Governments are reexamining technology as industrial policy confronts an era defined by artificial intelligence and uneven economic gains.
Dr. Michael Mandel, chief economist and vice president at the Progressive Policy Institute, told Competition Policy International (CPI), a PYMNTS company, that the core issue is not whether innovation is occurring. It is where the benefits are going.
“We’ve got rapid productivity growth in the information sector,” he said, “but productivity growth in the physical sector has slowed down to close to zero.” That gap has left entire industries and regions behind, creating what Mandel described as both an economic and political problem now at the center of industrial policy.
The Productivity Gap at the Heart of Modern Industrial Policy
Tech industrial policy is an effort to correct that imbalance. Over the past two decades, software and digital services have delivered measurable gains. Meanwhile, sectors like construction, agriculture and manufacturing have failed to keep pace.
Mandel stressed that this was not the expected outcome. Early in the information revolution, the assumption was that digital tools would lift productivity across the entire economy.
“Everybody needs to be focused on this as the main problem to be solved by industrial policy,” he said, pointing to the need to restore momentum in physical industries that underpin employment, affordability and regional stability.
The Limits of Relying on Markets Alone to Drive AI Growth
This challenge is taking shape as governments expand their role in technology markets, particularly in AI, semiconductors and advanced manufacturing. Mandel argued that the United States cannot rely solely on private capital to address these structural gaps.
Public policy must guide priorities while still allowing for decentralized execution and competition. In practice, that means setting national goals for AI infrastructure, semiconductor capacity and advanced manufacturing, then creating incentives for private and regional actors to compete in achieving them.
What the U.S. Can Learn From China’s Decentralized Industrial Model
China offers a reference point for how such a strategy can work at scale. Its system combines central direction with local competition, letting provinces and municipalities pursue national goals through their own funding and development efforts.
Mandel described this as a form of state-supported venture capital, where regions compete to build industries, absorb risk and accelerate innovation. The approach has produced inefficiencies, including overcapacity and uneven returns. But it has also driven real gains in productivity and global competitiveness.
For the United States, the lesson is not to copy China’s institutional model. It is to draw from its emphasis on decentralized investment and competition. Mandel suggested that federal policy should define strategic priorities, while state and local governments take a more active role in financing, experimentation and implementation.
Measuring Economic Performance in a Software-Driven Economy
U.S. policy must also evolve in how it measures economic performance, Mandel said. The tools used to assess output were built for a different era, when industrial production was the primary indicator of capacity.
“If you go back to the origins of the national income accounts, they were designed … to identify what the capacity of the U.S. was to produce to wage war … and maintain consumer spending at the same time,” he said. In an environment where software, data and computing power shape both economic and strategic outcomes, those metrics no longer capture what matters most.
“We do not have any way of tracking our software generating capabilities in a way that makes sense,” Mandel said. Without better tools, it becomes hard to assess whether AI investments are producing real gains or simply increasing activity in sectors that are already productive.
Spreading AI Gains Beyond the Digital Core of the Economy
The broader goal, in Mandel’s view, is to ensure that technological progress reaches beyond the digital core. The past two decades saw an unusual pattern: innovation failed to raise productivity where it was most needed.
Closing that gap requires a focus on diffusion, not just innovation. Mandel pointed to the need for policies that translate AI advances into practical use across industries.
Europe offers an additional lesson. Over the past decade, European regulators pursued extensive interventions targeting large technology firms, expecting those measures to stimulate competition and productivity. Mandel said the data do not support that outcome. “There’s no sign in the data at this point that multiple layers of regulation have done anything to accelerate European productivity growth,” he said.
The Actions the U.S. Must Take Over the Next Decade
Looking ahead, Mandel outlined what he believes should guide U.S. policy over the next decade. First is sustained investment in data centers and energy resources, which underpin both economic activity and national security. “We need to fund the AI infrastructure … that’s essential,” he said.
Beyond infrastructure, execution at the state and local level matters. Governors and regional leaders, he suggested, should take a more active role in deploying AI across industries through training programs, extension initiatives and targeted business support. That would help ensure productivity gains are distributed across sectors and regions, not concentrated in large technology firms.
Investment must also align with outcomes. Policies should be evaluated not just on spending volume, but on their ability to raise productivity in lagging industries, improve wages and expand economic participation.
Mandel returned to a lesson from recent history, when expectations about the spread of digital innovation proved too optimistic. “We can’t do that again,” he said. And on the necessity of investing in technology infrastructure: “We cannot increase the productivity of the lagging industries without it.”
