Industrial Policy and Competition

GAICIL Insights

Industrial policy has a famous failure mode: pick a champion, shower it with money, watch it get comfortable.¹ Industrial Policy and Competition, by Philippe Aghion, Jing Cai, Mathias Dewatripont, Luosha Du, Ann Harrison and Patrick Legros, shows how to avoid it. Industrial policy is back at the center of economic debate, and its effectiveness depends on design. Using a comprehensive dataset of all medium and large enterprises in China between 1998 and 2007, the authors show that industrial policies allocated to competitive sectors, or that foster competition within a sector, increase productivity growth.

For national AI programs, the sting is this: publish the allocation rule before the budget. If the rule funnels money to a handful of incumbents, the evidence says the program may underperform no matter how large the headline figure.

The policies include subsidies, tax holidays, loans and tariffs. Competition is measured with the Lerner Index. Competition-fostering measures include policies more dispersed across firms in a sector and measures encouraging younger and more productive enterprises. The central finding: sectoral state aid fosters productivity growth more when it targets more competitive sectors and when it is not concentrated on one or a few firms. Spread the money, and it works harder.

The magnitudes are notable. At the intensive margin of within-firm behavior, spreading instruments across more firms is associated with positive firm-level productivity increases. Allocating more benefits to more competitive, typically younger, firms is associated with even greater benefits, doubling or tripling the effects. Design can multiply results several times over. How the money is handed out matters as much as how much there is.

The result dissolves an old standoff. Industrial policy can be effective, and competition is essential. Policies supporting sectors while preserving or increasing competition harness both. Policies concentrating aid on a few incumbents may shield them from competition and reduce productivity growth. The debate between state and market turns into a debate about allocation rules.

Where does this leave AI? AI is the obvious application. Many governments are launching AI support programs through subsidies, tax incentives, research funding and procurement. Such programs will be most effective when targeting competitive sectors and spreading support across many firms, especially younger ones. Concentrating support on national champions may raise productivity less. A given budget spread across many AI startups may outperform the same sum handed to one flagship.

Young firms deserve the emphasis. AI innovation is often driven by startups and young firms. Industrial policies favoring established incumbents may miss the most dynamic sources of innovation. Designing AI programs to reach young, productive firms can amplify their effects. Grant processes built for large applicants quietly exclude the firms with the most to gain.

Market structure enters too. AI markets with strong competition may benefit more from public support than markets dominated by a few firms. Policymakers can pair industrial policy with competition policy, ensuring support does not entrench incumbents and that markets stay contestable. Subsidy and antitrust belong in the same meeting.

The data are comprehensive. Examining all medium and large enterprises over a decade lets the authors compare policy designs across many sectors, evidence valuable for designing industrial policy elsewhere, including AI.

Evaluation should be built in. Support programs can be structured so their productivity effects are measured and compared across allocation rules, letting governments shift resources toward designs that work best. An AI fund without an evaluation plan is spending blind.

The message for policymakers is constructive. Industrial policy can contribute to productivity growth if designed to work with competition. The tools, subsidies, tax holidays and loans, succeed or fail through their allocation. Spreading support, targeting competitive sectors and favoring young and productive firms produce the largest gains.

Pulled together, industrial policy and competition are complements when policy is well designed. In China, state aid allocated to competitive sectors or in competition-preserving ways increased productivity growth substantially. As governments invest in AI, this evidence offers a clear guide: support innovation broadly, favor young and productive firms and preserve the competitive pressure that drives improvement.