Loan applications, visa forms, health questionnaires — every one of them has a line labelled "occupation". You fill in two words and turn the page. That line is a ruler sociology spent seventy-odd years calibrating, and what it measures is not how much you earn.
This page takes apart three things: how that ruler was built, why the gap it measures is not the same thing as what happens to one person over a lifetime, and where the sieve that actually decides whether a position gets passed down is turning.
Start with the oldest ruler. In 1947 the National Opinion Research Center in the United States, under Cecil North and Paul Hatt, ran a survey: read ninety occupations out to ordinary people and ask them to rank them. That is the origin of the occupational prestige scale. In 1964 Robert Hodge, Paul Siegel and Peter Rossi repeated it as closely as they could, and the two sets of results correlated at 0.99 — sixteen years, almost no movement. In 1977 Donald Treiman put prestige surveys from about sixty countries side by side and found the orderings broadly similar there too.
Be clear about the strength of that evidence: it is a cross-sectional survey correlation. It establishes one thing only — that the consensus about how to rank occupations is unusually stable — and not that the occupations carry any intrinsic rank.
The second ruler is completely different. The classification built by the British sociologist John Goldthorpe and his collaborators (usually called the EGP schema, after the initials of Erikson, Goldthorpe and Portocarero) does not sort by income at all. It sorts by employment relation. The core distinction is between two very different contracts: a service relationship, in which the employer is buying long-term loyalty and judgement and pays with a salary, a promotion ladder and a cushion when things go wrong; and a labour contract, settled by the hour, the piece or the month, with nothing owed either way afterwards.
The power of that distinction: two people take home the same amount this month, one of them by hourly settlement, the other on a salary with a ladder attached — and one serious illness or one industry downturn sends their lives in different directions. Class is not an income bracket. It is the logic by which your money arrives.
The third ruler gets mentioned least: income is a flow, wealth is a stock. The two are correlated, but the shapes of their distributions are not alike — national statistics agree everywhere that wealth is far more concentrated than income. The reason is not mysterious: a stock has positive feedback, since money makes money, and a flow does not.
Three rulers, three different things, and none of them the "right" one. The Weber and rationalisation page described his insistence that class, status and power are three axes that cannot be reduced to one another — these three rulers are roughly the workable version of those axes. So when two reports on inequality contradict each other, first check whether they are measuring the same thing.
The relative at the dinner table who says "our kid is bound to do well" has a counterpart number in sociology: the intergenerational income elasticity. It asks how much higher a child's income is, on average, when a parent's income is 10% higher.
US estimates in the 1980s sat near 0.2, which made the society look rather mobile. In 1992 Gary Solon, using the long-running Michigan survey that has revisited the same families every year since 1968, pointed out a hard flaw in the older calculation: a single year of income contains a great deal of transitory noise — the year the bonus was unusually good, the year someone was ill. Use a noisy ruler as your explanatory variable and the correlation you measure is pulled systematically toward zero. Switching to a five-year average moved the estimate from roughly 0.3 to roughly 0.4.
That is the methodological lesson most worth keeping from this page: part of the conclusion "this society is quite mobile" was manufactured by measurement error. Noise does not push a result off in a random direction. It has a fixed one — it flattens the relationship.
One number for a whole country is still far too coarse. In 2014 Raj Chetty, Nathaniel Hendren, Patrick Kline and Emmanuel Saez used tens of millions of de-identified tax records to compute this at the level of commuting zones: the probability that a child born into the bottom fifth of the income distribution reaches the top fifth as an adult is 4.4% in Charlotte and 12.9% in San Jose — one country, one legal system, a gap of nearly three times. Evidence: whole-population administrative data, extremely solid as description.
Is the regional difference caused by the places themselves? That needs a different kind of evidence. In 2018 Chetty and Hendren built a quasi-experiment out of families who moved, treating a child's age at the move as close to random: the earlier a child moved to a high-mobility area, the higher their adult income, with the gain accumulating roughly linearly in years of residence. Stronger than a bare regional correlation, still not a randomised experiment.
Finally, the pair of numbers most often mixed up. In Science in 2017 Chetty and colleagues measured something else: the share of children earning more at thirty than their parents did at the same age fell from about 90% for those born in 1940 to about 50% for those born in 1980. Relative mobility — who swaps places with whom — has barely moved for decades, while absolute mobility, whether you live better than your parents, has halved. In the everyday argument about whether there is still any hope, each side is usually holding one of those numbers, and both numbers are true.
You send in a CV assuming that what follows is a contest of ability. Lauren Rivera wanted to see what actually happens at that step, so she sat inside the hiring process at several top investment banks, consultancies and law firms, from the campus presentation through to the final deliberation table, and added a large number of anonymised in-depth interviews. In 2015 she published that fieldwork as Pedigree.
The first thing she saw: the comparison does not begin at the CV. These firms present and collect applications at a very short list of "core schools", and everyone outside that list never enters the same pool at all. The second thing happens at the interview table: a large share of the assessment is so-called cultural fit, and the way interviewers judge fit is largely by looking for people whose leisure pursuits resemble their own — rowing, skiing, a particular style of travel.
Those pursuits are not ability signals. They are things that take years and a great deal of money to cultivate. Sociology calls this cultural capital: the competences picked up at home as a by-product, with no tuition fee but not without cost — which topics count as tasteful, how to talk to authority, which kind of joke you can hear. Its one decisive property is that in a screening room, it looks like ability. Evidence: ethnography of one industry and a handful of firms; the direction is credible, the proportions are not general.
Further upstream there is another gate. In 2023 Chetty, David Deming and John Friedman linked de-identified admissions records to tax data for the most selective private colleges in the United States. Holding standardised test scores equal, applicants from families in the top 1% of the income distribution were roughly twice as likely to attend those colleges as equally-scoring applicants from middle-class families; among those, legacy applicants were roughly five times as likely as comparable peers. Three channels: alumni preference, recruited athletics, and the non-academic ratings supplied by private high schools. Evidence: large-scale administrative data with a same-score comparison, very strong descriptively, though the authors themselves note it is not a randomised experiment.
How to interpret these facts splits sociology into two opposed readings, and both deserve their strongest version. In 1945 Kingsley Davis and Wilbert Moore argued that inequality is an incentive device: important and difficult positions must carry enough reward, or nobody will put in the years of training first. In 1953 Melvin Tumin's two hardest counters were these. First, "importance" cannot be measured independently of reward — saying a position is important because it pays well and pays well because it is important is a circle. Second, opportunity itself is unequally distributed, so the same device that allocates talent also systematically wastes the talent that never gets a chance to try.
That argument ran from the 1950s into the 1980s and never converged. Every number on this page is usable by both sides — the disagreement is not about the facts, but about which face of them counts as the main effect.