Share
Trade wars, industrial policy, and myriad other supply chain disruptions have fast become the ‘new normal’. Yet, given the highly opaque and complex nature of production processes, knowing how to navigate this rise of deglobalisation is a daunting task. This column introduces a tool that harnesses the power of frontier AI tools to recover input–output relationships across over 5,000 traded products. The findings highlight the usefulness for AI to help build data products that are suitable for understanding networks in global trade. By mapping production networks at a granular level, policymakers can better anticipate the impacts of trade policies and supply chain disruptions.
Trade wars, industrial policy, and myriad other supply chain disruptions have fast become the ‘new normal’. These trends look set to intensify in 2025 and beyond as the world prepares for Donald Trump’s second term in office. Just last week, China announced further restrictions on the supply of rare earth minerals, yet another fissure in global economic integration.
Policymakers and executives alike must act to mitigate these shocks and identify new opportunities at a time when trade frictions could encourage geo-economic fragmentation (Aiyar and Ilyina 2023). Yet, given the highly opaque and complex nature of production processes, knowing how to navigate the rise of deglobalisation is a daunting task.
In order to improve transparency of global production and support efforts to answer questions that loom over the world’s economic fortunes, our recent paper (Fetzer et al. 2024) introduces the AI-generated Production Network, or AIPNET.
AIPNET harnesses the power of frontier AI tools to recover input–output relationships across more than 5,000 traded products. These data (available publicly at http://aipnet.io) demystify global production and help decision makers and researchers make sense of the highly complex and interdependent nature of economic production.
Mapping the global production network with AI
Understanding the intricate web of global production and its organisation along global supply chains requires a granular view of production processes. In particular, one needs to know which products are required as inputs to produce other goods.
AIPNET offers a comprehensive mapping of over 5,000 product categories and their input–output relationships. By employing advanced generative AI models, we construct a detailed network capturing how goods are interconnected in production. This network allows us to analyse shifts in trade patterns with unprecedented granularity (Fetzer et al. 2024).
Modern production processes are incredibly inter-connected and complex
The granularity of AIPNET is best illustrated with examples. Figure 1 presents the production network that is associated with the raw mineral gallium which recently has been subject to export restrictions from China. Gallium as a rare element may be, in value terms, a small yet very critical input to many high complexity products as is illustrated.
Figure 1 Production is highly complex and interconnected, as is illustrated here in the context of the supply chain of rare element gallium
The control of supply chains of such rare elements, due to their centrality in the production of many downstream goods, can have significant implications for countries attempting to, for example, localise semiconductor manufacturing. Similarly, the if the production and trade of said resources is politically controlled, the strategic flooding or control of exports may undermine strategic attempts to build up alternative suppliers.
A recent shift towards upstream goods
Combining the AIPNET linkages with data on global trade, we also construct the Integrated Global Product Centrality (IGPC) index – a measure of each product’s ‘importance’ in the production network. This allows us to track recent shifts in the kinds of products being imported globally, a latent measure of various strategic decisions made over the past decades.
Contrasting trends in imports of goods between the US, Europe and China, we note in Panel A of Figure 2 that China has steadily increased the centrality of its imports, indicating a focus on acquiring critical upstream inputs. Conversely, the United States shows a decline. This divergence hints at strategic differences and may serve as an indication of the growing geopolitical tensions.
In Panel B of Figure 2 we note that across countries, there have been significant shifts in global trade over the past decade: countries are increasingly importing goods that have high centrality or importance in production. This suggests a global reorientation towards securing essential inputs, possibly as a hedge against supply chain disruptions.
Figure 2 Trends in import centrality
Note: The IGPC index tracks the centrality of imported goods in the production network.
This trend aligns with concerns about geoeconomic fragmentation with national security concerns fuelling greater protectionism (e.g. Aiyar and Ilyina 2023) or attempts to foster ‘onshoring’, ‘friend-shoring’, or ‘nearshoring’.
Onshoring in response to supply shocks
The increase in imports of upstream goods may indicate a trend towards ‘onshoring’, where countries bring production processes back within their borders. Supply shocks, such as sudden price increases or trade barriers, can incentivise nations to develop domestic production capabilities for critical goods.
To investigate this, we identify persistent supply shocks by detecting structural breaks in the unit prices of imported goods. We find that these shocks have become more prevalent since 2016, particularly affecting consumer goods and processed intermediates, as is illustrated in Figure 3.
Figure 3 Supply shocks across the production network
Note: The figure illustrates the distribution of supply shocks by product centrality.
We note that the dynamic seems to have set in particularly strongly from 2016 onwards, coinciding with a marked increase in supply shocks that were detected. This shift suggests that countries are investing more in importing foundational products that can feed into various production processes, a possible sign of onshoring.
This observation is consistent with findings from Schwellnus et al. (2023a), who note that industries with few supplying countries (high geographic concentration) and few supplying firms within the industry (high industry concentration) are more vulnerable to supply shocks. They emphasise that public policies can enhance resilience by promoting diversification and agility in supply chains.
We estimate the onshoring response to the post 2016 supply shocks flexible across each country. Figure 4 illustrates whether different countries see an increase in imports of upstream goods following downstream supply shocks. We note that this is the case for countries such as Qatar, China, and several in Central Europe, East Africa, and North America. This pattern underscores the role of supply shocks in accelerating the shift towards domestic production capabilities.
Figure 4 Estimated supply-shock induced onshoring response across countries
Note: The map displays the extent to which countries’ imports of upstream goods are increasing following downstream supply shocks, indicative of an onshoring response. It plots t-statistics from a country-by-country regression.
We supplement the cross-country evidence with evidence pertaining to a natural experiment: the case of the Qatar blockade.
The 2017 blockade of Qatar by neighbouring countries provides a natural experiment to study the effects of sudden trade disruptions that is not too dissimilar to the experience of the UK with Brexit vis-a-vis the EU. The blockade resulted in an immediate halt of imports for many goods, providing an unprecedented and unexpected supply shock.
In Figure 5 we present descriptive evidence for one particular product that Qatar was mostly importing from its neighbouring countries at the time: dairy. Following the start of the blockade, we observe a sharp decline in imports of dairy products. Yet, at the same time, we see a rapid and pronounced increase of upstream inputs necessary to establish and run domestic dairy production. This pattern is consistent across other affected goods, indicating a strategic move towards onshoring.
Figure 5 Qatar’s import patterns before and after the blockade
Note: The figure shows import indices for dairy products and their upstream inputs.
Our analysis employs a dyadic regression approach, examining pairs of upstream and downstream goods. We find that for products highly exposed to the blockade, Qatar significantly increased imports of upstream inputs while reducing imports of the final goods. This shift underscores how trade disruptions can accelerate domestic production capabilities.
Implications for policy and business
Our findings highlight the usefulness for AI to help build data products that are suitable for understanding networks in global trade. By mapping production networks at a granular level, policymakers can better anticipate the impacts of trade policies and supply chain disruptions.
The trend towards onshoring and increased importation of upstream goods suggests that countries are seeking greater control over critical production processes. This has implications for international trade agreements, supply chain resilience, and economic development strategies. It may also shed light on populist policy-induced deglobalisation that comes with significant economic and political dislocations, such as Brexit in the UK (Alabrese et. al. 2024) and the concern of further aggressive trade policy during President Trump’s second term in office.
Moreover, Schwellnus et al. (2023b) emphasise the importance of policies that enhance supply chain resilience, such as promoting management and worker skills, using targeted fiscal support, and increasing geographic diversification of supply chains. They suggest that while onshoring can reduce dependencies, it may come at significant costs and may not always be the most effective strategy.
In the broader context, Aiyar and Ilyina (2023) caution that geo-economic fragmentation could make it more difficult to provide global public goods, such as climate action and pandemic preparedness. They argue that the negative side effects of fragmentation could outweigh the benefits and emphasize the need for multilateral cooperation.
Source: cepr.org
