Industrial Policy Conditionality: A Guide for Users
Public debate treats conditionality as a single dial to be turned up or down to impose stricter discipline on private actors. That framing obscures the choices that actually matter.
By Luuk Schmitz, Max Planck Institute for the Study of Societies (MPIfG)
Almost every discussion of industrial policy conditionality treats it as a story about too little. Without strings attached, the worry goes, public money flows to politically connected incumbents, subsidies become corporate welfare, and the state ends up captive to the industries it was meant to steer. That concern is real, well-documented, and historically grounded. But it is only half the picture, and without thinking about the other end of the rope, these strings may yet undo the industrial policies they are meant to keep on track. Put differently, conditionality is not simply a safeguard against waste. It is also a design choice that affects whether firms participate, whether projects move quickly enough to matter, and whether public support builds real industrial capability. The challenge is not whether industrial policy should come with strings attached. It is which strings to attach, when to attach them, and how to ensure they discipline firms without deterring the investments policymakers are trying to mobilize.
Applying conditionalities in real-world settings can be costly. The procedures, performance benchmarks, and eligibility tests that align private behavior with public goals consume time, administrative capacity, and firm attention. They shape who participates and who doesn’t. They produce workarounds. And they mandate certainty in fundamentally uncertain processes.
Drawing from my own collaborative work on the European Union, which applies stringent conditionalities in its Important Projects of Common European Interest (IPCEIs), these trade-offs come into view. IPCEIs are one of the European Union’s main tools for allowing member states to support strategic industrial projects that would otherwise fall afoul of EU state aid rules. They are meant to enable large-scale, cross-border projects in areas such as batteries, hydrogen, microelectronics, and cloud infrastructure where private investment alone may be insufficient. But because they operate as an exception to the EU’s normal restrictions on subsidies, firms and governments must show that public support is necessary, proportionate, and directed toward an important common European objective. First deployed in 2018, IPCEIs allocate over €37 billion in public funding across ten initiatives in microelectronics, batteries, hydrogen, cloud, and health. But fourteen out of twenty-seven member states have never participated in one. Approval timelines routinely run four to five years. And policymakers who have overseen the process describe participants, not entirely in jest, as ‘damaged for life.’
IPCEIs are shaped by the EU’s unusual governance structure: member states provide much of the money, but Brussels sets the rules under which that support can be approved. That makes the instrument distinctively European. Still, the lessons travel. Every government pursuing industrial policy faces the same basic problem: how to attach conditions that prevent waste and capture without making strategically important projects too slow, too costly, or too difficult to pursue. Getting conditionality right means taking its costs as seriously as its benefits.
Conditionality is more than a dial
Public debate treats conditionality as a single dial to be turned up or down to impose stricter discipline on private actors. That framing obscures the choices that actually matter. Conditionality is a set of decisions about what to demand of firms, when to demand it, and whose interests to prioritize.
Industrial policy conditionalities can cover very different goals, each with its own failure mode. Directional conditions steer firms toward higher-value activities, novel technologies, or the commercialization of public R&D. They can help push firms beyond business-as-usual investment, but they risk locking in the wrong technological bet if policymakers demand too much certainty too early. Localization conditions require domestic production, local content, or job creation. They can build domestic capacity and political support, but they can also create sheltered industries that never become competitive. Social conditions secure worker protections, environmental standards, or affordable access to subsidized products. They can broaden the public value of industrial policy, but if designed poorly, they may make participation too burdensome. Geopolitical conditions restrict who can participate, where production can expand, and how intellectual property can travel. They can protect strategic capabilities, but they may also alienate allies or fragment supply chains.
A further problem in the public debate is that we underappreciate how these categories interact with technological uncertainty. Our comparative historical cases for good industrial policy and, by extension, good conditionality, all stem from East Asia in the 1980s-90s. But the conditionalities suited to late-developer catch-up — strict performance targets, output benchmarks, export quotas — were designed for a world in which the technology was known and the task was to climb a familiar ladder.
Frontier industrial policies targeting strategic sectors such as microelectronics, batteries, hydrogen, or advanced manufacturing face a different problem. When the technology path is uncertain, the supply chain is still forming, and the relevant capabilities are still being discovered, it is unrealistic to impose detailed ex-ante expectations of success. Market actors still need discipline, but that discipline should often come through milestones, learning requirements, repayable advances, and ex-post evaluation rather than rigid upfront tests. Here, Europe’s IPCEIs serve as a cautionary tale.
The IPCEI Experience
The way a state uses conditionalities in industrial policy involves more than technical design. Conditionalities inherit the institutional habits, legal traditions, and political assumptions of the system that produces them. Here, the European Union has layered a fifth logic on top of the four categories above.
State aid — public support that gives a selective advantage to a private firm — is restricted by default under the EU treaties. The purpose is to prevent subsidy races among member states and protect fair competition inside the single market. IPCEIs rest on an exception to that logic. They allow public support when projects address a major market failure, contribute to a common European objective, and would not happen in the same form without state aid. This is a coherent framework for limiting waste and preventing subsidy races. But it is a less natural framework for building strategic industries, where speed, scale, uncertainty, and capability-building often matter as much as proving that a narrow market failure exists. Much of the European tension comes from trying to reverse-engineer industrial policy through competition law.
That mismatch produces three costs, each visible in IPCEI projects. The result is not simply “too much conditionality.” It is a particular kind of conditionality, one optimized for legal defensibility and market-discipline tests, rather than for the practical needs of frontier industrial development.
The first cost is perverse outcomes: conditions that defeat the policy’s own purpose. To justify state aid, IPCEI projects must often demonstrate innovation “beyond the global state of the art.” The goal is understandable. Public money should support additional innovation, not subsidize routine investment. But in frontier sectors, the requirement can backfire. For a firm like TSMC, a European facility that exceeded the technological frontier of its Taiwanese operations would raise obvious strategic concerns, including the risk of weakening Taiwan’s “silicon shield.” A condition designed to push the innovation frontier can therefore exclude precisely the firms most capable of delivering strategic capacity.
The second cost is adverse selection. Much of the IPCEI approval process requires firms to provide detailed information about funding gaps, expected returns, project design, and why public support is needed. This can reduce unnecessary state aid, but it also creates delay and administrative burden. Approval timelines have often run four to five years. The result is that IPCEI participation tends to favor member states and firms with the bureaucratic infrastructure to manage the process. Conditionality becomes a filter for administrative capacity rather than strategic value.
The third cost is workarounds. When procedures become too burdensome, firms and governments adapt in ways that preserve the appearance of conditionality while weakening its substance. Germany’s Economy Ministry, for example, developed an “Early Start Mechanism” that allows firms to begin investment at their own risk before Commission approval. But this creates a paradox: if a firm can credibly start without aid, it becomes harder to prove that aid was necessary in the first place. Firms have also begun structuring projects to stay just below the €50 million threshold that would trigger future subsidy clawback rules. In each case, conditionality does not disappear; it is rerouted.
These outcomes are not simply implementation failures, but reflect a deeper design problem: conditionality that is calibrated to prevent distortion can become poorly calibrated for building industrial capacity under uncertainty.
Conditionality for a Second-Best World
The way conditionalities come about, what they demand from private actors, and how they are implemented reflect a fundamentally political process. There is no universal formula for conditionality. A condition that works well for a mature industry may fail in a frontier sector. A rule that is politically durable may be administratively difficult to enforce. A requirement that protects public value may also slow investment. Conditionality design is therefore a second-best problem: policymakers must choose among trade-offs rather than optimize in the abstract. Optimal design becomes a question of matching the rules to the political environment in which industrial policy actually operates.
Three design principles follow.
Match conditionality to uncertainty. The more uncertain the technology, market, or supply chain, the less conditionality should rely on rigid upfront specifications of success. Directional conditions in frontier sectors — steering firms toward new technologies, novel value chains, or capabilities still being discovered — should be milestone-based, iterative, and open to ex-post adjustment. By contrast, low-uncertainty conditions, where the relevant parameters are stable and measurable, can carry clearer upfront targets. The current EU regime often gets this calibration backwards: heavy ex-ante review of innovation criteria sits alongside relatively weak ex-post discipline.
The US Inflation Reduction Act tilted the other way, trading ex-ante directional rigour for performance-based and security-driven content rules. Even after the 2025 One Big Beautiful Bill Act repealed much of the IRA’s clean-energy machinery, the design philosophy that survived: escalating threshold percentages, material-assistance cost ratios, and milestone-based phase-ins more closely track what frontier industrial policy needs.
The IRA had its own weaknesses, but its conditionality was often easier for firms to understand and act on: meet a content threshold, locate in a qualifying community, satisfy labor standards, or produce an eligible good. That kind of rule may be cruder than the EU’s project-by-project assessment, but it can be more usable in fast-moving industrial sectors.
Sequence by political robustness. Some conditionalities are easier to sustain politically than others. In the current environment, geopolitical conditions — rules about who can participate, where production can expand, and how technology can travel — often enjoy broader support than more contested social or localization requirements. That makes them a durable starting point. But governments should not stop there. Directional, social, and localization conditions may be equally important for building public value, but they require greater enforcement capacity and more careful design. Sequencing conditionalities can therefore be a way to build state capacity over time rather than loading every objective into the first version of a program.
Tolerate bounded failure. The goal should not be to minimize the chances that mistakes will occur, but to minimize the costs of mistakes when they do. That insight matters especially for frontier industrial policy. If the state is supporting technologies, value chains, or capabilities that do not yet fully exist, some failure is inevitable. Conditionality should make failure informative and bounded, not impossible. A regime that produces zero failures is either not trying hard enough or is paying its costs invisibly, through firms that decline to participate at all.
The IPCEI experience is a particularly clear case, but the underlying problem is general. Every government now pursuing industrial policy faces the same calibration challenge. Too little conditionality risks waste, capture, and corporate welfare. Too much, or the wrong kind, can deter participation, slow investment, and select for bureaucratic sophistication rather than industrial capability. Conditionality is not a single dial. It is a series of choices about uncertainty, capacity, time, and political opportunity.
