Is industrial policy back, and is it justified?
A word economists used as an insult is now the law of the land. The harder question is whether the comeback corrects an old mistake or repeats one.
The revival: from slur to strategy
“The postulate that deep trade liberalization would help America export goods, not jobs and capacity, was a promise made but not kept. … To build. To build capacity, to build resilience, to build inclusiveness, at home and with partners abroad.”
— Jake Sullivan, Brookings “new Washington consensus” address, April 2023
Inside one decade, “industrial policy” went from an epithet to the explicit governing strategy of the world’s largest economies. The 2022 CHIPS and Science Act put roughly \$52 billion behind domestic semiconductor manufacturing; the Inflation Reduction Act committed around \$370 billion in clean-energy subsidies and credits; the EU answered with its Net-Zero Industry Act. The question is no longer whether the state should pick sectors. It already has.
Industrial policy is selective, sector-targeted state intervention to shape what an economy produces, subsidies, directed credit, tariffs, public investment, procurement, all aimed at specific industries rather than the economy as a whole. It differs from horizontal policy (a corporate-tax rate, an interest rate, a patent system), which treats every sector alike. For two generations the mainstream verdict on the vertical kind was close to unanimous: it was the thing governments did when they wanted to disguise favors as strategy.
The revival has answered “should the state ever target sectors?” with new programs and formerly skeptical economists saying a qualified yes. The live question is narrower: is this revival aimed at genuine market failures, and can the state actually execute on them? Which failures justify intervention in principle is what the next stage sets out.
Mariana Mazzucato’s The Entrepreneurial State (2013) reframed the state as the risk-bearing first investor whose early bets — the technologies behind the smartphone, the internet, GPS — private capital was too cautious to fund. Her lineage, and Dani Rodrik’s calibrated-industrial-policy position, belong to the post-2008 pluralist turn in economic thought, traced in History of Economic Thought Ch.17 (Modern pluralism).
“Industrial policy is back, and it’s the right answer to an economy that financialized its way out of building things.”
— the popular pro-revival case, after Mazzucato
Did industrial policy actually come back?
The empirical claim is easy to overstate in both directions. One camp says nothing changed but the branding; the other says the laissez-faire era is simply over. The reading separates “came back” (a fact about policy) from “is justified” (the rest of this walkthrough).
A correction, or a relapse?
“The state is not just fixing markets, it is actively creating and shaping them.”
— Mariana Mazzucato, The Entrepreneurial State, 2013
Mazzucato’s case is that the revival corrects a category error. The story that markets innovate and the state merely cleans up afterward gets the history backwards: the foundational technologies of the last half-century were funded into existence by mission-driven public agencies long before any private firm would touch them. On this reading the laissez-faire decades were a period of the public sector quietly carrying the risk while the private sector took the credit, and the revival makes that division of labor explicit and deliberate.
“Industrial policy is one of those things that economists and elites recoil from out of bitter experience. The label changes; the failure mode — politically directed capital, protected incumbents, no exit — does not.”
— the free-market skeptic’s rejoinder (Cato / Posen tradition)
The skeptic answers that the revival is protectionism wearing a green coat of paint. The names are new — resilience, friend-shoring, the green transition — but the machinery is the one that failed before: the legislature decides which firms get capital, the favored firms acquire lobbying power, and the subsidy outlives whatever rationale launched it. Mazzucato’s celebrated examples are real, but they are also survivors selected after the fact; for every DARPA there is a graveyard of public bets that quietly failed, and a planning apparatus that learns nothing from them because the failures are politically invisible. That the mood has shifted is evidence only that a generation has forgotten why the trap was there.
Where this leaves us
Industrial policy is empirically back, as governing strategy, bipartisan, and global. That half of the title is settled, and it settled fast. What is not settled is the word doing all the work in the question: justified. The skeptic and the advocate agree the comeback happened; they disagree about whether it should have happened. And that disagreement splits cleanly into the two questions the next two stages take in turn, whether there is a real theoretical case for targeting these sectors, and whether the state can act on that case without making things worse.
The advocates claim something narrow and hard to dismiss: there are specific places where markets, left alone, will under-build the things a country needs. To judge the revival you have to know what those places are, and whether the programs are actually aimed at them.
The case for industrial policy
“Comparative advantage is not a fact of nature. Countries can, and routinely do, create it, through strategic investment in capabilities that did not exist before.”
— after Dani Rodrik, The Globalization Paradox, 2011
The strongest case for industrial policy is narrow and hard to refute: there are specific, nameable market failures that markets provably under-provide, and strategic sectors are where they cluster.
The serious case rests on four market failures, each a textbook category. None of them says the market is generally wrong; each says there is a particular thing the market, left alone, will build too little of, and that the shortfall is real.
1. Coordination failures. A new industrial cluster is a chicken-and-egg problem. The component suppliers won’t locate near a factory that doesn’t exist yet; the factory won’t come without nearby suppliers, trained workers, and infrastructure. Each investment is profitable only if the others happen, so privately none happens, the market is stuck at a bad equilibrium it cannot escape on its own. A state that coordinates simultaneous entry, or commits to the anchor investment, can move the whole cluster to the good equilibrium. This is the big-push logic of development economics, and a genuine failure of decentralized markets.
2. Learning and agglomeration externalities. When an industry learns by doing, unit costs falling as cumulative output rises, much of what it learns leaks out: trained workers move to rivals, tacit process knowledge diffuses, the supplier base improves for everyone. The first firm pays the full cost of climbing the learning curve but captures only part of the benefit, so private investment under-provides the learning. This is the infant-industry argument in its modern, defensible form; the social return to early production exceeds the private return. Smith and List founded the two sides of this argument, and the comparison of them puts both frameworks at their strongest.
3. National-security strategic goods. Some dependencies are dangerous in a way prices don’t register. When a country sources its most advanced semiconductors, its rare earths, or its defense-critical inputs from a single supplier that happens to be a geopolitical rival, the market price reflects the cheapest source and misses the cost of being cut off in a crisis. That gap is a real externality: the firm placing the order bears none of the strategic risk it creates for everyone else. Jake Sullivan’s “small yard, high fence” is the policy articulation, intervene narrowly, only where the security stakes are genuine, and resist the temptation to fence the whole economy.
4. Climate-transition coordination. The green build-out is half market failure by construction. Clean-energy capacity carries a positive externality the carbon price rarely captures in full, and the transition needs the same simultaneous-entry coordination as any new cluster, the grid, the battery supply chain, the charging network, and the manufacturers all have to move together or none moves at all. The Inflation Reduction Act is as much climate policy as industrial policy. How a society should pay for that transition is its own large question, taken up at depth in the walkthrough on how we should pay for climate change.
The proof-of-concept that these failures can be acted on is the East Asian record — postwar Japan, Korea, Taiwan, Singapore, then China — where targeted protection, directed credit, and export promotion built world-leading industries that comparative advantage said should never have existed. That record is the empirical base for the case, but it is not this walkthrough’s to retell: the head-to-head comparison lives in the comparative sibling on Japan vs. Korea as two developmental states, the late-industrializer story in the sibling on Germany vs. Britain, and the deep history of state economic capacity across eras in the state-formation thread. The China extension is documented in Economic History Ch.17 (China’s reform and the Asian century).
“There are real strategic-sector market failures that laissez-faire simply cannot fix, coordination, learning, security, and the climate transition are textbook, not slogans.”
— the calibrated case for targeted intervention
Are these failures real, or just convenient?
The case for industrial policy stands or falls on whether the four failures are genuine market failures or post-hoc rationalizations for favors. They are genuine, which is why the harder question is whether the state can fix them better than the market gets them wrong.
Create the advantage, or just identify the failure?
“The most successful catch-up economies of the past half-century — Korea, Taiwan, Japan, China — did not wait for comparative advantage to appear. They built it, deliberately, with policy.”
— after Dani Rodrik, calibrated-industrial-policy position
Rodrik’s argument is that comparative advantage is a snapshot. A poor country’s current advantage is in low-wage assembly; if it accepts that as fixed, it stays there. The fast developers refused to, they used strategic investment to move up into industries they had no business entering on a static reading, and the gamble paid off because learning and coordination effects are real and policy can switch them on. The point is that in the presence of these specific failures, the no-policy default is itself a choice to leave the failure in place. The state’s job is to create the capabilities, then expose them to competition to find out whether the bet worked.
“I grant the failures. I doubt the cure. Identifying where the market falls short is the easy part; knowing which intervention improves on it requires the planner to possess information the failure itself guarantees nobody has.”
— the market-failure-conceding skeptic
The honest skeptic concedes every one of the four failures and then asks the question Rodrik’s optimism glides over: which industry, which firm, how much, and for how long? Each answer requires knowing the size of a learning curve and the breadth of a spillover before either has happened, estimates the state must produce in an environment where every potential recipient is lobbying to inflate them. Rodrik is right that the no-policy default is a choice. So is the policy default, and the skeptic’s claim is that the second set of errors is, historically, larger than the first. Which is the question the next stage has to settle with the record.
Where this leaves us
The four failures are genuine, and intervention targeted at them is theoretically justified. The blanket “industrial policy is always a category error” position does not survive the taxonomy, and this is the strongest half of “back and justified.” But the case for carried its own undoing inside it: A market failure existing does not mean the state can fix it. That is the question every critic asks next.
Identifying a market failure is the easy part. The hard part is whether the state can fix it without creating a worse government failure, and the historical record on that is brutal. The same theory that justifies CHIPS today justified Latin American import substitution sixty years ago. One built champions. The other built a graveyard. The difference wasn’t the theory.
The government-failure and execution critique
“The infant-industry argument has been used to justify protection for industries that have been infants for fifty years. Tariffs imposed to nurture learning became permanent rents defended by the firms they enriched.”
— the Latin American import-substitution record, as the critics read it (Krueger / Bhagwati tradition)
The same theoretical case — coordination failures, infant-industry learning, build the capability and competition will follow — was made for Latin American import substitution in the 1950s through 1970s. The theory was right. The execution produced sheltered, uncompetitive industries, fiscal crises, and capture.
Grant every market failure from Stage 2. The critique asks whether the state can act on them without making things worse, and it has four strands, each as live and unsolved today as when the skeptics first raised it.
1. The picking-winners epistemology problem. To intervene well, the state has to know which industries have the steep learning curves, the large spillovers, and the right market structure to justify the bet, before any of it has happened. The information requirement is severe, and it gets worse on contact with theory. In the strategic-trade models, the optimal policy is not even robust to the kind of competition: if firms compete in quantities (Cournot), a subsidy shifts profits home; if they compete in prices (Bertrand), the optimal policy flips sign to a tax. A planner who guesses the market structure wrong subsidizes exactly what it should be taxing.
2. Capture and rent-seeking. An infant industry that survives becomes a politically powerful adult, with payrolls, plants in key districts, and a lobby whose entire purpose is keeping the support flowing long after the learning is done. The subsidy outlives the market failure that justified it, because the constituency it created now defends it. This is the public-choice critique in its sharpest form: policy is made by self-interested actors inside institutions that reward concentrated benefits and diffuse costs. The distance from “strategic industrial policy” to “permanent protection for politically connected firms” is shorter than advocates admit. The lineage of this argument — Buchanan, Tullock, and the public-choice tradition — is the subject of History of Economic Thought Ch.14 (Public choice tradition).
3. The Latin American record. This is the canonical failure, and the theory behind it was impeccable. Postwar import substitution took the infant-industry argument at face value, protected domestic manufacturing behind high tariffs and licenses, and waited for the champions to emerge. They didn’t. Without the discipline of having to export and compete, the protected firms optimized for the one thing the system rewarded, keeping the protection, rather than becoming competitive. The result was inefficiency, indefinite subsidy, fiscal strain, and the eventual debt crises of the 1980s. The history sits in Economic History Ch.14 (The postwar golden age and decolonization). ISI was a defensible policy fatally executed.
4. The preconditions problem. So why did East Asia succeed where Latin America failed, on the same theory? The answer the critics give is damaging for the revival: East Asia had an unusual bundle of preconditions that don’t transfer. An insulated, technocratic bureaucracy that could resist the firms it subsidized. Export discipline, the willingness to let protected firms fail if they didn’t become globally competitive, which is what forced learning instead of rent-seeking. And a Cold War strategic context that bought the state both autonomy and patience. Most states most of the time lack at least one of these. There is even a sharper version, the Krugman and World Bank reading that East Asia’s growth was mostly high savings, human capital, and catch-up convergence, and that the developmental state succeeded despite its interventions. That export-discipline preconditions are the binding constraint is worked out in the comparative walkthroughs; here it is the empirical reading.
“The state can’t pick winners, and even when it picks right, the winners pick the state, they capture the policy that made them and never let it go.”
— the government-failure critique, in one line
Can the state pick winners without being captured by them?
The critique has two engines: an epistemology problem (the state can’t know which industries to back) and a capture problem (even a right bet becomes a permanent rent). Neither has been answered. Both are why “justified in principle” doesn’t equal “will work in practice.”
Government failure vs. the discipline that worked
“We remember POSCO and forget the dozens of state steel mills that consumed subsidies for decades and never became competitive. The selection bias is the whole story. Policy made by interested actors in capturable institutions trends toward the graveyard, not the exception.”
— the public-choice critique, at full current strength
This is the critique as a living position. It grants that the failures are real and that interventions sometimes work. Its force is in the base rate. For every developmental-state success there is a long tail of protected industries that absorbed public money indefinitely, defended by the constituencies the subsidy created, invisible in the success stories because failed bets don’t get written up. The epistemology problem guarantees a high error rate going in; the capture dynamic guarantees the errors are hard to reverse once made. A policy regime that picks wrong often and can’t cut its losses is not redeemed by its rare wins. The honest baseline for any new program is the graveyard, and the burden is on the advocate to show why this program escapes it.
“The Korean state did not protect firms and hope. It protected them and demanded exports, and cut off the ones that failed. Discipline, not targeting, is what the successful cases had and the failures lacked.”
— after Alice Amsden & Robert Wade, developmental-state scholarship
The rebuttal accepts the base rate and asks what the exceptions had in common. Amsden’s and Wade’s answer is export discipline: the successful developmental states made support contingent on performance in world markets, a test the favored firms could not lobby their way out of. That is the precise institutional answer to the capture problem, a sunset mechanism with teeth, enforced by competition the state could not fake on the firm’s behalf. So the conditions under which industrial policy works are demanding but nameable: an insulated bureaucracy, a hard performance test, a credible willingness to cut losers. The disagreement with the critique is narrower than it looks. Both sides agree the binding constraint is execution. They disagree about how often a given state can meet it, which is the calibrated question the verdict has to answer.
Where this leaves us
The critique survives the case. Identifying a market failure is necessary but not sufficient: the binding constraint is whether the state can execute — capacity, export discipline, accountability, and a credible willingness to let losers fail — without being captured. The historical record shows these conditions are demanding and rarely met; East Asia is the exception that demonstrates how hard the rule is, and it does not travel as a template. The picking-winners epistemology problem is unsolved and the capture dynamic is structural. But the rebuttal is equally real: the conditions can be met, they were met, and what met them was export discipline. So the disagreement has migrated to execution quality, and that is where the verdict lives.
So we have a real theoretical case and a brutal execution record. Put them together and the question — is industrial policy back, and is it justified? — resolves into a test you can run on any program.
The verdict: settled in principle, live on execution
“We are all industrial-policy people now.”
— after Dani Rodrik, on the mainstream’s shift, 2010s
The people now conceding a role for industrial policy include economists who were its skeptics a decade ago. The frame moved, and the question “is it back?” is, underneath, a question about that move.
Stage 2 gave the four market failures: coordination, learning, national-security, and climate-transition coordination. These are what justifies intervention, the conditions under which leaving the market alone is a choice to leave a failure in place. Stage 3 gave the four execution conditions: state capacity, export discipline, accountability, and hard sunsets. These are what determines whether intervention works, the conditions under which a justified intervention becomes a successful one rather than a captured one.
The in-principle question (is there a real failure here?) is largely settled. The execution question — can this state, with this program, act on the failure without making things worse? — is open, program by program, and is where every honest remaining disagreement now sits. Conflating the two is the error that produces both the breathless advocacy and the reflexive dismissal.
Inside the moved mainstream
“The right model is neither free trade nor protectionism. It is selective engagement, with industrial policies calibrated to institutional capacity, and capacity is the binding constraint, which, I admit, describes most of the developing world.”
— after Dani Rodrik, the moved-mainstream position
Rodrik holds the position the mainstream moved toward, and notice how much of the critique it has already absorbed. He claims comparative advantage is incomplete and can be deliberately created, but only where institutional capacity exists to enforce the discipline that makes creation work. His own qualification is the critique’s concession built into the case: capacity is the binding constraint, and most places lack it. It is a calibrated yes, justified where the failure is real and the institutions can execute, which is a smaller set of cases than the enthusiasts want and a larger set than the blanket skeptics allowed.
“The failures are real. The cure’s record is poor. We remember the successes and forget the dozens of failures, and the broad, undisciplined, unmonitored programs are the ones we should fear most.”
— the calibrated free-trade skeptic, holding the residual caution
The skeptic here speaks from inside the new consensus, voicing its residual caution. The concession is total on the theory: the failures are real, targeted intervention can be justified. The caution is about the base rate and the selection bias, we write up POSCO and forget the graveyard, so our intuitions about how often this works are systematically too optimistic. And the programs most likely to fail are the broad, undisciplined, unmonitored ones, the ones that subsidize widely without a hard performance test or a credible sunset. The recent record of sweeping, poorly-targeted intervention, tariffs imposed broadly and never reviewed, is the cautionary tale for what undisciplined industrial policy looks like in a rich country with strong institutions. Even there, the discipline is the hard part.
The verdict
Industrial policy is back, and it has a legitimate theoretical basis in specific market failures: coordination, learning, national-security, and climate-transition coordination. The in-principle question is largely settled, yes for those specific failures. But success is conditional on demanding execution — state capacity, export discipline, accountability, hard sunsets — the conditions that separated East Asian success from Latin American failure, and which the record shows are hard to maintain. The live question is execution quality.
Three findings hold that verdict together. The headline is that the mainstream frame moved over a decade, from “industrial policy is a category error” to “industrial policy is justified for specific failures, conditional on execution,” and the proof is that the people saying so include the skeptics of ten years ago. On the evidence, the East-Asia-versus-Latin-America comparison is decisive: export discipline was the difference; it was necessary, demanding, and not portable. The magnitudes, how large the industrial-policy multiplier is, how much of CHIPS and the IRA is genuine additionality versus crowding out private investment, how much of the East Asian miracle was the developmental state versus savings and convergence, remain unsettled, and the 2020s programs are too new for an outcome verdict. This is not a “you decide” punt: the frame has moved, the in-principle case is real for specific failures, and the live action is execution quality.
Where this leaves us
We started with a word economists used as an insult that is now the law of the land, CHIPS, the IRA, the EU’s answer to both. The four stages took the title apart in order:
- The revival is real. Industrial policy came back as bipartisan governing strategy, in the US and abroad, inside a decade. “Back” is settled as a fact about policy.
- The in-principle case is real and specific. Four textbook market failures — coordination, learning, national security, climate — justify targeted intervention in particular sectors. The blanket “category error” position does not survive them.
- The execution record is brutal. The same theory that justifies CHIPS justified Latin American import substitution. The difference was execution — export discipline, state capacity, the willingness to cut losers — and those conditions are demanding and rarely met.
- Settled in principle, live on execution. The frame moved; the in-principle case holds for specific failures; the genuine remaining disagreement is about execution quality, program by program.
The headline is the frame shift itself. A decade ago, “industrial policy” ended an argument; today it starts one. That move is the mainstream conceding that a horizontal-policy-only toolkit left real, nameable holes, and the evidence of the concession is that its loudest critics now make qualified versions of the case. The question “is it back?” is, at bottom, a question about that concession, and the answer is yes.
The takeaway is a test you can run on any program, in any country, in any decade. Ask two questions. First, the market-failure question: is there a real, specific failure here — a coordination problem, a learning spillover, a strategic dependence, a climate externality — or is “strategic” just a label on a favor? Second, the execution question: does this state have the capacity, the export-or-equivalent discipline, the accountability, and the credible sunset to act on the failure without being captured by the firms it helps? A program that fails the first test is a category error. A program that passes the first and fails the second is Latin America. A program that passes both is the rare, demanding thing East Asia managed, and the whole debate, once you separate “back” from “justified,” is the work of telling the three apart.