The tenth-best violinist in the world is, by any honest measure, astonishingly good. Blindfold you, play her against the very best, and you’d struggle to tell them apart — a shade less warmth in a high passage, a hair of imprecision you’d need a trained ear to catch. She is not “a bit worse.” She is 99% as good. And she earns, maybe, 2% of what the best earns. A rounding error in talent becomes a chasm in reward.
That is the puzzle of this lesson. Everywhere the previous lessons pointed — cities, wealth, plays on a song — we found a lopsided power law. Now we ask why the market for talent itself is shaped that way: why the gap between #1 and #10 in ability is trivial while the gap in their paychecks is monstrous. The answer isn’t that #1 is secretly 50 times more talented. It’s that the market is built to magnify a tiny edge into a landslide. Economists call this the winner-take-all or superstar market, and it runs on two ingredients you can name.
Rosen’s superstars
In 1981 the economist Sherwin Rosen published “The Economics of Superstars” and put his finger on exactly what makes a market winner-take-all. Two conditions, both required.
- Scalability. The performer’s output can serve a vast audience at roughly zero extra cost per person. One recording, one film, one app, one book — pressed, streamed, or downloaded to millions with no meaningful ceiling. The best surgeon can only operate on one patient at a time; the best singer can sell to everyone at once. Technology is what removes the old capacity limit that used to let many performers each serve a local market.
- Imperfect substitutability. People prefer the slightly-best, and a lesser performer is not a good substitute. Ten mediocre singers do not add up to one great one — you’d rather hear the best once than the tenth-best ten times. Quality doesn’t average; it ranks.
Put those two together and the arithmetic gets brutal. If the best is even fractionally preferred, and serving everyone costs almost nothing, then why would anyone settle for #10 when #1 is one click away? The audience floods toward the top. The best captures a share of the market wildly out of proportion to her edge in ability. Rosen’s phrase for it: small differences in talent become large differences in reward.
The one-sentence version
A superstar market appears when output is scalable (one unit serves millions at ~zero marginal cost) and imperfectly substitutable (people want the best, and second-best won’t do). Then a razor-thin edge in quality captures a wildly disproportionate slice of the reward — the tenth-best is nearly as good and earns a pittance.
Before you read — take a guess
A concert pianist and a top heart surgeon are both, say, the very best in their country and roughly equal in raw skill. The pianist can earn tens of millions from recordings and streams; the surgeon, however brilliant, earns a comfortable-but-bounded salary. What best explains the gap?
The ingredients that turn 1% into 100:1
Rosen gave us the two preconditions. But several forces stack on top, each turning a razor-thin edge into a larger and larger gap. These are the amplifiers.
Scalability (the capacity limit dies)
Before recorded sound, the best singer alive could only fill one hall a night. There was room in the world for thousands of very good singers, each serving a town. Recording killed that ceiling. Now one voice fills every hall, every phone, every café, everywhere, forever. Technology removes the capacity constraint that used to protect the merely-excellent — and when the ceiling lifts, the audience that used to be split among a thousand locals collapses onto a handful of globals.
Network effects (the lead self-reinforces)
A product with network effects gets more valuable as more people use it — a phone network, a social app, a marketplace, a file format everyone can open. This is where winner-take-all fuses directly with cumulative advantage from lesson 2. The leader’s lead makes the leader more attractive, which grows the lead: preferential attachment, wearing a business suit. Once a platform is “the one everyone’s on,” being 5% better isn’t enough to pull people off it — so the near-tie hardens into a near-monopoly, and it self-locks.
Search and attention costs (converge on the known name)
Nobody has time to audition every option. Faced with a thousand plausible choices, people converge on the name they already know — the bestseller, the top result, the friend’s recommendation, the brand. That shortcut is rational (searching is costly) and it’s ferociously self-reinforcing: the known get known-er. Attention is the scarce resource, and it pools around whoever already has it.
Reduced friction and globalisation (one market, not many)
Cheap shipping, instant streaming, a global internet — each melts the walls that used to carve the world into many local markets, each with its own local champion. Melt the walls and those thousand local contests merge into one global contest with one global winner. Fewer walls, fewer winners, bigger prizes for the survivors.
Match each term to its precise definition.
Tournament markets: paid for your rank, not your work
In 1995 Robert Frank and Philip Cook wrote The Winner-Take-All Society and added a sharp second lens. In these markets, they argued, rewards depend on relative rank, not absolute performance. You are not paid for how good you are in some absolute sense — you are paid for where you finish. It’s a tournament: the winner takes the purse, and finishing a hair behind pays a small fraction of finishing a hair ahead, even though the two performances were nearly identical.
This “small rank difference, huge pay gap” pattern shows up far beyond entertainment:
- CEOs — the person who gets the top job earns many multiples of the runner-up who was, on the shortlist, basically tied with them.
- Athletes — the gold medalist and the fourth-place finisher can be separated by hundredths of a second and by millions in endorsements.
- Authors and apps — the bestseller and the platform “everyone” uses swallow the market; the near-identical alternative gets scraps.
- Academic prestige — citations, prizes, and faculty jobs concentrate on a few names (the Matthew effect from lesson 2, again), though the tenth-ranked researcher’s work is often excellent.
The deep point: in a tournament market, the reward gap tells you almost nothing about the ability gap. A landslide of pay can sit on top of a coin-flip of talent.
A friend is deciding between two side ventures and asks which will have superstar, winner-take-all dynamics. Venture A: a mobile game she'll publish once to a global app store. Venture B: a local dog-grooming business in her town. Which is the superstar market, and why?
The control case: why your plumber isn’t a superstar
The cleanest way to see superstar dynamics is to look at a trade where they don’t apply. Consider a local plumber, or a dentist, or a hair stylist.
Their output can’t scale: a plumber can only fix one leak at a time, bounded by hours in the day and roads they can drive. And their service is a fine substitute: the plumber three streets over is perfectly good — you don’t need the world’s best plumber, you need a competent one who can come Tuesday. Both of Rosen’s preconditions fail. So there’s no mechanism to funnel the whole market onto one super-plumber. Instead, thousands of plumbers each serve a local patch, and their incomes cluster in a comfortable bell curve — the best earns maybe two or three times the median, not ten thousand times.
That contrast is the whole lesson in one table:
| Scalable / superstar | Non-scalable / local | |
|---|---|---|
| Example | Recording artist, app developer, novelist, pro athlete | Plumber, dentist, hair stylist, local chef |
| Output serves… | Millions at once, ~zero marginal cost | One customer at a time, capped by hours |
| Good substitutes? | No — people want the best | Yes — a competent local one is fine |
| Market shape | One global winner, long tail of also-rans | Many local providers, each with a patch |
| Income spread | Power law — #1 earns 1000× the median | Bell curve — #1 earns ~2–3× the median |
| Reward vs. skill | Wildly disproportionate to ability gap | Roughly proportional to ability |
The same person, with the same 1% edge in skill, gets paid two conditions differently depending only on whether their output scales and whether substitutes are good enough. Change the technology — let the plumber somehow bottle and sell their expertise to millions — and the plumber’s income distribution would snap from bell curve to power law overnight.
Here’s the part that stings. Because the reward gap between #1 and #2 is enormous but the ability gap is essentially a coin flip, “just be the best” is brutal, near-useless advice. In a bell-curve trade, being 10% better reliably earns you ~10% more — effort maps to reward. In a winner-take-all market, being 1% better might earn you 100× more or nothing at all, and which one you get is decided by luck, timing, and the compounding loops from lesson 2 — who got the early playlist spot, the first review, the lucky break that network effects then locked in. So thousands of nearly-equal contenders pour their lives into chasing one slot, most of them lose, and the winner — genuinely excellent but no more excellent than the runner-up — pockets a fortune the market attributes to “talent.” The distribution is manufactured by the market’s structure, not handed down by merit.
The costly downside: arms races for the top slot
Winner-take-all markets don’t just distribute rewards unequally — they distort behavior, and expensively. Because the prize is enormous and rank is everything, hordes of nearly-equal contenders rationally over-invest in the fight for the top slot. This is positional competition (tie it to the arms-race idea): when only relative rank pays, everyone escalates effort and spending to out-rank everyone else — and because rank is zero-sum, most of that effort cancels out. A thousand musicians can’t all be #1 no matter how hard each grinds; a thousand startups chase one category-winning slot and most of the collective investment is, in aggregate, burned.
Two consequences fall out of this:
- Enormous inequality from small skill gaps. The reward distribution is a power law even though the underlying talent is nearly bell-curved. A near-tie at the top opens into a chasm of pay.
- Wasteful over-investment (arms races). Because “the winner takes the purse,” too many talented people crowd into the tournament, and society over-spends on the contest relative to what it produces — the classic too many contestants for one prize problem.
The engine that picks the winner
Which near-equal contender becomes the winner-take-all winner? Rarely the one who is provably best — the field is too close for that. Usually it’s whoever the cumulative-advantage loop from lesson 2 happened to favor: an early lead in visibility, plays, downloads, or citations that preferential attachment then compounds and network effects then lock in. Winner-take-all is why the prize is so lopsided; preferential attachment is often the engine that selects who wins it. The two lessons are two halves of one machine.
Superstars, tournaments, and the tail
What are the two conditions Rosen identified as necessary for a superstar (winner-take-all) market?
Check your answer to continue.
Recap
We answered why the market for talent is itself power-law shaped. Rosen’s superstars: when output is scalable (one unit serves millions at ~zero marginal cost) and imperfectly substitutable (people want the best, and second-best won’t do), a razor-thin edge in quality captures a wildly disproportionate share of reward. Four amplifiers stack on top — scalability killing the old capacity limit, network effects self-reinforcing the leader’s lead (cumulative advantage in a suit), search/attention costs converging on the known name, and globalisation merging many local markets into one global contest. Frank & Cook’s tournament markets sharpen the point: reward tracks relative rank, not absolute performance, so a coin-flip gap in ability sits under a landslide gap in pay — visible in CEOs, athletes, authors, apps, and academic prestige. The control case — the local plumber whose output can’t scale and whose substitutes are fine — stays reassuringly bell-curved. And the costs are real: enormous inequality from small skill gaps plus wasteful arms races as too many contenders over-invest chasing one slot that the preferential-attachment loop from lesson 2 usually decides.
Winner-take-all explains where the giant prize comes from. But it hides a stranger consequence: when a handful of outcomes are that gargantuan, the ordinary tools of reasoning — the average, the typical case, the sample so far — quietly stop working. Next up: lesson 5, “The Tyranny of the Tail” — why in a power-law world the average is meaningless, the sample mean never settles, and one event can outweigh all of recorded history.