Wage Differentials, by the Numbers: Why Pay Varies So Dramatically Across Jobs

Start with two real numbers from the same government survey. In May 2023, fast food and counter workers earned a median annual wage of about $30,110, while the typical worker in legal occupations earned roughly $99,220 — and at the top of the distribution, occupations like physicians, surgeons, and chief executives carried median pay well above $200,000. These figures come from the Bureau of Labor Statistics Occupational Employment and Wage Statistics program, which surveys hundreds of thousands of establishments to price nearly every job in the economy. The median across all occupations was $48,060.

That is a spread of more than seven-to-one between the bottom and the middle-high, and far more at the extremes. Why? Not because surgeons work seven times harder than line cooks — anyone who has worked a dinner rush knows effort is not the variable. The answer is a small set of forces economists can name and, to a surprising degree, measure.

The headline number: a vast, structured spread

The first thing to understand is that the wage distribution is wide and structured — it is not random scatter. Pull a handful of occupations from the May 2023 OEWS national estimates and a pattern emerges:

Line these up and the ladder is obvious: pay rises with the training, licensing, and rarity required to do the job. The question is what specifically is being priced at each rung.

What the numbers behind the numbers say

Three forces account for most of the spread, and each leaves a measurable fingerprint in the data.

Skill and training. The biggest driver is human capital — the years of education and training a job requires. The BLS data on earnings by educational attainment shows a clean staircase: more schooling, higher median pay, lower unemployment. Surgeons sit at the top partly because the path to the job is a decade-plus of training that few complete. Fast food work sits near the bottom partly because it requires little formal training, which means the pool of people who can do it is enormous.

Scarcity. Training requirements matter largely because they restrict supply. Two jobs requiring similar effort can pay very differently if one has a thin labor pool and the other a deep one. Airline pilots earn premium wages not only because the job is demanding but because certification — thousands of flight hours, rigorous medical and testing standards — keeps the supply of qualified pilots tight relative to demand. Loosen the licensing and, all else equal, the wage would fall as the pool widened.

Value of output. A job attached to high-value output can support high pay. Software developers command six-figure medians in part because the products they build can generate enormous revenue per worker — the marginal revenue product of a developer at a profitable tech firm is high, so the firm can afford to bid for scarce talent. The same skill applied to lower-value output would pay less.

The honesty layer: compensating wage differentials

Here is where the data gets genuinely interesting, because it reveals a force most people overlook. Adam Smith proposed in 1776 that wages must rise to compensate workers for the unpleasant, dangerous, or inconvenient features of a job. Economists call this a compensating wage differential — extra pay that is the price of misery, risk, or inconvenience, holding skill constant. The Library of Economics and Liberty's discussion of how labor markets price job characteristics treats this as a foundational result: identical workers will demand higher pay to take on worse conditions.

The fingerprint shows up across the data. Within similar skill levels, jobs that are dangerous, dirty, isolated, or scheduled at brutal hours tend to pay more than comfortable jobs requiring the same training. Deep-sea fishing, logging, and offshore drilling — consistently among the most fatal occupations the BLS Census of Fatal Occupational Injuries tracks — pay premiums above what their formal skill requirements alone would predict. Night-shift and remote-location work carries a similar bump. Workers are, in effect, being paid to accept what others won't.

This also explains a puzzle in the opposite direction: pleasant, high-status, intrinsically rewarding jobs can pay less than their skill level suggests, because the agreeable conditions are part of the compensation. Museum curators, many academics, and nonprofit professionals often accept lower wages for work they find meaningful — a negative compensating differential. The job pays partly in satisfaction, so it pays less in cash.

What the data doesn't show

The compensating-differential story comes with a crucial caveat that the raw wage tables hide: the mechanism only works when workers have alternatives and information. A line cook in a town with one employer cannot extract a premium for a hot, hazardous kitchen, because there is nowhere else to go — the monopsony problem. And many of the most dangerous jobs in the world pay poorly, not well, because the people doing them lack options. Compensating differentials are a force that operates in competitive markets with mobile, informed workers; where those conditions fail, danger and difficulty go uncompensated. The theory describes a tendency, not a guarantee, and treating it as a guarantee leads to the false and cruel conclusion that anyone in a bad job is being fairly paid for it.

The OEWS medians also mask enormous within-occupation spread. "Software developers" earn a median around $130,000, but the 10th-percentile developer and the 90th-percentile developer can differ by well over $100,000 depending on firm, location, and specialization. The single median number compresses a distribution that, for the individual worker, is what actually matters — and part of that within-occupation spread reflects employer wage-setting power, which the Richmond Fed's research on measuring employers' market power finds is widespread, especially in low-wage work.

What it means for you

Read honestly, the wage data points to where real leverage lives — and where it doesn't. The spread is dominated by skill scarcity and the value of output, which means the durable way to move up the distribution is to acquire skills that are both genuinely hard to obtain and attached to high-value work. Hours and hustle barely move the needle, because the line cook working sixty hours is still priced against a deep, easily-replaced labor pool.

Compensating differentials offer a narrower, real opportunity: if you have alternatives and can tolerate conditions others avoid — night shifts, hazardous trades, remote postings — the market will pay you a premium for it, and that premium is yours to capture precisely because most people won't. But the same logic warns you off the trap: a hard or risky job is not automatically a well-paid one. Always check the actual number against the BLS occupational wage data before you assume danger pays. Sometimes it does. Sometimes it just costs.