This post was originally published on this site.
Supply chain disruptions have been a defining feature of the global economy since 2020. Geopolitical risks, diplomatic tensions, regional instability, and environmental risks are exposing critical vulnerabilities in global sourcing models. These crises underscore that import dependencies, especially on a single partner, can be a major source of fragility and lead to large macroeconomic effects through supply chains (Alessandria et al. 2023). In response, major economies have implemented strategic resilience frameworks, such as the US Inflation Reduction Act, the EU’s Net-Zero Industry Act, and the upcoming Industrial Accelerator Act. The relevance and design of such economic policies require identifying trade vulnerabilities and a deeper understanding of how firms hedge against supply-chain risk.
In response to this supply risk, firms primarily adopt two main strategies: building strategic buffer stocks and diversifying sourcing origins. There is mounting empirical evidence that these strategies are often adopted ex-post. Firms exposed to supply chain shocks adapt by revising their inventory strategies (Zhang and Doan 2023), diversifying their supply sources (Castro-Vincenzi et al. 2024), or even relocating production (Castro-Vincenzi 2024, Balboni et al. 2025).
In turn, documenting the recourse to those strategies ex-ante informs about the exposure and robustness of sourcing. In a recent paper (Lafrogne-Joussier 2025), I use comprehensive firm-level data on stockpiling and sourcing diversification to document the extent to which French manufacturing firms utilise both strategies. By focusing on importers — where direct exposure to international shocks is most acute — this research demonstrates that accounting for inventories significantly alters our assessment of trade vulnerabilities.
Inventories and diversification practices differ widely across firms
In the manufacturing industry, the median firm holds enough stock to sustain production for over two months (63 days) in the event of a total supply cutoff. This median masks vast disparities: the top 10% of firms hold at least five months of stock, compared to a maximum of six days for the bottom 10% — a significant 7.5% of importing firms report zero inventories of inputs.
Practices vary by sector, with a median stock of nearly 114 days in pharmaceuticals versus 37 days in the agri-food sector (Figure 1a). These disparities are not merely a function of input perishability; even within narrowly defined industries, firm-level idiosyncrasies remain considerable.
Consistent with previous literature, I show that the level of inventories at the firm-level is increasing in the share of imported inputs (Alessandria et al. 2010) and the geographical distance inputs travel to reach France (Carreras-Valle 2024). While the former is typically attributed to fixed costs per shipment, the latter is intrinsically linked to delivery-time volatility. This reinforces the conceptualisation of inventories as a critical buffer against exogenous supply risk.
Figure 1 Inventories and diversification across and within industries
A) Inventory ratio
B) Diversification
Note: This figure displays the median, p25, and p75 within two-digit industries of the number of days of inventories of inputs (Panel A) and of the average number of sourcing countries per product, called diversification (Panel B). The sample contains one data point per importing manufacturing firm in France, per year, between 2012 and 2023.
Using individual customs data, I measure the diversification of sourcing of French importers as the (weighted) average number of origin countries per imported input.
Much like inventory management, diversification strategies exhibit significant variance across firms: a quarter of manufacturing importers source each product from a single foreign origin on average, whereas the most diversified decile sources from more than four countries. The automotive industry appears as the most diversified sector, with the median firm importing inputs from an average of over 2.3 countries (Figure 1b). In contrast, in the metallurgy and apparel sectors, the median firm relies on a single foreign source for its entire imported input basket.
Two factors influencing this variability across firms are worth discussing. First, some of these differences are due to economies of scale in importing: diversification increases with the number of imported inputs. Second, diversification is constrained by the underlying supply architecture of specific inputs. Firms are significantly more likely to diversify when the product in question is supplied by a larger number of countries globally.
Building up inventories and diversifying supply sources are substitute strategies
A central finding of this paper is that diversification and stockpiling are substitutes. While both aim to reduce production volatility, they operate through distinct mechanisms: diversification mitigates upstream risk, while inventories buffer against realised shocks.
Empirically, geographic diversification significantly stabilises import flows. Firms sourcing from a single country exhibit a dispersion of monthly imports almost three times higher than that of those sourcing from over eight countries. Conversely, inventories preserve operational continuity during acute supply stress. For instance, the shortage of Chinese inputs due to the early lockdown in 2020 did not significantly affect French firms sourcing from China that held high levels of inventories (Lafrogne-Joussier et al. 2022).
A robust negative correlation emerges: as inventory intensity increases, geographic diversification declines. Firms with one month of stock average 2.5 origin countries, whereas the top 5% (with 240 days of autonomy) average fewer than two. This trade-off persists when controlling for sectoral fixed effects and within-firm longitudinal variation.
The substitutability between these strategies is partly driven by diverging cost structures.
Diversification entails significant fixed costs — such as supplier identification and contract negotiation — which are proportionally lower for larger firms. Furthermore, greater market power allows large firms to secure higher supplier reliability, reducing the need for precautionary holdings. Empirically, diversification scales with firm size: the smallest firms average approximately one origin country per product, whereas the top 5% source from more than five (Figure 2).
Conversely, stockpiling intensity decreases with firm size. While the smallest 5% of firms maintain an average of 96 days of inventory, the largest 5% hold only 53 days. This is consistent with the theory that inventory costs — including storage and management — are convex (Ramey and West, 1999), making large-scale stockpiling disproportionately expensive for high-volume firms.
Figure 2 Inventories and diversification along firm size
Note: This figure displays a binscatter of the number of days of inventories of inputs on firms’ (log of) annual sales, in black, and a binscatter of diversification on firms’ (log of) annual sales, in grey. The pattern also holds after removing industry-year and firm fixed effects, and for various measures of firm size (number of employees, value-added, labour productivity). The sample contains one data point per importing manufacturing firm in France, per year, between 2012 and 2023.
Refining trade vulnerabilities assessments: The role of stocks
Securing international supply chains is a major economic policy goal, requiring the identification of ‘vulnerable’ products whose supply is particularly fragile. Existing analyses often retain the concentration of sourcing origins as a measure of exposure to international supply shocks. As less diversified firms may compensate with sufficient inventory to manage short-term disruptions, taking inventories into account may reduce the number of vulnerable products.
I consider an input as vulnerable if it exhibits three combined characteristics: low diversification, low diversifiability, and low stocking.
- Low diversification: Imported from a small number of countries. Out of 3,175 inputs, 62 are imported from just one country, 624 from two countries or less.
- Low diversifiability: Global production is highly concentrated (e.g. HHI > 0.25), limiting alternatives (the case for 496 inputs).
- Low stocking: Imported by firms with low average stock levels. Out of 3,175 inputs, 1,839 are imported by firms with less than one month of inventories of inputs, meaning that any disruption exceeding one month would immediately propagate through the domestic economy.
By combining the first two criteria (low diversification and concentrated global supply), 174 inputs appear at risk. However, accounting for inventory buffers reduces this figure to 79 highly vulnerable products. Changing the thresholds on the first two criteria does not change the effect of taking inventories into account: around half of them are sufficiently stockpiled (more than one month).
These highly vulnerable products account for only 0.2% of the total value of input imports, but a supply disruption of some of them may have larger effects. Some of the 79 vulnerable products are upstream inputs, like several minerals such as cobalt (essential for battery manufacturing), whose disruption may ripple downstream along the supply chain. A shortage of some others may create large non-economic effects, as the largest group of inputs among the 79 is chemical products, including organic chemical products vital for the pharmaceutical and cosmetic industries.
Figure 3 Number of trade vulnerabilities
Note: This figure displays the number of imported inputs according to the vulnerability criteria described in the text. Statistics are computed on imports over the five years between 2015 and 2019, at the HS6-level, and restricted to intermediate products, according to the BEC classification.
Towards resilience policies
Integrating inventory data significantly refines trade vulnerability assessments. However, while stockpiling mitigates transitory shocks, it is an ineffective hedge against protracted disruptions, such as permanent shifts in trade policy or geopolitical alignments.
More generally, it remains unclear whether current private incentives for supply resilience levels align with public resilience goals. Since there is no established benchmark for optimal stocks or optimal diversification, further research is required to quantify the wedge between private incentives and public welfare. Last, the high heterogeneity in supply-chain risk management practices uncovered in this column suggests that effective resilience policies will not be uniform across firms.
References
Alessandria, G, J P Kaboski and V Midrigan (2010), “Inventories, lumpy trade, and large devaluations”, American Economic Review 100(5): 2304-2339.
Alessandria, G, S Y Khan, A Khederlarian, C Mix and K Ruhl (2023), “The aggregate effects of local and global supply chain disruptions, 2020-2022”, VoxEU.org, 12 March.
Balboni, C, J Boehm and M Waseem (2025), “Firm adaptation and production networks: Structural evidence from extreme weather events in Pakistan”, Mimeo.
Carreras-Valle, M-J (2024), “Increasing inventories: The role of delivery times”, Technical report, Penn State University.
Castro-Vincenzi, J (2024), “Climate hazards and resilience in the global car industry”, Mimeo.
Castro-Vincenzi, J, G Khanna, N Morales and N Pandalai-Nayar (2024), “Weathering the storm: Supply chains and climate risk”, NBER Working Paper 32218.
Lafrogne-Joussier, R (2025), “Inventories, Diversification, and Trade Vulnerabilities”, Insee Working Paper 2025-22.
Lafrogne-Joussier, R, J Martin and I Mejean (2022), “Supply chain disruptions and mitigation strategies”, VoxEU.org, 5 February.
Ramey, V A and K D West (1999), “Inventories”, Handbook of Macroeconomics 1: 863-923.
Zhang, H and T T H Doan (2023), “From just-in-time to just-in-case: Global sourcing and firm inventory after the pandemic”, VoxEU.org, 1 September.