Opposing Firm-Level Responses to the China Shock: Output Competition versus Input Supply

GAICIL Insights

The China shock is usually told as a single blow.¹ This paper finds two punches thrown in opposite directions. Opposing Firm-Level Responses to the China Shock: Output Competition versus Input Supply, by Philippe Aghion, Antonin Bergeaud, Matthieu Lequien, Marc J. Melitz and Thomas Zuber, splits the surge of Chinese exports after 2000, a central case in debates about trade, into two components. An output shock hits firms selling goods that compete with Chinese imports. An input supply shock reaches firms using inputs similar to imported Chinese goods.

One question to put to anyone measuring AI's labor-market damage at the industry level: the data may be canceling a wound against a windfall. Look firm by firm, and the same technology may show up as both.

Combining French accounting, customs and patent information at the firm level, the authors find the output shock detrimental to sales, employment and innovation, with the damage concentrated in low-productivity firms. The output shock also strongly induces firms to switch their product mix toward products where France's comparative advantage relative to China is stronger. The input supply shock has reversed effects, benefiting firms that gain access to cheaper inputs. One import is a rival to some firms and a gift to others.

The method carries a warning for the whole literature. At the industry level, output and input shocks are highly correlated, making industry-level analyses hard to interpret. Correct identification requires disentangling the two at the firm level while controlling for industry-wide trends. Contrasting findings in prior studies about the China shock's effect on innovation might be explained by differences in input-output structures between the United States and Europe. Aggregate data can average a wound and a windfall into nothing much.

The net verdict is measured. The overall negative effect on French domestic innovation is broadly in line with prior US evidence, yet quantitatively small and concentrated among low-productivity firms. Trade shocks hurt some firms and help others, depending on supply-chain position and productivity.

Here is where the paper meets the present. Artificial intelligence is a two-component shock of exactly this kind. For some firms, AI-enabled competitors threaten their markets. For others, AI supplies cheaper, more powerful inputs that raise productivity. Analyses lumping these effects together may miss the reallocation inside industries. Separating competitive and input effects at the firm level offers a template for studying AI's economic impact. The question for any firm is whether AI arrives as a rival's weapon or as its own tool.

The pattern of harm is instructive. Competitive shocks hit the weakest firms hardest, prompting reallocation toward stronger ones. For AI, low-productivity firms may struggle against AI-enabled rivals while productive firms adapt. Policies helping firms upgrade capabilities could reduce adjustment costs.

Product switching is the hopeful response. Firms facing competition shifted toward products where they held stronger comparative advantage, a form of creative reallocation through which economies adjust. AI may prompt similar shifts as firms move toward activities where human skills and domestic capabilities provide an edge. Retreat from one product line can be an advance into another.

Trade policy should take note. Import competition harms some firms and benefits others through cheaper inputs. Restricting imports to protect firms facing output competition may harm firms relying on imported inputs. Evaluating trade policy requires attention to both sides. A tariff that shields one factory may raise costs for the factory next door.

Economic structure shapes outcomes. A shock's effects depend on input-output linkages, which vary across countries, so policies effective in one context may not translate directly to another. AI policy should likewise rest on evidence about how AI affects firms within each economy's structure. Importing another country's AI playbook means importing assumptions about its supply chains.

Taken together, shocks have multiple components with opposing effects, and disentangling them requires detailed firm-level data. The China shock hurt firms facing output competition and helped those gaining cheaper inputs. As AI transforms competition and inputs simultaneously, this approach will be essential for understanding its effects and designing policies that ease adjustment while capturing productivity gains.