The number arrives in a proxy statement, or a headline drawn from one, and it is built to stop you: the chief executive of a large company was paid two hundred and eighty-one times what the typical worker in that company’s industry earned. Not twice, not ten times. Read it against your own pay and the arithmetic turns personal and a little insulting, because the ratio invites a comparison of persons: is one human being worth two hundred and eighty-one of another. From there it is a short step to one of two slogans, and most people arrive at one or the other before the sentence is finished. The first says the package is theft, extracted by people who set their own pay. The second says it is a market price, what a rare talent commands, and that resenting it is resenting how value works. This chapter is about why both slogans are answers to a question the number has not actually posed yet.
The place to start is with the number itself, because “two hundred and eighty-one times” is a choice among several rather than one fact, and a different and equally official way of counting the same executive’s pay gives a different figure that has, in some years, moved the opposite way. Get the measure straight first, then the climb, and the two competing explanations for it come into focus as what they are: a real disagreement, with one particular piece of evidence sitting on the seam between them.
Two ways to count a pay package
An executive’s pay is mostly not salary. It is stock: shares awarded outright and options to buy shares later at today’s price, which together dwarf the cash. And stock can be counted at two different moments, which is where the trouble begins. One way, call it the granted measure, values the shares and options at the moment the board hands them over, using a standard formula to price an option nobody has yet exercised. It answers the question, what did the board decide to pay this year. The other way, the realized measure, counts what the executive actually collected in a given year: options cashed in, restricted shares that vested and became sellable. It answers a different question, what did this executive actually take home, and because today’s cash-outs come from grants made years earlier, it carries the whole run-up of the share price since. The two are not variants of one number. They are two different measurements of two different things, and they diverge.
How far they diverge, and in which direction, is the part that undoes the slogan. In most recent years the realized figure runs higher than the granted one, because a long bull market keeps turning old grants into large cash-outs. But at the peak of the dot-com boom in 2000 the order reversed: boards were granting enormous option packages, so the granted ratio reached 393 to one while the realized ratio, reflecting older and smaller grants, sat at 380. The gap between the two measures does not merely differ in size from year to year; it changes sign. The headline “two hundred and eighty-one to one” is specifically the realized figure for 2024, and that year’s value is itself a projection, estimated from partial data before the year was out. Quote it, and you have quoted one measure, in one year, still provisional. The granted figure for the same year is 213 to one. Neither is wrong. But a story that splices the high years of one measure onto the low years of the other, to make the steepest possible climb, is telling you about its own choices as much as about executive pay.
With the measure pinned, the shape of the climb is clear enough. On the realized measure the ratio was around 20 to one in 1965 and roughly 30 to one as late as 1978; it reached the low hundreds in the 1990s and has swung between about 170 and 410 to one since. Over the stretch from 1978 to 2024 executive pay by this measure grew nearly twelvefold, while the pay of the typical private-sector worker grew by roughly a quarter. That is the raw material both slogans are working with. The question is what drove the climb, and here the composition of the package is the first clue: the rise is overwhelmingly a rise in the value of stock. That fact does not by itself decide between the two stories, but it tells you what any account of the climb has to explain, which is an ownership stake that grew with the companies and the market underneath it, rather than a salary that ballooned.
The scale story
One account begins with a plain observation: the biggest companies got very much bigger, and pay tracks size. A chief executive at a firm worth two hundred billion dollars is making decisions over a far larger base than one at a firm worth two billion, so a small difference in the quality of those decisions is worth a great deal more in dollars at the larger firm. If the market for executives is even roughly competitive, it will bid a modest edge in talent up to a large premium in pay, simply because the premium is spread over so much more capital. On this reading the climb in pay is a shadow of the climb in firm size, and the thing to explain is growth, not greed.
The economists Xavier Gabaix and Augustin Landier built this argument out into numbers, and their paper draws a distinction the slogan flattens. Look across firms at a single moment, they find, and pay rises with a firm’s own size only gently: the elasticity is about a third, so a firm twice as large pays its chief executive roughly a quarter more, not twice as much. That is the cross-sectional pattern, and on its own it would make a three-hundred-fold pay gap hard to credit as a reward for three-hundred-fold personal productivity. The steeper relationship lives in time. When all the large firms grow together, the average size of the biggest firms, the reference level the market prices against, pulls pay up roughly one-for-one. By that time-series logic Gabaix and Landier attribute the sixfold rise in US executive pay from 1980 to 2003 to the sixfold rise in the market value of large firms over the same period. The two elasticities, the gentle one across firms and the steep one over time, are different quantities that answer different questions, and the paper’s point is the contrast between them.
The capture story, and the test between them
The other account starts from a different fact about the room where pay is set. In a public company with no controlling owner, the people who approve the chief executive’s pay are the directors, and the directors are not the disinterested agents of far-off shareholders the market story assumes. The legal scholars Lucian Bebchuk and Jesse Fried made this the center of their work: the board, they argue, cannot be expected to bargain with the executive at arm’s length, because directors owe their seats in part to the executive’s goodwill, hold only nominal stakes in the firm, and lack the independent information a hard negotiation would need. The forces that are supposed to discipline pay from outside, the threat of takeover, the pressure of capital and labour markets, are real but loose, leaving what they call managerial power: the executive’s ability to influence the size and shape of their own package. On this reading a chunk of the pay is not the price of talent at all; it is rent, and packages are structured to blur how much is being paid and how weakly it is tied to performance.
Stated at that level, the two stories can each explain the climb, which is why a slogan can be built on either. What separates them is a prediction they do not share, and Bertrand and Mullainathan found the place to test it. Suppose an executive’s pay is a market price for skill and judgment. Then pay should respond to the things the executive actually controls, and not to pure windfalls, gains that arrive for reasons outside anyone’s decisions. Oil companies offer a natural case, because a large part of an oil firm’s fortunes rides on the world price of crude, which no single executive moves. The finding, drawn on the shareholder-wealth measure the bottom panel of the figure plots, is that a windfall gain is rewarded about as much as an ordinary one: a rise in shareholder wealth lifts pay by nearly the same fraction whether or not the executive caused it, and boards do not strip the windfall out before paying on it. Pay for luck, in their phrase, is real: the statistical test rejects the idea that boards filter luck away completely.
The finding is two-sided, and both sides have to be kept in view. The luck estimate carries a wide error band, so the claim is that a lucky gain is paid about like a general one, and not that it is paid more; the same paper’s other measure, which reads as if luck is rewarded several times over, rests on an estimate too imprecise to bear that weight, and its authors decline to draw it. The other half of the result is in the paper’s subtitle: pay for luck is weaker where governance is tighter, at firms with a large outside shareholder watching or a smaller, more independent board. That is the seam between the two stories drawn as a measurement. A world of pure talent pricing would not pay executives for the weather in the oil market; a world of pure capture would pay for it everywhere and equally. What the data show is neither: pay does respond to luck, which the talent story alone does not predict, and it responds less where the board is in a stronger position to say no, which is what the managerial-power story predicts. That is as far as the test itself speaks, and the contest stays open.
What the number does and does not settle
Put the pieces back together and the executive’s package resolves into parts that answer to different logics. Most of the dollar figure is stock, so most of the climb is the stock market and the growth of the companies it prices. A firm many times larger genuinely makes each executive decision worth more, so scale explains why the level rose as firms grew, though the gentle cross-sectional relationship means scale does not license reading a three-hundred-fold pay ratio as three-hundred-fold personal worth. And sitting on top of the scale is a residue that behaves the way the capture account predicts and the talent account does not: pay that moves with luck, and moves less where someone is positioned to check it. None of this makes the number theft, and none of it makes it a clean market price. It makes it a mix, whose proportions the evidence can constrain but not resolve to a point.
CEOs are paid 300 times more because they’re worth 300 times more.
Oversimplified Moderate confidence
The claim takes a real relationship and dresses it as precise individual desert. It has a true kernel, which is why the ruling is not the reverse of the claim: pay does track firm size, bigger firms genuinely put more capital under each executive decision, and a market that prices talent against the size of the largest firms will lift pay as those firms grow, which is much of what the long climb records. But three things break the “worth three hundred times more” step. First, the “three hundred times” is one measure of the gap, the realized one, which swings with the stock market and in some years runs below the granted measure rather than above it, so the exact multiple is a choice as much as a fact. Second, the relationship the scale story actually establishes is gentle across firms, an elasticity near a third, so a much larger firm pays only somewhat more per executive; the steep climb over time comes from all large firms growing together, which explains a rising level of pay without certifying any individual as hundreds of times more productive than a typical worker. Third, and most directly against the word “worth,” executive pay demonstrably rises on luck the executive did not create, an oil-price windfall being the standard example, and it rises less where a board is well placed to object. Pay that moves with the weather is not paying for worth. So the claim is not simply false, and it is not the opposite of the truth; it takes a real link between pay and firm scale and inflates it into a statement about personal worth that the measure’s own instability, the gentleness of the cross-sectional relationship, and the evidence on luck will not support.
Sources
- The two measures and the climb: Economic Policy Institute, CEO Pay report (2025), Table 1 — the realized CEO-to-worker ratio at the top 350 US firms was 280.7 to one in 2024 (a projected figure) against a granted ratio of 212.6, and the two measures cross, with granted above realized at the 2000–02 peak (393.1 against 379.6 in 2000) and realized above granted in most recent years. Executive pay grew about 1,094 percent from 1978 to 2024 on the realized measure against about 26 percent for the typical worker, and the package is roughly 79 percent stock.
- The scale story: X. Gabaix and A. Landier, “Why Has CEO Pay Increased So Much?,” Quarterly Journal of Economics 123(1), 2008 — the cross-sectional elasticity of pay to a firm’s own size is about 0.37 (Roberts’ law exponent near 1/3), while the time-series elasticity of aggregate pay to the size of the largest firms is near one; the authors attribute the roughly sixfold rise in pay from 1980 to 2003 to the sixfold rise in large-firm market value. The two elasticities are different constructs.
- The capture story: L. Bebchuk and J. Fried, “Executive Compensation as an Agency Problem,” Journal of Economic Perspectives 17(3), 2003 (NBER Working Paper 9813) — boards of widely held firms cannot be assumed to bargain at arm’s length, so managerial power lets executives influence their own pay, and part of the package is rent rather than the price of talent.
- The discriminating evidence: M. Bertrand and S. Mullainathan, “Are CEOs Rewarded for Luck? The Ones Without Principals Are,” Quarterly Journal of Economics 116(3), 2001, Table 1 — among oil-company executives, luck (shareholder-wealth gains instrumented by the world oil price) is rewarded with an elasticity of about 0.35, close to the 0.38 for general performance; complete filtering of luck is statistically rejected, and pay for luck is weaker where governance is tighter. The accounting-return specification is too imprecise to read as “paid more for luck,” and the chapter does not read it that way.
- Confidence is moderate under the rubric, which scores the weaker of evidence directness and construct match. Evidence directness is reasonable: the pay-size and pay-for-luck relationships are estimated from large panels of firms. The binding, weaker leg is construct match. The claim asserts individual “worth,” a marginal product that is not directly observable, while the evidence measures elasticities of pay to firm size and to luck, which bear on the claim without being a measure of worth, and the luck finding cuts directly against reading pay as worth. Because the claim’s central construct is unobservable and the nearest evidence partly contradicts it, the construct mismatch is what caps the confidence.
Why the far end of the scale looks like this
This chapter closes a run of four questions about who is paid what, and it sits at the opposite end of the scale from the ones before it. The volume opened by arguing that pay tracks how replaceable a worker is and how credibly they can walk away, and the flashpoints since have mostly been about workers whose leverage is thin: the shift-worker facing a wage floor, the two groups paid differently, the newcomer feared as cheap labour. The chief executive is the same lens turned all the way over. Here the worker is treated as close to irreplaceable, the bargaining runs almost entirely one way, and the board that is supposed to sit across the table has, in Bebchuk and Fried’s account, reasons to sit alongside instead. That the pay is mostly stock is a fact about leverage too: it ties the executive’s fortune to the owners’, which is the whole point of the arrangement, and it is also what lets the package ride the market up to numbers that strain any story about a single person’s worth. The machinery that makes a company able to grow to that size and to be owned by strangers who never meet, and that hands its directors their awkward double role, is itself the subject of the chapter on how the corporation was engineered.
Whether all of this adds up to the rich getting rich off the rest, and how such concentrations have risen and broken across history, are larger questions this book takes up elsewhere: in Frameworks, on whether one person’s gain is another’s loss, and in the Pathologies volume, on when inequality stops being tolerated and cracks a regime. The next chapter stays inside the boundary of work but moves to its edge, to the labour that does not show up in any pay figure at all because no money changes hands for it, and to why the economy counts a nanny but not a mother.