Policy, Not Tech Alone, Drives True Economic Growth

The National Bureau of Economic Research piece titled “Technology, Productivity, and Economic Growth” does something valuable: it walks through the mechanisms, the production functions, the channels through which innovation shows up in GDP. It reads like a clean model. But models have a habit of assuming away the mess that decides who actually wins. Who pockets the productivity dividend? Follow the money.

The temptation—reasonable, even comforting—is to treat technology as a neutral multiplier. Push a button; output rises. The article rightly insists technology matters. But technology is an accelerant only when someone already has fuel and a road. Drop the same AI system into a logistics giant and a mid-size local retailer; one will rewire its supply chain, the other will stare at a dashboard it doesn’t have the staff or systems to use.

Why? Because technology needs complements—capital, managerial retooling, trained workers, and institutions that let firms scale. Those are not evenly distributed across places or companies. Convenient, isn't it, that the firms with the cash and legal teams are also the ones best placed to turn new tools into monopoly-like returns.

Think of productivity as a highway. Technology is the pavement. But there are toll booths everywhere—access to finance, skilled labor pools, regulatory regimes that favor incumbents, procurement processes that reward scale. The article charts technology’s role; it underplays these tolls. A small manufacturer in Cleveland or a public school in rural Texas may get the same software available to a Bay Area unicorn, and yet the outcomes diverge. The unicorn can buy complementary data, attract engineers, and litigate around intellectual property; the smaller actor cannot.

Those toll booths aren’t abstract.

Look at how cloud computing played out. Amazon didn’t just roll out servers; it bundled them with pricing power, tie-ins to its marketplace, and a hiring pipeline that vacuumed up specialized talent. Smaller firms gained access to tools but also found themselves renting the infrastructure of a future rival. The technology lifted productivity—on Amazon’s terms.

The NBER framing is useful when it insists on mechanisms; less useful when it treats diffusion as something that “eventually” happens, as if time alone melts barriers. Technology does diffuse—but unevenly and often slowly. The piece nods at diffusion without interrogating the bottlenecks: mismatched skills, archaic labor rules, and market concentration that lets a few platforms internalize most of the gains. Here’s what they won’t tell you: patents, network effects, and exclusive data deals are modern toll booths. They shape who benefits from innovation as much as the invention itself does.

History already wrote an early draft of this story.

Electrification transformed factories, but only after managers rewired floor layouts, retrained workers, and scrapped old routines. For years, many plants simply swapped steam for motors and saw little gain. The constraint wasn’t volts; it was organization. Today’s AI and automation tools are in that same limbo in thousands of offices and schools—installed, not integrated.

If technology’s promise depends on context, then policy is not a side note; it’s the missing complement. Education and training are obvious levers—no argument there. But policy also has to lower those invisible tolls: competition rules that actually bite; public procurement that doesn’t automatically default to the biggest vendor; broadband treated as basic infrastructure, not a luxury; worker mobility so skills can actually move to where new tools are deployed; a tax code that doesn’t reward parking intangible assets in legal limbo.

That’s not about punishing successful firms. It’s about making sure the highway has more than one on-ramp.

Proponents will counter that this is too cynical, that technology has historically lifted living standards broadly and markets will spread gains in time. There’s some truth there—printing presses, vaccines, and semiconductors all started narrow and went wide. But history also shows long lags and winners-take-most dynamics. Diffusion is not destiny. Treating those lags as harmless “adjustment periods” quietly writes off towns, sectors, and generations as acceptable collateral.

Measurement muddies the water further.

If productivity is recorded mainly in goods-producing firms and misses quality improvements in services, health, or education, the assessment is incomplete. If an AI tutor helps a struggling student grasp algebra but test scores and wage data don’t register that boost for years, the official story will lag the lived one. When the article leans on canonical productivity accounting without highlighting these gaps, it hands policymakers a partial map.

And maps tell you where to build roads—and where they think the road ends.

I agree with the article’s core: technology is central to growth. But treating technological change as a self-executing engine minimizes the institutional work required to spread gains. That’s not academic hair-splitting; it’s a recipe for policy complacency, the kind that lets officials boast about “innovation ecosystems” while half their local firms never get past email and spreadsheets.

So yes, read “Technology, Productivity, and Economic Growth.” Then read your city budget, your school board’s procurement rules, and your antitrust filings side by side. The distance between those documents will tell you exactly how much of that promised productivity will ever show up in your neighborhood.