CEO pay at 290 times a typical worker: a ratio is not a verdict

A director on your board remarks that in America, CEOs earn 300 times what their workers do. The board is about to vote on the managing director's pay. How accurate is that figure, and does it say anything about your own director's pay? This article separates two questions: what the number measures, and whether the arrangement that sets pay holds up.

What is the 300-to-1 ratio, and who measured it?

The EPI, a Washington policy institute, reports 290, not 300. In 2023 the CEOs of the 350 largest US public firms by revenue earned 290 times a typical worker on realised pay, and 192 times on granted pay. In 1965 the realised figure was 21 (Economic Policy Institute, 2024, p. 1, 2). If a slide says '300:1', ask which year and which measure.

How it is built matters. The EPI computes a ratio for each firm and then averages them. The denominator is the average compensation (wages and benefits) of production and non-supervisory workers in the firm's main industry, because pay data for each firm's own workers are not available (Economic Policy Institute, 2024, p. 6). The figure swings with the stock market, since CEO pay is mostly stock-based (Economic Policy Institute, 2024, p. 1, 4): the realised ratio was 360 in 2022 and 405 in 2021, which the report calls a historic high (Economic Policy Institute, 2024, p. 5, 9).

The EPI does not merely count. Its report says CEO pay has risen because executives use their power to set their own pay, not because their skills or contribution have grown (Economic Policy Institute, 2024, p. 1). That is one side's case. To read a ratio like this, see five questions before you trust a chart.

What does Rawls say about pay gaps, and why does the question move to institutions?

Rawls does not demand equal pay. Justice as fairness has equal basic liberties, then fair equality of opportunity and the difference principle: inequality is acceptable only if it most benefits the least advantaged. People choose these principles behind a veil of ignorance, not knowing their own position (Rawls, 2001). So high pay is not unjust in itself, and no single number is enough to judge.

In the chapter-one excerpt I read, Rawls aims his principles at "the main political and social institutions and the way they fit together as one scheme of cooperation" of a society (§1.3) (Rawls, 2001). That is a whole society's institutions, not one company's; carrying the test into a boardroom is an analogy made by the library's framework, and I flag it as one.

The last check is the hard one, because it needs two real structures to compare. The library names two traps: treating the 'average worker' as the 'least advantaged', and claiming that pay exceeds the minimum effective compensation without data.

How well do the four arguments for high pay hold up?

Only the incentive argument holds, conditionally; the library rates the other three weak, or weak to moderate.

Talent. The library's objection is that talent is natural luck and its market value depends on supply and demand. Gabaix and Landier sit on the other side. In their model the spread in talent at the top is tiny (swapping CEO number 1 for CEO number 250 lowers a firm's value by only 0.016 per cent), yet firm size magnifies it into a pay gap of more than 500 per cent, and they write that this small spread still 'justifies' large pay differences (Gabaix and Landier, 2008). My reading is that this is justification within the model's market equilibrium, not a claim that the pay is deserved. Scanlon defines a desert-based justification as one that holds a treatment justified simply by what the person is like and has done, independent of good effects, of an institution calling for it, or of whether the person could have avoided it; the abstract says the paper explores economic reward but states no conclusion (Scanlon, 2013).

Effort. The library lists three objections (hard to measure, partly a product of talent, lower risk than many jobs); no source I read goes deeper.

Contribution. Dietsch argues that where parties are equally interdependent in the division of labour, the co-operative surplus should be shared equally (Dietsch, 2008). The abstract does not mention CEOs; applying it here is the library's step.

Incentive. Boatright writes that high pay can be justified in a free-market system if certain conditions hold, including that bargaining between CEO and shareholders involves no fraud or breach of fiduciary duty (Boatright, 2010). Edmans and Gabaix survey optimal contracting theories and argue that many features of pay packages that look inefficient can be consistent with efficiency (Edmans and Gabaix, 2009). Moriarty puts the limit on the CEO: their fiduciary duties cap the pay they can accept, because excess leaves the beneficiaries, such as shareholders, worse off (Moriarty, 2009). The EPI treats the CEO-to-top-0.1-per-cent ratio (9.4 in 2022, against an average of 2.6 in 1965–1978) as a sign of rent, and cites Kaplan as the most prominent voice for the market-for-talent reading (Economic Policy Institute, 2024, p. 10). Check 4 depends on it, because whether any pay above the minimum can be redirected to the least advantaged is an empirical question.

What does a small family company's board do with the test?

Hypothetical example. Ví Dụ Packaging ('ví dụ' means 'example'; the firm, the people and every figure are invented by me). It makes cardboard boxes and has 85 employees. Mr T. owns 60 per cent and is chairman and managing director; his wife holds 15 per cent; his sister, Ms M., holds 25 per cent and works elsewhere. The board has three members: Mr T., his wife and a cousin who is chief accountant. Mr T. is paid VND 1.8 billion a year (before dividends) and the median employee VND 120 million, a ratio of 15 to 1; it cannot be set directly against the EPI figure because the denominator differs. The 40 lowest-paid warehouse and line staff earn VND 84 million each. A recruiter quotes VND 1.0 billion for a hired general manager. Mr T. has guaranteed a VND 5 billion bank loan with his own assets.

Check 1, who sets pay? Mr T. sits on the board approving his own pay, with his wife and a cousin who keeps his books. Ms M., who bears a quarter of the cost and receives none of the pay, is absent. Check 2, political liberty, does not apply to an 85-person firm.

Check 3, fair equality of opportunity. Assume all six management posts are held by relatives and no outsider has been promoted into management in ten years. That is a red flag; see also succession in family firms.

Check 4, a feasible comparison. Structure A: keep VND 1.8 billion. Structure B: pay Mr T. VND 1.0 billion, the price of a hired manager, and give VND 0.8 billion to the 40 lowest-paid staff, VND 20 million each, about 24 per cent more. Total spend is unchanged. Mr T. already keeps 60 per cent of profit, so the amount above the quote buys no extra incentive. On these numbers B is feasible and A fails the test, but the evidence is a single quote.

The outcome is a decision, not a verdict. The board could keep VND 1.8 billion but let Ms M. see three quotes and vote, or cut to VND 1.0 billion and raise the lowest-paid group's pay. The test makes whoever sets pay put numbers on the table before someone who does not depend on him.

Can an institution cap the gap by rule?

The 2023 annual report of the Spanish cooperative group Mondragon states a wage scale of 1 to 6 between minimum and maximum salaries (MONDRAGON Corporation, 2023, p. 8). It is self-reported and measures lowest against highest, not CEO against the average worker, so it cannot be compared with the EPI figure. The two pages I read give no actual gaps; the library's Mondragon case records other sources with other figures. See the Mondragon case.

Going deeper: what did I read, and what did I not?

EPI, 'CEO pay declined in 2023' (Bivens, Gould and Kandra, 19 September 2024; the library lists the title as 'CEO pay in 2023'): I read the 18-page PDF. Page numbers are the printed ones; the cover and overview are unnumbered.

Rawls (2001): I read the chapter-one excerpt (§1 and the start of §2) published by The New York Times, and the book's contents list from a bookseller's page. The two principles and the veil of ignorance come from the library's framework text; I did not check them against the book.

Boatright, Moriarty, Scanlon, Dietsch, and Edmans and Gabaix (2009): abstracts only. Gabaix and Landier: the abstract and introduction of the July 2006 working paper, not the 2008 journal version, so no page numbers. Mondragon 2023: pages 7 and 8. I relay the library table's objections without checking the originals.