Fair500 → Analysis → The denominator problem

The widest pay ratios are mostly a story about the denominator

Among the 25 widest pay ratios in the S&P 500, CEO pay sits at the 72nd percentile. Median worker pay sits at the 10th. The extremes are made at the bottom of the fraction, not the top.

By the Fair500 editors · Published 18 July 2026 · All figures from SEC filings · How these numbers are built

The CEO-to-worker pay ratio is a fraction, and a fraction can be made large two ways. Public argument about executive pay almost always assumes the first: that a big ratio means a big pay package. Across the S&P 500, the most extreme ratios are more often produced by the second.

Consider two companies with almost the same ratio. Ross Stores reports 3,225:1. Its chief executive is paid about $32 million on a three-year average, a large sum but an unremarkable one for a company of its size, and well below what several hundred-to-one companies pay. What makes the ratio extreme is the other number: a median employee compensation of $10,059.

Now take Welltower, at 2,290:1. Its median employee earns $124,995, which is among the better-paid workforces in the index. The ratio is enormous because the chief executive's three-year average compensation is $286 million, inflated by a single equity grant valued at around $821 million.

These two companies sit next to each other in a ranking of pay ratios and have essentially nothing in common. One is a story about executive compensation. The other is a story about who Ross Stores employs.

Measuring which half does the work

We can put a number on this. Across the 492 companies with a disclosed ratio, the log of the pay ratio correlates with CEO pay at +0.73 and with median worker pay at −0.60. Taken across the whole index, then, executive pay is the stronger single driver, which is roughly what you would expect, and worth stating plainly before the caveat.

But the picture inverts at the extremes. Look only at the 25 widest ratios in the S&P 500 and rank each company against the full index:

The companies with the most extreme ratios do not, as a group, pay their chief executives extraordinarily. They pay them well, comfortably above the middle, but nowhere near the top of the index. What sets them apart is that their median employee is paid less than roughly nine out of ten companies in the S&P 500. The tail of the distribution is manufactured almost entirely in the denominator.

So why are those medians so low?

Three reasons, none of which is "this company suppresses American wages," and all of which are visible in the filings.

The median employee is often not in the United States

This is the dominant factor at the very bottom. The lowest median pay figures in the index belong overwhelmingly to companies whose workforce is concentrated in manufacturing outside the US:

Western Digital, Lumentum, Jabil, Flex, ON Semiconductor, Amphenol, Seagate: these are contract manufacturers and component makers whose headcount sits largely in Malaysia, Thailand, the Philippines, Mexico, China and Eastern Europe. Flex employs 150,000 people and the large majority are production staff outside the United States. A median of $12,939 is an accurate description of the midpoint of that workforce, and it is not a wage that any American employee of Flex is likely to be earning.

The pay-ratio rule requires exactly this. Companies must include their global workforce, with a narrow exemption allowing them to exclude up to 5% of non-US employees. So a US-listed company with a genuinely global manufacturing base is required to divide an American executive salary by a global median, and the resulting number spans labour markets with radically different costs of living. The arithmetic is correct. The comparison it implies is not one anybody would defend if stated aloud.

Part-time and seasonal staff pull the midpoint down

The second cluster is domestic and quite different: retailers, restaurant chains and hospitality companies. Starbucks at $17,279, McDonald's at $19,020, TJX at $14,994, Ulta Beauty at $11,883, Chipotle at $17,446, Royal Caribbean at $19,027.

These medians are low not primarily because hourly rates are low, but because the median employee does not work a full year at full hours. The rule takes a snapshot of everyone employed on a chosen date, part-time and seasonal staff included, and reports their actual compensation for the period. A barista working twenty hours a week for part of the year enters the calculation at what they were actually paid. Starbucks employs 381,000 people on that basis, and the median of that population is a part-time wage.

That is a real and meaningful fact about the business. It tells you the company is built on part-time labour, which is worth knowing. It is not the same claim as "the typical Starbucks worker earns $17,279 a year for full-time work," which is how a 3,669:1 ratio tends to get read.

A few companies have almost no rank-and-file workforce at all

The third case is rarer and stranger. ON Semiconductor reports a median of $14,060 across just 2,400 employees; Vici Properties, at the opposite extreme, reports a median of $468,119 across a workforce small enough that profit per employee reaches $95 million. Where headcount is tiny or unusually structured, the median employee is not a representative worker in any ordinary sense, and both the ratio and the profit-per-employee figure become close to meaningless.

Where the ratio does tell you something

None of this makes the disclosure worthless. It makes it a measure that has to be read in context, and there are two contexts where it is genuinely informative.

The first is within an industry. Comparing Ross Stores to TJX, or McDonald's to Yum! Brands, holds the workforce structure roughly constant, so a difference in ratio reflects something real about pay policy rather than about geography or scheduling. This is why Fair500 offers sector pages: a within-sector comparison is far more defensible than an index-wide league table, and the sector medians vary enormously, from 81:1 in Energy and 87:1 in Utilities to 455:1 in Consumer Discretionary.

The second is where the median is high and the ratio is still extreme. When a company pays its median employee $124,995 and still posts a four-figure ratio, the denominator explanation is unavailable. That is the case at Welltower, and at Intel, whose median employee earns $114,900 against a ratio of 809:1. Those numbers are about executive compensation, and there is no confound to hide behind.

Why we publish the number anyway

There is a reasonable argument that a statistic this easy to misread should not be given a league table at all. We publish it for two reasons.

The first is that the alternative, adjusting the medians for geography or hours, would require assumptions we cannot verify from filings. Companies do not disclose the country distribution of their median calculation, and any correction we invented would be less trustworthy than the disclosed number, while looking more authoritative.

The second is that the composition of a workforce is itself a business decision. A company that structures itself around part-time domestic labour, or around offshore manufacturing, has made a choice with real consequences for the people it employs. The ratio does not measure executive greed at those firms, but it is not measuring nothing either. It is measuring the distance between the top of the organisation and the middle of it, and that distance is partly created by the decision about who counts as an employee in the first place.

What we ask is only that the number be read for what it is. That is also why Fair500 scores companies on a second measure alongside the ratio: median worker pay set against the profit each employee generates. It asks whether workers capture a reasonable share of what the business earns, and it is not distorted by executive pay at all.

← More analysis · Explore the map