Industrial Policy Under Uncertainty
How governments steer innovation when they no longer know the winners in advance
“Industrial Policy for an Age of Uncertainty”, Project Syndicate, 25 May 2026, by Yuen Yuen Ang
WASHINGTON, DC—Industrial policy is back, and with a vengeance. After decades of preaching neoliberalism, Western policymakers and intellectuals have rediscovered the role of the state in economic development. Even the World Bank has jumped on the bandwagon, conceding that its old advice now has the “practical value of a floppy disk,” and that every country should make industrial policy part of its national toolkit.
But swinging from one end of the spectrum to the other is not the same as thinking anew. Much of the hype over industrial policy misses a new reality: today, simultaneous disruptions—AI, deglobalization, climate shocks, and geopolitical frictions—make it harder for policymakers to know which winner to bet on.
In the 20th century, the playbook for late industrializing countries demanded disciplined execution, but at least it was clear: build infrastructure, attract factories, manufacture for export, and climb the value chain. Today, the path is far less obvious. Moreover, disruptive innovation, by definition, cannot be predicted. By the time policymakers realize which technologies truly matter (such as semiconductors), the train may already have left the station.
The debate is not whether governments should intervene—most already do. Instead, the real question is: what can governments do when they do not know in advance what will work?
Selecting winners is even more difficult among advanced Western economies, because their policy reflects not only the pursuit of competitiveness, but also security concerns. Once the logic of national security takes hold, almost everything can be declared “critical” or “strategic”—be it semiconductors, electric vehicles, fertilizer, masks, baby formula, painkillers, or even lumber.
The debate is not whether governments should intervene—most already do. Instead, the real question is: what can governments do when they do not know in advance what will work?
One answer from the World Bank and others is to adapt and experiment. But who could disagree with that? Left unsaid is how exactly governments should enable adaptation.
Ten years ago, in How China Escaped the Poverty Trap, I introduced the concept of “directed improvisation.” Governments that rely on this approach do not dictate outcomes or pick winners in advance. Instead, they act more like the director of an improvised play, or like the programmer of a self-generating computer game. Their job is to create the conditions for bottom-up, spontaneous creation and shape a variety of possibilities by drawing boundaries, defining and rewarding success, and scaling up what works.
My premise is that uncertainty—all the unknown possibilities, both good and bad—is distinct from risk, meaning negative outcomes whose probability of occurrence can be calculated and controlled. Under uncertainty, preselecting outcomes is not only futile but potentially counterproductive, since it may lead governments to miss opportunities that defy their initial expectations. Instead of controlling, governments must exercise influence.
Ten years ago…I introduced the concept of “directed improvisation.” Governments…act more like the director of an improvised play...Their job is to create the conditions for bottom-up, spontaneous creation and shape a variety of possibilities…
One way to do that is through what I call adaptive policy communication. When faced with a novel challenge, governments can initially issue grey (ambiguous) signals that neither endorse nor prohibit activity. Then, when pilots prove effective, they can endorse them explicitly, while drawing red lines to prevent intolerable risks.
This mechanism can be systematically measured. In a recent study, my co-authors and I used large language models to classify Chinese central policy directives and track how signals have evolved over time and across policy domains. What we found undercuts the stereotype of China as the Darth Vader of industrial policy—an omnipresent state that thinks in decades, picks winners, pours in subsidies, and aims to dominate the world.
Consider e-commerce, a major engine of inclusive growth. China now has the world’s largest e-commerce industry, accounting for roughly 50% of global sales and generating an estimated 50-70 million jobs. E-commerce platforms accelerate the sales of countless retailers and manufacturers of consumer goods like lighters, lamps, and socks, none of which matter geopolitically, but all of which are indispensable to modern life—not least in America.
Instead of controlling, governments must exercise influence. One way to do that is through what I call adaptive policy communication. When faced with a novel challenge, governments can initially issue grey (ambiguous) signals…Then, when pilots prove effective, they can endorse them explicitly....
But was e-commerce one of the sectors targeted by the government’s Made in China 2025? The first central directive related to e-commerce was issued in 2000, when the State Council referred to “internet information services,” defined narrowly as providing information online and designing webpages for fees. Central planners did not foresee the rise of e-commerce.
Nor did the government take a hands-off approach. As the sector grew, the state took notice and adapted its strategy. In 2015, the State Council approved a request from the Zhejiang provincial government to build experimental zones for cross-border e-commerce, urging it to “support bold innovation” and “adjust according to the times.” In plain language, this was a green light to scale up an industry that had shown proof of concept, with local authorities lobbying the central government for support.
The lesson for other countries is not that they should copy China. It is that under uncertainty, the challenge for industrial policy is not simply picking winners but discovering them. And once they are discovered, it is essential to tailor support beyond the usual toolkit of tariffs and subsidies. Some of the most effective government support in e-commerce came from policies that encouraged traditional industries to integrate themselves into digital platforms.
China is not the only illustration. Who would have predicted that Nigeria would produce the world’s second-largest film industry by volume? Or that fashion, empowered by e-commerce, would become one of Africa’s fastest-growing creative industries?
The lesson for other countries is not that they should copy China. It is that under uncertainty, the challenge for industrial policy is not simply picking winners but discovering them.
More broadly, many vibrant, creative, digital sectors that are emerging from the bottom up in Africa defy the old industrial-policy vision of foreign-invested factories and industrial parks. Yet through my own research in Nigeria, I found that entrepreneurs in these emerging sectors face obstacles when they seek support to scale up operations. Banks hesitate to lend to those without physical collateral, and technocrats overlook businesses that do not appear “strategic.”
In the 21st century, uncertainty—unknown possibilities—is not a bug, but a feature. Designing industrial policy under such conditions requires more than pivoting away from neoliberalism. It demands a new way of thinking that I call AIM (adaptive, inclusive, moral) political economy.
AIM challenges a core assumption of top-down planning: that policymakers know the answers in advance. Increasingly, they do not. But this does not mean helplessness. It means crafting meta-institutions that enable discovery and innovation—which often emerge unexpectedly.
Reprinted from Project Syndicate. Read the original article here.
Learn More
Read Yuen’s introductory essay on AIM — Adaptive, Inclusive, Moral Political Economy at Oxford Development Studies
Read “Ambiguity and Clarity in China’s Adaptive Policy Communication” (2023) open access at The China Quarterly
Read “CAPC-CG,” an empirical extension of the above using LLM annotation, accepted at ACL (Association of Computational Linguistics) 2026
Interesting article which has parallels with hybrid agile project management which I have implemented for international clients. Simply put. Identify the key things needing attention. Form a diagonal slice team. Identify the working process and then iteratively improve. Experimentation and trying new things being key. It it works adopt and extend. If not bin and put efforts elsewhere.
A great read, thank you for sharing!
The idea of “directed improvisation” is compelling precisely because it treats uncertainty as the central condition governments must manage, not something they can simply plan away.
A thoughtful and timely argument.