end of the day, these are supply shocks, so more research is coming on this. Central banks can, if it's relative price changes, it's possible that they can fight against it. Obviously, there's tradeoffs to do it.
Minoiu: Let me follow up with another question that ties us back to an earlier discussion about the very, very fast technological improvement we're seeing. AI and so on. Do you feel that there's an interaction between the great reallocation and technology transformation? In particular, what do you think is the role of new technologies in explaining some of the speed and pattern of the allocation that you see in your work?
Alfaro: I don't think the speed that we see is driven by AI. This starts in 2017; it might be too early to pick up AI. We might see it later. The work we're doing also matching the credit registry is before COVID, so it is not likely to see the effects of AI.
It is asking a broader question of what's going to happen. The US leads the way in trading services, and it has been multinationals that lead the way in trading services. They are also the ones leading the way in AI, so it is possible that the effect will be seen more there than in other parts.
That is the question: What's going to happen there? That concern is being felt by countries that are on the other side. I was at a panel discussing India. In India, they worry about that: What is AI going to do to all the trading services? I don't think the data has yet to pick it up, but it's coming.
Minoiu: Speaking to the second paper you presented, on the role of the banking sector and the resilient banking sector in helping buffer some of these shocks: Does the analysis of banks' role suggest anything about how banking structure relates to lowering the cost of adjustment, through the provision of credit? For example, the relative role of large versus small banks, or domestic banks versus foreign banks. When you think about specialized banks, is it more the domestic ones, the ones with a global footprint. How should we think about how the US banking structure speaks to that reallocation?
Alfaro: I'm going to answer this question, but you need to check whether what I'm saying is correct or not. The data set has a threshold, so it's relatively larger banks to begin with. It's not about the little Brookline Bank in the corner of Brookline. It does speak about banks that have a global footprint, because we do many robustness checks and some are related to banks that do trade activities. Those tend to have a global footprint.
The specialized banks as well, they tend to have a lot of their activities in trading, so it does seem to go there. Again, my own coauthor should check on it. You had a question about trade credits. We actually checked, and there was no reaction by Chinese firms, once you put all the fixed effects, to try to keep US firms with trade credits. There doesn't seem to be a lot of action there, but again, it's a relatively recent result and we need to explore more.
Minoiu: Excellent. I'm going to use the last 40 seconds or so to just ask the mother of all questions: What are the policy implications of the great descriptive evidence you have provided? I'm going to let you put meat on that question, because it's sufficiently broad.
Alfaro: I don't think the US has done a proper cost evaluation of any of these policies. As I have said in the past, I'm a student of Harberger. We do have in the economic profession tools to evaluate cost benefits of policy intervention. It is true that a lot of the policy is still taking place and through evaluation is exposed once the policy has been taken out. So, it's hard to know. We do have evidence that the consumers paid for all the tariffs, which is bizarre if you think about it because in general it's the country that has less possibility to substitute that takes the cost of a tariff. It's bizarre that it was the US.
One would have thought that it would be China, but it was the US. We know that consumers paid for it, and we know that there was welfare reduction. There's a paper by Amiti, Weinstein, and Steve Redding. The only way I'm going to qualify what I just said is that we economists do not have good quantitative models to think about national security. As I said in the paper I have done with Chor and Maggie Chen, there is this big concern in the US that people cannot disentangle jobs from China, and some broader concerns of national security. We economists are very good at coming up with many things and upsetting all the other social sciences, but we should do some effort to try to quantify the effects of national security to get a sense whether consumers paying more is justified by the concerns of national security.
Minoiu: With that I'll say thank you Laura, thank you Julian. We have lunch next, and I will invite you all to join us back at 7 p.m. for dinner and a keynote address with Professor Ed Glaeser. Thank you.