
Alex Tabarrok - Prizes, Prices, and Public Goods
What this covers
Alex Tabarrok, an economist at George Mason University, examines three interconnected mechanisms for solving coordination failures in innovation, public goods provision, and education. Hosted by Dwarkesh Patel, the conversation spans prizes and advance market commitments as alternatives to conventional grants, the Baumol effect as an explanation for rising healthcare and education costs, and dominant assurance contracts as a solution to free-rider problems. Tabarrok argues these market designs work better than government-grant approaches at generating speed and breadth of innovation, especially in vaccine development where multiple approaches and funding streams matter more than targeted bets on a single candidate.
The discussion covers territory from pandemic economics to labor market displacement. On the Baumol effect, Tabarrok contends that rising prices in stagnant sectors like healthcare and education reflect not dysfunction but rather the opportunity cost of skilled labor drawn away by productivity gains elsewhere—a theory that reframes slowing price inflation as a symptom of slower overall progress rather than improvement. The conversation turns to online education, where Georgia Tech's scaled computer science degree serves as both proof of concept and warning: educational content can be de-bundled, with lectures moving online at a fraction of traditional cost while credentialing and coaching remain geographically distributed. On the darker side, Tabarrok discusses how automation and trade can displace workers whose skills don't complement technology, and how declining research productivity—requiring roughly 18 times more engineers to sustain semiconductor advances than decades past—hints at fundamental constraints on innovation. The episode also explores dominant assurance contracts, which solve contribution problems by making participation a dominant strategy through refund bonuses, and sketches how blockchain-based online worlds might enable thousands of governance experiments previously impossible.
Tabarrok argues that well-designed economic mechanisms—prizes and advance market commitments for innovation, dominant assurance contracts for public goods, and online education tied to progressive sectors—can solve coordination and free-rider problems that conventional grant-and-government approaches handle poorly, and that the Baumol effect explains rising prices in stagnant sectors as a side effect of progress rather than dysfunction.
- Prizes and advance market commitments give firms stronger incentives to move quickly on vaccines than ordinary markets do, especially during a pandemic where speed has enormous payoffs.
- The Baumol effect shows rising prices in education and healthcare result from productivity gains elsewhere (opportunity cost of skilled labor), not from regulation or unions.
- Dominant assurance contracts solve the public good contribution problem by making contribution a dominant strategy via a refund bonus, enabling experimentation with new governance mechanisms.
This asset isn't compiled yet
You're seeing its claims, ranked. Compile it to build the argument threads, weight them, and check each claim against your library — the full view.
Because rising stagnant-sector prices are caused by progressive-sector productivity growth, prices in stagnant sectors rise less quickly when the progressive sector slows—so the slowing of healthcare price inflation today (after exploding ~10% per year in the 1950s-60s when productivity was high) is actually a bad sign reflecting the great stagnation, meaning lower price inflation can be a symptom of slower progress.
“when did healthcare prices really go up the big increase in healthcare prices when it was in the 60s 50s and 60s... that's when productivity was going up so now productivity is slowing down and we're seeing prices are not going up so quickly”
The dominant assurance contract solves the public goods free-rider problem (which Paul Samuelson deemed unsolvable) by adding one twist to a Kickstarter-style assurance contract: if the funding threshold is not reached, everyone who agreed to pay receives a refund bonus, making contribution a dominant strategy—you benefit whether the project succeeds (you get the public good) or fails (you get the bonus). Lab experiments with Cason and Zubrickas show it roughly doubles the number of successful projects.
“if uh not if you don't reach the threshold then everybody who agreed to pay gets a refund bonus okay so... either way you're actually better off contributing than not contributing... it's now a dominant strategy”
The Baumol effect explains rising prices in stagnant sectors (education, healthcare, repairs) as a single unified theory: because progressive sectors gain productivity rapidly, the opportunity cost of resources (especially skilled labor) used in stagnant sectors rises, so stagnant-sector prices climb—not because of unions, regulation, or sector-specific dysfunction. A string quartet takes the same four people 40 minutes in 1826 and 2020, but what you give up to hire them has risen enormously.
“the slower productivity sector doesn't have to be negative productivity just it's growing more slowly... if you can get a lot more goods in the progressive sector well that means if you want more of the stagnant sector you have to give up a lot”
Georgia Tech's online Master's in Computer Science demonstrates the future: ~7000 online students treated identically to residential students (same grading, professors don't know who is online), graduating about 7% of all US CS masters at less than a quarter of the residential price—serving as a trailblazer or warning sign for higher education.
“georgia tech has the largest computer science program in mass masters program in the world has something like out of 7000 online students... it graduates i think like seven percent of all the computer science masters uh in in the world... a quarter of the price”
The Baumol effect implies we won't run out of work because stagnant service sectors grow as a share of the economy, but it does not mean automation fears are unfounded: less-skilled labor can be displaced by automation and trade (their wages have not risen), and this will worsen with remote work and automation, so those who cannot raise their skill level face real trouble—analogous to horses, which did not all find alternative employment after the automobile.
“i think about horses right uh you know the horse when we introduced the automobile it's not like horses all found alternative employment uh they found the glue factory”
The best career advice for a 20-year-old is to gain skills complementary to technology rather than competing against it ('racing with the machine')—data science, machine learning, causal inference, electrical engineering, computer science, or economics—because the volume of data is growing faster than almost anything and extracting meaning from it is difficult and valuable; design and marketing also remain valuable when combined with technology (as Apple shows), so even artistic impulses pay off if fused with a technological field rather than pursued as, say, pure poetry.
“you don't want to be competing against technology you want to be racing with the machine uh to use eric brynjelson's... term... get educated in a sector like electrical engineering computer science or economics... which is complementary to technology data um data science”
The Baumol effect contains both a substitution effect and an income effect, which jointly explain why spending on education and healthcare rises even as their prices rise—a pattern that contradicts most rival theories (which require two separate explanations) because in Baumol's framework the productivity growth in progressive sectors simultaneously raises stagnant-sector prices and makes society richer, pushing out the production possibility frontier.
“in baumel the you're pushing out the production possibility frontier as it is becoming um more curved at the same time so the price is going up but you're also becoming richer so bommel explains both of these things very very nicely”
The dominant assurance contract solves only the contribution problem for public goods of known optimal scale (a bridge, a lighthouse) where the right size is easy to determine; it does not solve the preference-revelation problem for goods like defense where the optimal quantity ($100B vs $500B) is genuinely uncertain.
“it solves the contribution problem there are other public goods where you're not quite sure how much of it should you actually produce like defense... this does not solve that but it does solve the contribution problem”
Democracy is a poor preference-aggregation mechanism but a good mechanism for limiting government and excluding dictatorial rule—democratic governments do not starve their own citizens, a low bar that many governments fail to meet—so its value lies more in governance constraint than in accurately collecting and aggregating preferences.
“it's not a great preference aggregation mechanism it is actually a pretty good way of limiting government... democratic governments do not starve their own citizens that's a very low bar but it's a low bar which many governments fail to meet”
Education will be de-bundled: the lecture/information-transmission component will go online with far fewer teachers reaching many more students, while the coaching and tutoring component will be delivered through hierarchical structures (a star teacher with tiers of TAs) and through a growing 'life coach' industry—analogous to how athletes all have coaches, suggesting everyone could benefit from coaching to map skills and provide encouragement.
“education will be de-bundled right... the kind of traditional lecture part of it sage on the stage... that part will go online and a many a much smaller number of teachers will teach many more people but the coaching part of it”
Online worlds (especially blockchain projects, each with its own governance system) enable an unprecedented explosion of low-cost experimentation in collective-governance mechanisms because they satisfy Tiebout conditions—people can move between 'worlds' at low cost—allowing thousands of governance experiments versus the handful (monarchy, oligarchy, democracy, anarchism, futarchy) seen in all prior human history.
“the nice thing about online is that it satisfies the taboo conditions... you can move from world to world to world to world very easily at low cost online so it's much less costly to set up a new type of government”
Online education will make teaching a winner-take-all field like professional sports: a small number of teachers will be paid like basketball stars while most teachers are driven out of business, because one person teaching online can reach hundreds of thousands rather than 30 students; this is good because cracking the scaling problem yields massive returns, tying education to a progressive (technology) sector instead of a labor-intensive one.
“we're gonna pay our teachers more like basketball stars and and baseball stars but there's just gonna be a lot fewer of them right because when you teach online... i can teach 300 students and i do um so hundreds of hundreds of thousands of students”
MOOCs will not replace universities, but universities will go online because students still need the credentialing 'stamp' (signaling) that MOOCs can't yet provide; high-reputation universities like George Mason or Georgia Tech can expand their markets globally, while many universities without strong reputations will fall by the wayside, especially under pandemic pressure.
“the moocs as it were you know coursera and things like that they're not gonna replace universities but universities are gonna go online right so you still need that stamp”
Globalization should boost innovation by increasing both the supply of global talent (more scientists and engineers as China and India grow richer) and the demand for R&D (a larger global market where, e.g., everyone wants a cancer cure), yet this coincided with the great stagnation because of a countervailing 'ideas are getting harder to find' dynamic—possibly a fundamental, regulatory, or merely temporary lull between technological quantum leaps (electricity, internal combustion, computers, perhaps biology or quantum computing next).
“globalization has been good for um research and development uh we would be much worse off if we didn't have that... there is this countervailing wind which we now know from the bloom and van renin work”
Diversifying across a portfolio of vaccine types and funding mechanisms is essential because any vaccine can fail at any stage—exemplified by the AstraZeneca/Oxford trial being paused over a negative effect despite 2 billion pre-ordered doses—so policy should take many 'shots on goal' since the winning candidate is unknown.
“you really want to have a diverse set of uh vaccines in your uh portfolio... take a lot of shots on goal because we don't know which shot on goal is actually going to be the one which uh succeeds”
A prize is superior to targeted grants for vaccine innovation because government and even VCs are bad at picking winners, the most successful vaccine type is unknown (live attenuated, killed virus, mRNA, DNA, or therapeutics), and a prize opens the field to unconventional 'crazy' innovators who wouldn't pass through NIH committees—analogous to the Longitude Prize won by a clockmaker rather than Newton.
“when you have a prize you really are opening up the field to these to the crazy ones right uh to the ones who wouldn't necessarily get through the nih uh committees right and that's of course the classic uh longitude uh story”
The US government has lost state capacity—it has become too slow, complacent, and bound by legalism and bureaucracy to do what it is supposed to do when needed—exemplified by the CDC botching the first COVID test and the FDA blocking private companies from using their own tests; this failure spans Trump, Congress (which produced no testing or vaccine plan), and the broader system, and is the central concern of 'state capacity libertarianism': wanting a small government that can nonetheless act effectively when required.
“the cdc their entire reason for existing is to stop a pandemic and yet they completely failed they botched the first test okay and then the fda came in and said oh private companies you cannot use your testing”
Howard Hughes Medical Institute grants, which give researchers money with minimal strings ('come back in five years and tell us what you did'), produce more highly cited research and more patents than conventional grants even after controlling for researcher quality—evidence that funding with fewer strings attached is more successful.
“those uh grants are much more effective uh the researchers who get them uh that research tends to be much higher cited more patents... even when you set up some fine controls it looks like giving money with less strings attached actually is more successful”
Much of the great stagnation is explained by the economy shifting toward service/stagnant sectors that grow in productivity the least, so buying more slow-growing goods naturally lowers the aggregate growth rate—possibly an 'optimal stagnation' that is a natural consequence of getting richer rather than a pure pathology.
“as you shift to the service sector you see a decline in your growth rate simply because you're buying more of the things which are growing more slowly”
Helland and Tabarrok refine Baumol's labor-focused theory by specifying it is skilled labor in particular whose price has risen, because skilled labor became far more valuable in Silicon Valley; the opportunity cost of a PhD teaching 30 students rose because that person could instead earn far more at firms like Uber or Amazon, even though their classroom output is unchanged.
“we just give it a slight twist in our our book we say look uh it's skilled labor uh it's skilled labor which in particular has uh gone up in price because... a skilled laborer is much more valuable in silicon valley right than they used to be”
Even progressive sectors now require far more input per unit of output: per the Bloom, Van Reenen, and Cowen-cited research, about 18 times as many semiconductor engineers are needed to sustain Moore's Law doubling as in the past, and pharmaceutical R&D budgets per new drug and life-expectancy gain have risen sharply—not merely due to FDA regulation but because the low-hanging fruit has been picked, with research productivity reportedly halving roughly every 13 years.
“we need 18 times as many semiconductor engineers just to uh get the same... moore's law doubling of uh transistors every two years”
International competition (with China and India) is a promising way to restore American effectiveness and innovation, just as Sputnik-era competition with the Soviet Union drove the US to devote up to 12% of its budget to NASA; the US has become a 'warfare-welfare state' that should instead become an 'innovation state,' shifting federal spending from military and welfare toward R&D, where federal investment has shrunk as a share of the budget.
“what we have today is a warfare welfare state okay we do two things we invest a lot in the military and we invest a lot in various forms of welfare... what i would like to see us be is an innovation state”
Tying educational content to a progressive technology sector creates automatic compounding improvements: MRU captioned its videos in English, YouTube automatically translates captions into hundreds of languages, and every improvement in AI (e.g., DeepMind) automatically improves the product's captions without any additional work by the creators.
“every improvement in deep mind right every improvement in artificial intelligence automatically turns into an improvement in our product because periodically you know the captions they get better without us having to do any work”
Vaccine manufacturers have weaker incentives than is socially optimal to ramp up production quickly, because most vaccines fail and firms won't build factories until a vaccine is proven safe, effective, and approved; therefore society should add incentives via prizes, advance market commitments, or paying for at-risk manufacturing capacity to speed up production.
“a vaccine manufacturer they're not going to be willing to build a factory to ramp up a factory to get the doses flowing right until the vaccine has been proven safe and effective and it's going to be approved”
The US education system is irrationally configured: high school graduation rates are low compared to other countries (around 75-85% for men) yet 60-70% of high school graduates go to college (vs ~40% in Germany, which has ~97% high school completion); more people should finish high school and become highly skilled, but not necessarily via college, especially given many enter low-payoff fields like journalism and psychology while computer science enrollment has been flat for 30 years with growth coming only from foreign students.
“the high school graduation rate is actually quite low compared to other countries... but then of those who graduate high school we send a huge proportion to college way more than in europe”
Artificial tutors are already as good as real tutors in randomized controlled trials, and AI homework-assessment systems can analyze hundreds of thousands of student responses to pinpoint exactly which concept a student is failing to grasp and direct them to that precise piece of knowledge, allowing students to advance much faster than with generic lesson repetition.
“artificial tutors are already as good in randomized controlled trials as real tutors okay... machine learning i could see the machine will be able to see what concept you're not grasping”
A prize can be announced and committed to before funding is gathered or a committee is formed, making it faster to deploy than grant processes; under existing US code every agency can establish up to a $50 million prize and these can be combined, so multiple agencies could create a $100-200 million prize immediately, yet none have done so—illustrating government complacency and slowness in the pandemic.
“there is uh in the code uh i think from obama's time every agency has the right to set up a 50 million dollar prize... they can be accumulated so you could have you know two or three agencies or four agencies could right now set up a 100 million 200 million dollar prize”
Only individual researchers who get less-structured funding tend to do more cited, higher-patent work; combined with fast COVID grants, this suggests speed and autonomy in funding research outperform slow committee processes—Cowen distributed over $20 million in COVID grants faster than the NIH gave out a single grant.
“tyler cowan has given out more than 20 million dollars in covid grants and given them out incredibly quickly i think tyler gave out 20 million dollars faster than the nih had given out a single grant”
Online learning has structural advantages over classroom instruction because students learn at different paces and can pause, rewind, and vary speed (like listening to podcasts at 1.25-1.5x), giving them control that the fixed time-and-place classroom lacks—which is why many students report preferring online courses.
“with online something you didn't quite hear you didn't understand you just pause it... think about podcasts... at 1.5 or 1.25 speed... so this means that online students can go at their own pace”
Small differences in a cluster of technologies coming together can produce huge effects, as shown by the Apple Newton versus the iPhone: technologically very similar, but the Newton failed because it wasn't fast enough or connected enough, while the iPhone succeeded—suggesting online education may now be crossing a similar threshold where animations, music, AI, and speed converge to make it boom despite past failed predictions about radio, film, and TV transforming education.
“it's really just a small step from the newton to the iphone and yet the iphone was huge... technologically there's very little difference between the newton and the iphone the noon was a failure the iphone was a huge success”
VCs have a comparative advantage over government at picking winners because they have skin in the game, but they are still not great at it—learning enough about a complex field like vaccines to distinguish scams from real prospects takes significant time, which is why market-based funding can be too slow in a fast-moving crisis.
“the truth actually is that the vcs aren't that great at it either right uh they have a comparative advantage they have an advantage over the government... but for a vc to learn enough about vaccines... this takes time”
Prizes are particularly valuable when experts have failed and a problem requires a broad, out-of-the-box view, even though experts are usually the best people to consult; this is a situational rather than universal superiority of prizes over grants.
“prizes are particularly good when sort of the experts have kind of failed uh usually the experts are right... but then sometimes you have a problem where the experts have run against the wall and you need to take a very broad view and you need to get the crazies in”
US infrastructure projects (subways, tunnels) are far more expensive than in the rest of the world due to unions, legalism, historical-preservation committees, and accumulated procedural requirements; cost overruns are a general project-management problem that the dominant assurance contract does not address.
“there's a problem especially in the united states with why uh you know subways and construction and uh tunnels why they're so expensive here compared to the rest of the world... part of that problem is unions... but there's a variety of reasons”
Market-designed mechanisms for producing public goods (dominant assurance contracts, and the Weyl-Buterin quadratic-funding schemes) offer a 'progressive' libertarian path: rather than convincing a majority to adopt one's ideas, one can build alternative institutions that produce more public goods—especially local public goods—proving by demonstration that more public goods are desirable.
“creating a market mechanism it sounds contradictory but designing a market to produce public goods um could change you know everything... instead of saying you know the government's doing it wrong all the time um let's create an alternative institution which can do things better”
Robin Hanson's futarchy—governance by prediction/futures markets—is a genuinely novel fifth type of government, rare in being entirely original given that human history has produced essentially only monarchy/dictatorship, oligarchy, democracy, and Rothbardian/Friedman anarchism; futarchy deserves real-world experimentation.
“robin has his a new one which is incredibly rare right so that makes five which is futarki it's an entirely original uh idea for governance by futures markets”
Despite having more teachers per capita, math scores have not increased, because the student population has changed: far more and more varied students now attend college than the self-sufficient elite of the past, so the same teaching resources spread across a wider-ability population that needs more hand-holding yield no measurable gains in standardized outcomes.
“many many more people are going to college than ever did before and that just means like that the standards are going down and the students... just need a lot more hand-holding today”
For the COVID vaccine, paying upfront for a large fraction of manufacturers' factory-building or repurposing costs is cheaper in context than a fixed-price advance market commitment (the tool used for the pneumococcus vaccine), because the enormous trillion-dollar costs of the pandemic justify advancing on all fronts and tolerating waste from funding vaccines that ultimately fail.
“we're more pushing towards paying upfront for a large fraction of the manufacturer's costs for producing a factory or for repurposing capacity and it turns out it's cheaper in in our context to do that in this way”
Prizes may have declined relative to grants in the 20th century partly because dispensing grant money is desirable bureaucratic power—committees fund people who then become committee members—whereas a well-designed prize with strict criteria gives less discretionary power to the prize-givers; but this remains a genuine puzzle the speaker is unsure about.
“giving out money is kind of a powerful uh job... a prize if it's set up correctly... that kind of gives less power to the prize givers and maybe that's part of it but i'm not really sure actually it's a puzzle”
A change of administration risks worsening complacency: when an opposition party wins, the energy for fundamental reform dissipates even though underlying problems persist—just as anti-war protests vanished when Obama replaced Bush while the wars continued—so a Biden victory could take the wind out of demands to fix the government's legalism, bureaucracy, and slowness without solving them.
“under bush there was a big anti-war effort... and obama came into power and that all went away and the war kept going... if biden is elected that's not going to solve the problem of complacency”
Per Ronald Coase, an economist could justify his entire lifetime wages by stopping a single bad policy idea; economists' general, widely applicable analytical skills give them credibility in policy circles even when their specific proposals are rarely adopted, which Tabarrok finds comforting against the frustration of being ignored.
“coast said that uh that a an economist could earn his entire lifetime wages just by stopping one bad idea okay so i guess i feel i've done that”
Healthy international competition—'in the sense of the Olympics' rather than warfare—could reignite American passion for innovation and leadership in freedom, liberty, and racial and sexual equality; the US should welcome talent (e.g., people from Hong Kong, the Uyghurs) and be a 'shiny beacon on the hill,' since interests across nations largely align and cooperation is preferable to literal conflict.
“competition in the sense of the olympics a good kind of competition if we can keep it... that kind of competition i think would do a lot of good uh in reigniting america”
Tabarrok and Cowen aim to teach more people economics than anyone in the history of the world; via online scaling they have surpassed many historical economists and may overtake Mankiw, illustrating the unprecedented reach a single teacher can now achieve through online education.
“between tyler and i our goal is to teach more people economics than anyone has ever thought in the entire history of the world... we're catching up to mancu okay so i think we might overtake mancu”