What this covers

Michael Kremer makes the case that development economists must go beyond producing research to actively founding and operating institutions that test and scale solutions. Rather than conducting studies in isolation, he argues economists should use experimental method to identify what works, then iterate toward institutional designs that direct innovation toward the poorest. The conversation centers on two main tools: advance market commitment, which uses government purchase guarantees to incentivize vaccine and technology development in neglected areas, and field experiments that identify and measure which interventions—like deworming or merit scholarships—produce lasting gains. Throughout, Kremer illustrates how the disciplines of entrepreneurship and research can reinforce each other: running real organizations forces economists to understand actual constraints and customer needs rather than abstract problems.

The substance ranges across vaccines, education, and agricultural health. On vaccines, he details how a $1.5 billion advance market commitment for pneumococcus strains succeeded in development and vaccination where markets alone had failed, and proposes the same mechanism for clean cookstoves and other innovations facing commitment and public-good problems. In education, he examines why poor-country curricula often target high performers while most students fall behind, and presents evidence that remedial systems and merit scholarships can produce substantial learning and lifetime earnings gains. A recurring theme is Ideas getting harder to find—following Chad Jones, he notes that innovation increasingly requires larger teams and greater investment to achieve the same progress, creating a race between resources and difficulty that institutions must navigate. He also contests the once-common claim that culture bars development: rapid convergence between poor and rich countries in recent decades, visible in India, China, and beyond, suggests institutional change matters far more than the historical or cultural factors economists once emphasized. Throughout, he positions development innovation using venture capital portfolio logic, where most interventions fail but successful ones yield returns far exceeding costs.

Sharpest takeaway

Kremer argues that development economists should not merely produce research but build institutions and use experimental, iterative methods to design innovation-funding mechanisms (like advance market commitments) that direct technological change toward human needs and scale proven interventions.

  • The experimental method has a natural affinity with founding and operating organizations
  • Advance market commitments successfully incentivized vaccines (pneumococcus) by committing to buy efficient quantities
  • Poor countries are now converging with rich ones, showing development is achievable through better institutions and interventions

The argument · threads7 threads · 26 claims
0.80

Advance market commitments solve vaccine market failures by guaranteeing purchase of efficient quantities, incentivizing development while avoiding monopoly quantity restrictions.

5 pointscentrality 5/5
  • Although patent buyouts (buying patents and placing IP in the public domain to reward innovators without monopoly distortions) were academically attractive, they failed politically—likely because pharmaceutical firms feared it would delegitimize the IP system—whereas advance market commitments, which involve selling products rather than patents, proved politically acceptable.

    “They may have felt that if they sold their patents and put them in the public domain, that would weaken the legitimacy of that system”

  • Vaccine markets contain a commitment problem and a global public-good problem: governments may not pay the full social value of a vaccine, under-incentivizing firms, and for diseases like malaria spread across many small African countries, no single country has an incentive to pay a price generating socially appropriate development incentives.

    “No one of those countries has an incentive to say, “Well, let’s pay a price that would generate the socially appropriate incentives for developing a malaria vaccine.””

  • The first market failure with vaccines is the communicable-disease externality: when one person is vaccinated they avoid infecting others who then don't infect still others, eventually protecting people the vaccinated person never directly contacts, so individuals buying vaccines on their own would underinvest relative to social value.

    “when I take a vaccine that would benefit you if you were closer or nearby, but indirectly, it even benefits you. Because I don’t infect somebody, they don’t infect somebody, and eventually, they don’t infect you.”

  • A key distortion of monopoly is that high monopoly pricing reduces the quantity sold, which for pharmaceuticals can be very harmful to human health; advance market commitments avoid this by committing in advance to buy the efficient quantity, simultaneously creating an incentive for vaccine development and ensuring the efficient quantity is purchased.

    “when there’s monopoly pricing, less of this gets sold. When less of this gets sold, that’s harmful for many products, but in the case of pharmaceuticals, that can be very harmful to human health.”

  • A $1.5 billion advance market commitment pledged by a number of countries together with the Gates Foundation incentivized the development of three pneumococcus vaccines targeting strains common in the developing world, vaccinating hundreds of millions and saving an estimated 700,000 lives.

    “Hundreds of millions of people have been vaccinated, and 700,000 lives are estimated to have been saved by those pneumococcus vaccines.”

0.75

Institutions governing innovation—IP rules and research funding—should be experimentally redesigned to accelerate technological change and direct it toward human needs.

4 pointscentrality 5/5
  • In general, innovators capture only a fairly small fraction of the value they create—estimates of the social rate of return relative to private return on R&D investment often reach numbers like 70 percent—which justifies boosting R&D incentives across a wide variety of fields, not just vaccines.

    “People who’ve estimated the social rate of return to private investment and R&D often estimate numbers like 70 percent.”

  • A fundamental driver of long-run growth is technological change, so economists should focus more on which institutions—primarily intellectual property and research-funding institutions—can increase the rate of technological change and direct it toward human needs, experimenting with new approaches to improve them over time.

    “a fundamental driver of growth in the long run is technological change. One thing that we economists should be thinking about more is, what are the institutions to increase the rate of technological change and to try to direct it to human needs?”

  • Research-funding institutions should be experimented with—e.g., funding researchers five years past PhD based on what they have accomplished rather than on grant proposals, varying committee sizes, or funding work if any single reviewer strongly likes it rather than taking the committee average—because optimizing how research is funded should be a priority for anyone who cares about economic growth.

    “we’re going to fund you based on what you’ve accomplished, not based on your grant proposal.”

  • There may be something fundamental about technological progress such that we invent more technology over time but do not increase the growth rate of technology; this could change if AI works well enough to create a virtuous cycle toward a singularity, but since we cannot control whether it is a fundamental constraint, the actionable focus should be on the institutions we can control—IP rules and research-funding design.

    “We invent more technology. We don’t increase the growth rate over time. That’s just something inherent about technology. Obviously, that could change.”

0.61

Low-cost, high-impact interventions like deworming and merit scholarships demonstrate scalable solutions to development problems when behavioral and institutional barriers are addressed.

4 pointscentrality 4/5
  • Deworming, which can cost as little as 50 cents a year per person, produced phenomenal health and education gains in early Kenya trials, and tracking people over 20 years now shows they have a better standard of living and earn more; the intervention has scaled to roughly 150 million people a year in India and many other countries.

    “Now we’ve tracked people over 20 years, and we’re seeing people have a better standard of living or earning more.”

  • Demand for deworming is lower than the scale of the problem warrants partly for behavioral reasons: worms build up gradually and chronically rather than producing the dramatic, salient suffering and relief of malaria, so individuals do not feel demand commensurate with the harm—compounded by political factors (worm-afflicted poor are less politically influential) and bureaucratic turf wars between Ministries of Health and Education.

    “With worms, it’s much more like a chronic thing... I don’t think, even at the individual level, do you have quite the demand that would be commensurate with the scale of the problem. That’s a behavioral economics explanation.”

  • A Kenya program covering seventh and eighth grade fees for poor students who did well in sixth grade not only let more kids advance but increased effort throughout the distribution—not just top students—and also appeared to motivate teachers to work harder, suggesting merit scholarships deserve scaling and further testing.

    “we also saw that kids were putting in more effort in response to this, and not just the kids at the top of the distribution, but throughout the distribution. It also looked like teachers were putting in more effort”

  • In developing countries, curricula tend to be set at a high level similar to developed countries and focused on top-of-distribution students, while disadvantages like high teacher absence, sick children, and poor home preparation cause kids to fall behind; remedial education or adaptive technology-aided systems that meet kids where they are can produce very large learning gains at low cost.

    “in developing countries, education system is focused on kids at the top of the distribution. What’s been found is, if you can set up... remedial education systems or some technology-aided systems that are adaptive, that go to where the kid is... that can have a very big impact at quite low cost.”

0.51

Experimental development economics creates direct entrepreneurial and institutional-management skills through real-world problem solving with actual organizations and feedback.

2 pointscentrality 4/5
  • The experimental method in contemporary development economics has a natural affinity with founding and operating organizations, because experiments are typically conducted on concrete problems with real institutions, forcing researchers to know the customer, work with NGOs and governments, raise money, and iterate—skills directly applicable to entrepreneurship.

    “There’s a lot of relationship between the experimental method and the things that are needed to help found organizations.”

  • Experimental development economics teaches skills directly transferable to managing institutions: isolating causal impact via experiments, knowing the customer by actually talking to kids, teachers, and principals, getting and incorporating critical feedback from seminars rather than self-deceiving that your way is the only way, and treating improvement as a constant iterative cycle.

    “There’s a lot of relationship between the experimental method and the things that are needed to help found organizations.”

0.44

Culture is remarkably persistent but does not preclude rapid development; poor countries are broadly catching up to rich ones despite long-standing cultural differences.

6 pointscentrality 2/5
  • While economic history shows cultural factors are remarkably long-lasting (e.g., whether societies plowed thousands of years ago affects them today), the rapid catch-up of low-income countries shows culture should not be read as precluding development or change—earlier simplistic claims like 'Confucian societies will never grow' or an inherent 'Hindu rate of growth' turned out to be wrong.

    “what we’ve also seen is low-income countries really catching up... That suggests that we shouldn’t take talking about culture as indicating that there’s no possibility for development or for change.”

  • Rewarding a classroom as a whole for doing well—rather than individuals—can plausibly create a class culture where students want each other to succeed, addressing the US problem where it can be uncool to do well in school; preliminary Kenya results on this approach are extremely promising.

    “they reward the classroom as a whole for doing well. You can imagine that can possibly create a culture in the class as a whole, where you want each other to do well. The initial results on this are actually extremely promising.”

  • Transient populations that turn over a lot may be associated with worse-quality food, possibly explaining why Harvard Square has poor food despite high purchasing power and education.

    “I think transient populations that turn over a lot may be associated with worse-quality food, but I’m not sure.”

  • Contrary to the 1990s finding of no unconditional convergence, the pattern changed after the 1990s: poor countries are now broadly catching up to rich countries—not just India and China—representing a tremendous lifting of people out of poverty even as within-country inequality has risen.

    “after the ’90s, this pattern changed. Now, poor countries are catching up to rich countries. This isn’t just a matter of India and China. This is a broad-based phenomenon.”

  • Kremer's mother wrote about literature on the Holocaust, which conveyed the value that when there are issues we can do something about, we must act and not wait—influencing his choice to work on international development where problems can be addressed at fairly low cost to ourselves.

    “when there are issues that we can do something about, we need to do something about them. We shouldn’t wait.”

  • Following Chad Jones's insight, while existing technologies help invent new ones, an improvement in existing technology does not help one-for-one in developing new technologies—it is getting harder over time to invent new technologies (seen in the rising team sizes needed for innovation), creating a race between rising total innovation resources and the increasing difficulty of the task.

    “In some sense, it’s getting harder and harder to invent new technologies... We see that in the increasing size of teams needed to develop new innovations.”

0.43

Advance market commitments work for innovations with known endpoints and measurable thresholds but cannot incentivize truly unexpected technological breakthroughs.

3 pointscentrality 3/5
  • Clean cookstoves are a good candidate for an advance market commitment because indoor air pollution from household cooking fires kills millions yearly and produces potent global-warming chemicals, and an AMC could combine a technological threshold with a consumer-willingness requirement—e.g., subsidizing $20-30 if consumers will pay $10—solving the prior problem that developed stoves didn't attract consumer demand.

    “a lot of people in developing countries cook with a small fire in their house, and that creates incredible indoor air pollution. That’s estimated to kill millions of people a year.”

  • Advance market commitments work for innovations you know you need ahead of time and where you can define a passing threshold (like a malaria vaccine via FDA trials), but they cannot incentivize innovations you never knew you needed until invented (like the Post-it Note).

    “We didn’t know we needed the Post-it Note until it was developed.”

  • There is a tendency for advance market commitments to target technologically close products (like pneumococcus) rather than technologically distant ones (like malaria), because politicians and bureaucrats want quick wins and firms push products where they have a technological advantage—even though, economically, technologically distant products offer greater advantages from AMCs.

    “The politicians want a quick win. The bureaucrats want a quick win, and their firms or other entities that are going to be pushing the things that they think they’ve got a technological advantage on — they’re technologically close.”

0.38

Development innovation operates like venture capital—most efforts fail but successful winners generate returns far exceeding total portfolio costs.

2 pointscentrality 3/5
  • Investing in development innovation works like venture capital—most innovations don't succeed but a few do, and the winners pay for the losers; an analysis of Development Innovation Ventures' early portfolio of 41 innovations yielded a superconservative 5:1 lower-bound rate of return, computed by setting the quantifiable benefits of just 4 of the 9-10 projects that reached over a million users against the cost of all 41 investments.

    “we had 41 innovations in our early portfolio. We couldn’t quantify the benefits for all of them, but of the 9 or 10 that reached over a million users, we could quantify the benefits for 4. If you take the benefits of those 4 and set them against the cost of our investments in all 41, the ratio is 5:1.”

  • Because the world economy is losing $500 billion every month due to COVID, it is worth investing heavily in developing a vaccine and getting many 'shots on goal' despite high failure risk; since private returns from building large factories before a vaccine is proven fall far below social returns, governments should use a mix of push and pull incentives—up-front financing to build factories ahead of time plus rewards for success—and invest in supply-chain capacity, not just final vaccine production.

    “the world economy is losing $500 billion every month due to COVID, and that means that it’s worth investing a lot in trying to develop a COVID vaccine, even though there’s a pretty high risk of failure, and it’s also worth getting lots of shots on goal.”