Innovation, Incentives, and Technology

Session 6 · Thu Sep 17

Why does steady technological progress require more researchers over time, and what does that mean for institutions that support innovation?

The law that became harder to obey

In April 1965, Gordon Moore was director of research at Fairchild Semiconductor, one of the young firms making the Santa Clara Valley into a center of electronics. Asked to predict the future of integrated circuits, Moore noticed a striking pattern: the number of components that could be packed onto a chip had roughly doubled each year. He expected the trend to continue.

The prediction became famous as Moore’s Law. For decades, chip density kept doubling on a regular schedule. Computers became smaller, cheaper, and more powerful. The compounding was astonishing: the laptop on a desk and the phone in a pocket descend from a manufacturing process that learned to place billions of transistors where earlier engineers placed thousands.

But the law did not obey itself. Each doubling required fabs, clean rooms, precision machines, materials science, design software, and teams of engineers. Bloom, Jones, Van Reenen, and Webb estimate that maintaining Moore’s Law today requires more than 18 times as many researchers as it did in the early 1970s.

The surface story is one of effortless exponential progress. The underlying story is stranger: the same rate of improvement took more and more effort to sustain.

The question this session is about: why do ideas seem to get harder to find, even in sectors where progress looks steady from the outside?

To answer that, we need to connect two readings. UOE Chapter 4 explains why technological innovation is central to rising living standards. Bloom et al. ask what happens when the search for new technologies itself becomes more difficult.

Use Moore’s Law as the bridge from Session 5 to Session 6. Session 5 asks why basic knowledge is underprovided; Session 6 asks what happens after societies build innovation systems and discover that sustaining progress requires increasing effort. Write the Bloom equation on the board before showing any graph: growth = research productivity x researchers.

The Red Queen problem of research

Bloom et al. start with a simple accounting identity from growth theory:

Economic growth depends on both the number of researchers and the productivity of those researchers at finding useful new ideas.

That identity helps explain the puzzle. If research productivity were constant, a constant number of researchers could keep generating constant economic growth. But if research productivity falls, then society has to add more researchers just to keep the growth rate from slowing.

This is sometimes called a Red Queen dynamic, after the character in Through the Looking-Glass who has to run as fast as she can just to stay in the same place. The economy may be doing the same thing: adding scientists, engineers, labs, and R&D budgets in order to preserve growth rates that look stable from the outside.

The chart below shows how R&D spending has become a larger part of many advanced economies. It does not prove Bloom et al.’s argument by itself, but it makes the scale of the effort visible: modern growth depends on an increasingly large research enterprise.

Source: UNESCO UIS / World Bank via Our World in Data. Note: the chart shows total R&D spending as a share of GDP; Bloom et al.’s claim is about research productivity, not spending alone.

If ideas are getting harder to find, should that make us more pessimistic about progress or more serious about funding and organizing research? What would be the wrong conclusion to draw from Bloom et al.?

Bloom et al. decompose long-run growth into which two parts?

  1. Capital accumulation and population growth.
  2. Research productivity and the effective number of researchers.
  3. Patent counts and venture-capital investment.
  4. Prices and property rights.

Students may hear “ideas are getting harder to find” as “innovation is over.” Push the opposite reading: if research productivity is falling, then institutions, funding, research organization, and talent allocation matter more, not less. The paper is an argument for taking the innovation system seriously.

Incentives change what gets invented

UOE Chapter 4 adds a second piece: invention is not just a technical process. It is shaped by incentives.

If a firm expects to earn an innovation rent, it has a reason to search for a better product or process. If a scientist expects credit for being first, she has a reason to publish. If a government offers a prize, grant, procurement contract, or patent, it changes what people work on.

That means the direction of technological change is not automatic. Societies do not simply receive “technology” as a neutral force. They build rules that steer talent toward some problems and away from others. A patent system may reward drug discovery, but not necessarily clean water infrastructure. Venture capital may reward software with fast scaling, but not slow, capital-intensive technologies. Public science funding may support basic research, but not the messy work of turning a discovery into a usable product.

Innovation is not only a matter of having clever people. It is a matter of what society rewards clever people for doing.

Imagine two problems: a better social media recommendation algorithm and a cheaper way to retrofit old apartment buildings for heat pumps. Which one is more likely to attract private investment quickly? What does your answer reveal about incentives?

Why does the concept of innovation rent matter for understanding technological progress?

  1. It guarantees that successful innovators keep monopoly profits forever.
  2. It explains why firms have no reason to invest in research.
  3. It gives firms a reason to search for new products and processes before competitors do.
  4. It only applies when governments directly fund invention.

This is a useful moment to connect back to Nelson without repeating him. Nelson’s problem was that basic research creates spillovers. UOE’s point is that incentives can still power innovation when benefits are appropriable enough, or when institutions make them appropriable. The hard policy question is choosing the right incentive for the type of knowledge being produced.

Harder ideas, bigger teams

Bloom et al.’s paper is especially useful because it makes an abstract worry concrete. In semiconductors, agriculture, medical research, and firm-level data, they find a repeated pattern: more research effort is needed to generate the same rate of improvement. Their aggregate estimate implies that research productivity in the United States has been falling sharply over time.

Why might this happen? One possibility is that the easiest discoveries are made first. Another is that frontier work requires deeper specialization, longer training, more expensive equipment, and larger teams. A student in 1900 could reach the frontier of some scientific fields with a few years of study and modest equipment. A student in 2026 may need a PhD, access to a specialized lab, massive datasets, or a production facility worth billions of dollars.

That does not make progress impossible. It changes its institutional requirements. If ideas are harder to find, then growth depends on how well we organize teams, fund public goods, share knowledge, train researchers, and connect science to production.

The harder ideas are to find, the more progress depends on institutions that help many people search together.

Do bigger teams and more specialized knowledge make innovation more democratic, because more people are involved, or less democratic, because only large organizations can reach the frontier?

What is the best interpretation of Bloom et al.'s Moore's Law example?

  1. Moore's Law stopped working in the 1970s.
  2. A stable rate of chip improvement can hide a large decline in research productivity.
  3. Semiconductors are uniquely immune to diminishing returns.
  4. Consumers stopped wanting faster computers.

The “bigger teams” point sets up Session 8 on Silicon Valley. Students should begin to see that innovation systems solve coordination problems: who knows what, who works with whom, who funds whom, and how quickly ideas move across boundaries.

Connections

Builds on: Session 5 explained why basic research has public-good properties. Session 6 asks what happens when the innovation frontier advances and the search for useful ideas becomes increasingly resource-intensive.

Sets up: Session 8 moves from incentives in general to the geography of innovation: why some places, like Silicon Valley, organize the search for new ideas better than others.

Arc note: This session is the hinge between science policy and innovation geography. Students should leave with the idea that progress is neither automatic nor exhausted. It is produced by institutions that shape incentives, coordinate teams, and sustain the search when each additional idea is harder to find than the last.

If research productivity is falling, what follows most directly?

  1. Economic growth must immediately stop.
  2. Sustaining growth may require more effort or better institutions for producing ideas.
  3. Patents have no effect on innovation.
  4. Competition alone will automatically solve the problem.

Review cards

Work through these cards now, then Orbit will schedule them for review over the coming weeks.

Reading guide

Required

Halliday, Simon D., and Eric Bottorff. Understanding Our Economy, Chapter 4: “Innovation, incentives, and technology” (selections). Available at core-understanding-our-economy.netlify.app.

What to look for: Track the incentive mechanism. What makes someone search for a new technology rather than keep using the old one? Pay attention to the difference between invention itself and the institutional conditions that make invention worthwhile.

Key argument: Technological progress is shaped by incentives: the possibility of innovation rents, and the institutions that allocate rewards, help determine both the pace and direction of innovation.

Prepare to discuss: One example of a socially valuable technology that markets seem to reward strongly, and one that markets seem to under-reward.

Bloom, Nicholas, Charles I. Jones, John Van Reenen, and Michael Webb. “Are Ideas Getting Harder to Find?” American Economic Review 110, no. 4 (2020): 1104-1144. Read the introduction and conclusion only.

What to look for: Focus on the simple equation in the introduction and the Moore’s Law example. You do not need the technical sections to understand the central claim. Ask yourself: if the number of researchers has to rise just to keep growth constant, what does that imply about the future of progress?

Key argument: Research effort is rising while research productivity is falling; ideas are getting harder to find, so sustained growth requires increasing effort or better institutions for producing ideas.

Prepare to discuss: The Moore’s Law statistic: maintaining the doubling of chip density now requires more than 18 times as many researchers as in the early 1970s. What does that change about how you see technological progress?