Climbing the Ladder: What the Data Says About Social Mobility

Follows: Castes, Estates and Classes (Society 3) · Clase Social y Dinero (Español) · The Political Spectrum, Fairly Presented (Politics)

Picture two American babies. One is born in 1940, the other in 1984. By the time each turns thirty, what are the odds they earn more than their parents did at the same age, after adjusting for inflation?

For the 1940 baby, about 90%. For the 1984 baby, about a coin flip. That single comparison, from a 2017 paper by Raj Chetty and colleagues titled "The Fading American Dream", did more to change how people talk about getting ahead than a generation of speeches. It also hides almost everything interesting about the subject, which is what this piece is for.

Castes, Estates and Classes, the previous piece in this series, ended where modern societies began: legal ranks abolished, the promise of rising by talent, and Michael Young's warning about a meritocracy that believes it deserves its place. It closed with that 90-to-50 headline and a nod to the "Great Gatsby curve". Clase Social y Dinero, in the Spanish library, looked at class as something you feel. This piece opens up the data behind the headline. How much does the family you're born into decide where you end up? How does that differ between countries and neighbourhoods? And why do serious people, looking at the same numbers, still disagree about what they mean?

Part 1 — Two questions that sound like one

"Social mobility" is really two different questions, and most arguments about it go wrong by mixing them up.

Absolute mobility asks: are you better off than your parents were? It's about living standards. If the whole economy grows and the gains are shared, almost everyone can do better than the generation before, even if nobody changes places.

Relative mobility asks: did you end up in a different position from your parents? If your parents were in the bottom fifth of incomes, where did you land? This is about rank, and it's zero-sum by construction. For one person to climb from the 20th percentile to the 60th, someone else has to slip down. There are only a hundred percentiles to go around.

The OECD, in its 2018 report A Broken Social Elevator?, uses a neat image: mobility works like a set of escalators. Everybody can be carried upward together (absolute mobility), but some escalators move faster than others, so people's positions relative to each other shift over time (relative mobility).

Back to the two babies. Chetty's team (Hendren, Grusky, Hell, Manduca and Narang, writing in Science) found that absolute mobility in the US fell from about 90% for children born in 1940 to about 50% for those born in the 1980s. Comparing sons to fathers alone, it fell from 95% to 41%. The decline happened across the income distribution and was largest for the middle class.

The interesting part is the why. Two things could have caused it: slower economic growth, or growth whose gains went disproportionately to the top. The researchers ran two counterfactuals. If the US had grown as fast as it did in the 1940s and 1950s but shared the gains the way it does now, absolute mobility for the 1980s cohort would have reached only about 62%. If the economy had grown at its actual, slower rate but shared the gains the way it did for the 1940 cohort, absolute mobility would have reached about 80%. In other words, more than two-thirds of the fall came from how growth was divided, not from how much there was. The authors estimate that restoring 1940s-style mobility with today's distribution would require sustained real growth above 6% a year.

That finding gets cited by people across the political spectrum, but for different lessons. Some read it as an argument for redistribution or stronger wage bargaining. Others read it as an argument that growth itself is still the main lever, since no amount of reshuffling makes up for a stagnant economy, and point out that any income-based measure misses some gains, such as non-cash benefits and better products. Both readings take the numbers seriously. They disagree about what to do with them.

Part 2 — How sticky is the ladder? Measuring the link

Relative mobility needs a number for "how much of a parent's advantage passes to the child". Economists have three main ways of putting one on it.

The elasticity

The oldest is the intergenerational earnings elasticity, usually written as the Greek letter β (beta). It comes from a simple regression of children's log earnings on their parents'. Alan Krueger, chairing the White House Council of Economic Advisers, gave a clean example in 2012: with an elasticity of 0.4, if your parents earned 50% more than average, you would be expected to earn about 20% more than average in your own generation. Some of the advantage survives; most of it washes out. A β of 0 means your parents' income tells us nothing about yours. A β of 1 means the family's position is passed on intact.

Measuring β well turns out to be hard, and the difficulty is instructive. Any single year's income is noisy: a father might have had an unusually bad or good year. Noise in the parents' income makes the link to the children look weaker than it really is. Bhashkar Mazumder showed this with Social Security earnings records: earlier American studies, which typically averaged fathers' earnings over five years at most, had found elasticities around 0.4. Using many more years of fathers' earnings, Mazumder's estimate rose to around 0.6, implying much less mobility than previously thought. The general lesson: a short snapshot of income mistakes luck for status, and makes societies look more fluid than they are.

The rank correlation

Chetty's group popularised a second measure: compare ranks instead of incomes. Line up all parents from poorest to richest, line up all their grown children the same way, and ask how closely the two rankings match. Using tax records on more than 40 million American children, Chetty, Hendren, Kline and Saez found that a 10-percentile rise in parents' rank is associated, on average, with a 3.4-percentile rise in the child's rank. The rank measure has a practical virtue: it isn't affected by how far apart the rungs are. The elasticity mixes two things together, how often families change rungs and how far apart the rungs are. The rank measure isolates the first.

The transition table

The third way is the most intuitive: a table of where children from each fifth (or tenth) of the parental distribution end up. It shows something averages hide, which is that mobility can differ enormously at the top and bottom. Miles Corak compared the US and Canada this way. America's elasticity, about 0.5 in his comparable estimates, was roughly twice Canada's, yet in the broad middle both countries looked fairly mobile. The difference was at the edges. More than half of American sons raised by top-tenth fathers stayed in the top three tenths, and about half of those raised by bottom-tenth fathers rose no higher than the third tenth. Canada had less stickiness at both ends.

Krueger offered a memorable way to feel the size of these numbers. A reasonable summary of the US research, he said, was a parent-child income correlation of around 0.5, close to the correlation Francis Galton found between parents' and children's heights more than a century ago. So the chance of a child born in the bottom 10% reaching the top 10% is roughly the chance of a 5'6" father (168 cm) having a son who grows to over 6'1" (185 cm). It happens. Not often.

One caveat runs through this whole literature: the classic comparisons are of fathers and sons, because women's paid work changed so much over the twentieth century that mother-daughter comparisons are harder to interpret. Newer tax-record studies use household income and include daughters, but cross-country tables often still rest on fathers and sons.

Part 3 — The simulator: how long does advantage last?

Here is the elasticity put to work. If each generation keeps a share β of the previous generation's gap from the average (measured in proportional, log terms), then a family's disadvantage shrinks like compound interest in reverse: after one generation β of the gap is left, after two β × β, and so on. This is the logic behind the OECD's famous estimate of how many generations it takes a poor family to reach the average.

Play with it for a minute and three things stand out.

First, small differences in β compound. Using the OECD's own average inputs (an elasticity of 0.38, and a bottom-tenth family on a little under a third of mean income, a ratio of 1 to 3.5), the typical descendant reaches 62% of average in one generation, 83% in two, 93% in three and 97% in four. That's the source of the OECD's headline: in a typical member country it takes "around four to five generations" for children from the bottom tenth to reach mean earnings. Slide β to Denmark's 0.12 and the gap is nearly gone in two. Slide it to Colombia's 0.76 and you are counting in centuries.

Second, the averages hide the families. Switch on luck. With β around 0.4, parental income explains only a modest share of the spread in children's incomes (in this simple model, β squared, about 16%). Most of what happens to any one family is everything else: talent, health, choices, timing, the local job market, plain chance. Mobility statistics describe the odds, not anyone's fate.

Third, the start matters as well as the speed. In more unequal countries the bottom tenth starts further below the average, so even at the same β the climb takes longer. This double penalty is part of what the next section is about.

A dissent: maybe advantage lasts much longer

The simple model has a well-known weakness. It assumes that only your parents matter, so a family's advantage decays smoothly. The economic historian Gregory Clark has argued, in The Son Also Rises (2014) and papers with Neil Cummins, that this badly underestimates persistence. Tracking rare surnames among Oxford and Cambridge students in England from 1170 to 2012, they estimate an underlying status correlation of 0.70 to 0.90, "unchanged over centuries", and conclude that social mobility in England in 2012 was little greater than in preindustrial times. If that is right, the simulator's β should be set near the top of its range, and policy matters less than people hope.

Critics reply that surname averages measure something different. Florencia Torche and Alejandro Corvalan showed that group-level persistence can't be compared with individual persistence, because a surname group's average income blends individual and group factors that can't be separated without more information. On the other hand, studies that follow real families for longer do find more persistence than the simple model predicts. Using Swedish records linking three generations' lifetime earnings and four generations' education, Mats Lindahl, Mårten Palme and colleagues found that two-generation estimates "severely under-predict" persistence over three generations. That supports part of Clark's point, though not his near-total immobility. The honest summary: two-generation numbers probably understate how long advantage lasts, and nobody yet agrees by how much.

Part 4 — The Great Gatsby curve and its critics

On 12 January 2012, Alan Krueger gave a speech at the Center for American Progress in Washington. He showed a scatter plot: countries' income inequality in the mid-1980s on one axis, their intergenerational earnings elasticity on the other. The dots climbed from bottom left to top right. More unequal countries had stickier ladders. Krueger gave it a name: "I call this the 'Great Gatsby Curve'", after F. Scott Fitzgerald's novel about a self-made man who could never quite buy his way into old money.

The data underneath came from Miles Corak, who has said the curve had been drawn in various forms by several researchers before Krueger named it. Krueger used it to make a forecast. Because US inequality had risen sharply since the 1980s, he projected that the American elasticity would rise from 0.47 to 0.56 for the next generation, meaning the persistence of advantage would grow by about a quarter.

The pattern is real and has been reproduced. The OECD's 2018 report found that across generations, "earnings mobility prospects tend to be usually weaker in countries where income inequality is high", and found "no evidence that greater inequality would bring higher income mobility". What the curve means is where the argument starts.

Then came the test of time. In 2014, Chetty and colleagues used tax records to track Americans born from 1971 to 1993, the very cohorts who grew up as inequality rose. They found that rank-based mobility had stayed essentially unchanged. A child born into the bottom fifth in the 1990s had about the same chance of climbing, relative to others, as one born in the 1970s (for the youngest cohorts, measured by the chance of going to college). That looks like a point against the causal reading.

But the same paper added a sting. Because the rungs had moved further apart, the consequences of where you start had grown even if the odds of changing places hadn't. In the authors' words, the "birth lottery" now matters more. Defenders of the curve also point out that the children of the most unequal decades are still too young to measure fully; the forecast may not yet have had time to come true. This is a live debate, and both sides have serious economists in it.

Part 5 — Place: the neighbourhood as a ladder

If you want to see how much geography matters, look inside a single country. In "Where Is the Land of Opportunity?" (2014), Chetty, Hendren, Kline and Saez mapped upward mobility across the US. A child from the bottom fifth had a 4.4% chance of reaching the top fifth in Charlotte, North Carolina, and a 12.9% chance in San Jose, California: almost three times the odds in the same country, under the same federal laws.

High-mobility areas shared five traits: less residential segregation, less income inequality, better primary schools, more social capital and more stable families. The authors were careful to say this was descriptive. It could be that places shape people, or that different kinds of families live in different places.

Is it the place, or the people?

To separate the two, Chetty and Nathaniel Hendren studied more than seven million families who moved. The trick is to compare siblings who moved at different ages. If places matter, the younger sibling, who spends more of childhood in the new area, should look more like the area's native children. That's what they found: children's outcomes converged towards those of the new area at about 4% per year of childhood spent there. Neighbourhoods mattered in proportion to exposure.

The strongest evidence comes from a real experiment. Between 1994 and 1998, the US Moving to Opportunity programme randomly offered about 4,600 low-income families in high-poverty public housing, in five cities, a housing voucher that could be used only in a low-poverty area. For years the results looked disappointing. Early studies found adults' health improved, but not their earnings.

Then Chetty, Hendren and Lawrence Katz looked at the children once they were grown. Those who moved before age 13 earned about 31% more in their mid-twenties than the control group, were more likely to go to college, less likely to become single parents, and lived in better neighbourhoods as adults themselves. They estimated the gain at about $302,000 over a lifetime for a child who moved at around eight. Children who moved as teenagers gained nothing, and possibly lost, perhaps because of the disruption. The lesson for research is as important as the lesson for policy: the effect was there all along, in a group nobody had yet been able to measure.

Who your friends are

Why do places matter? One answer came in 2022 from a study in Nature using 21 billion Facebook friendships. Its key measure, economic connectedness, is the share of a lower-income person's friends who are higher-income. It turned out to be among the strongest predictors of upward mobility yet found. If children from low-income families grew up in counties as connected as the average higher-income child's, their adult incomes would be about 20% higher, the authors estimated. Other kinds of social capital, such as how tight-knit a community is, were not strongly linked to mobility. Whom you know across class lines seems to matter more than how many you know.

Some barriers turn out to be surprisingly practical. In Seattle, an experiment called Creating Moves to Opportunity gave voucher families information, short-term financial help, a navigator for the housing search and introductions to landlords. The share moving to high-mobility areas rose from 15% in the control group to 53%. The researchers argued that the main obstacles were in the search process itself, not a preference for staying put.

Race shows how place and history interact. In a large study of US tax data, Chetty, Hendren, Maggie Jones and Sonya Porter found that among children from families with the same income, Black boys grew up to earn less than white boys in 99% of Census tracts. Parental marriage, education and wealth explained little of the gap. The few areas where the gap was small had low poverty, less racial bias among white residents and more Black fathers present in the community, and fewer than 5% of Black children grew up in them.

Most of this neighbourhood evidence is American, because the US tax data made it possible. Whether the same patterns hold elsewhere, and how strongly, is still being studied.

Part 6 — Education: the great equaliser, with an asterisk

Education is the ladder most societies officially rely on, and it does a great deal of lifting. It also carries more family advantage than its reputation suggests.

Start with the floors. Across OECD countries, four in ten people whose parents had little education end up with only lower-secondary education themselves, and only about twelve in a hundred earn a university degree, compared with roughly two-thirds of children of highly educated parents. Adults whose parents had higher socio-economic status score almost 20% higher in numeracy, the equivalent of more than three extra years of schooling.

In poorer countries the stakes are starker. The World Bank's 2018 report Fair Progress?, built on a database covering 96% of the world's population, found that developing economies make up 46 of the 50 worst performers for educational mobility from the bottom to the top. In some low-income or fragile African countries, only 12% of young adults born in the 1980s had more education than their parents. The report also found reasons for hope: rising school enrolment in Africa and South Asia over the last two decades, and girls overtaking boys in tertiary education in high-income countries, with the developing world moving the same way.

The view from the top colleges

American tax-record research gives a precise picture of elite universities. A 2017 "mobility report card" study found that children from the top 1% of families were 77 times more likely to attend an Ivy League college than children from the bottom fifth. Yet once there, low-income and high-income students went on to earn similar amounts. Elite colleges work as ladders; very few low-income students get to stand on them. The highest rates of bottom-to-top mobility were at some mid-tier public universities, which combine decent outcomes with real access.

A 2023 study by Chetty, David Deming and John Friedman went further. At "Ivy-Plus" colleges, children from the top 1% were more than twice as likely to attend as middle-class children with the same SAT or ACT scores. About two-thirds of that gap came from higher admission rates, and the advantage traced to three things: preferences for alumni children, weight on non-academic ratings, and athletic recruitment. None of the three predicted better outcomes after college, while test scores did. And attending such a college instead of an average state flagship mattered: it raised the chance of reaching the top 1% of earners by about half, nearly doubled the chance of attending an elite graduate school, and almost tripled the chance of working at a prestigious firm.

The asterisk

There are two limits on education as an equaliser. The first is that expanding education is mostly absolute mobility. When everyone's education rises, children can out-learn their parents without anyone changing rank, and credentials begin to work like a queue: what counts is having more than the next applicant. (Piece 6 in this series returns to this idea of "positional goods".)

The second is the Solon–Corak mechanism from Part 4. When the pay gap between graduates and non-graduates is wide, families with money invest more heavily in their children's schooling, from tutoring to the right neighbourhood. Education can carry advantage forward as easily as it can break it. That's why the OECD's recommendations put heavy weight on things that reach children early and broadly: high-quality early childhood education and care, preventing school drop-out, and targeting spending on disadvantaged pupils rather than simply spending more overall.

Part 7 — Income is a flow, wealth is a reservoir

Almost all of the numbers so far are about income: what people earn in a year. But the family advantage that people feel most sharply is often wealth: the house, the savings, the parent who can lend the deposit or cover a gap year. Income is a flow; wealth is a reservoir.

The World Inequality Lab, which runs the World Inequality Database, published its World Inequality Report 2026 in 2025. Its global estimates for 2025:

The report finds the same pattern in every region: "wealth is much more concentrated at the top than income". At the very top it has grown. The share of global wealth held by the richest 0.001% of adults, fewer than 60,000 people, rose from about 3.8% in 1995 to about 6.1% in 2025. (These estimates combine tax records, surveys and national accounts, and their methods are argued over; in the US, for example, economists Gerald Auten and David Splinter dispute the size of the rise in top income shares estimated by Thomas Piketty, Emmanuel Saez and Gabriel Zucman, with unreported income the biggest point of disagreement.)

Why does wealth matter so much for mobility? Because it is insurance and launch fuel at once. A family with savings can absorb a job loss without a crisis, move to the high-opportunity neighbourhood from Part 5, pay for an unpaid internship, or let a child take a risk on a business. Mazumder found that the earnings elasticity was higher for families with low net worth, which is what you would expect if poorer families can't borrow to invest in their children. The OECD found that having grown up in a family with little or no wealth is one of the two main predictors of poor health in adulthood, alongside parents' poor health.

Wealth also makes positions sticky within a lifetime. Over a four-year period in OECD countries, about 60% of people in the bottom fifth of incomes stay there, while at the top 70% remain. And the middle is more exposed than it feels: about one in seven middle-income households slides into the bottom fifth over four years, a risk the OECD links to events such as unemployment or divorce.

Part 8 — Why? The structural and the individual case

Everyone in this debate agrees that family background matters. The disagreement is about how it matters, and so about what, if anything, should change. Here are the two broad explanations, each at its strongest.

Each side has answers to the other. Structuralists note that the "success sequence" describes the people who managed to follow it, and that finding stable full-time work, or a partner who has it, is itself much easier in some places than others. Individualists note that a voucher helped young children but not teenagers, and that no policy has made Denmark-style mobility appear in a large, diverse country. Even the genetics evidence cuts both ways: Belsky's team also found that people with higher scores tended to come from better-off families, and that a mother's genes predicted her child's attainment beyond the child's own, which suggests genetic and environmental advantage travel together.

What people believe about mobility is tangled up with politics, too. Alberto Alesina, Stefanie Stantcheva and Edoardo Teso surveyed people in France, Italy, Sweden, the UK and the US, and found Americans more optimistic about mobility than Europeans. When the researchers showed people pessimistic information about mobility, left-leaning respondents became more supportive of redistribution, especially policies aimed at equal opportunity. Right-leaning respondents didn't change their views, which the authors suggest may reflect a belief that government is the problem rather than the solution. Both groups were reacting to the same facts through different values, which is exactly the pattern described in The Political Spectrum, Fairly Presented.

Underneath sits a question that data can't settle. If a society removed every unfair barrier and children still tended to land near their parents because talents and temperaments run in families, would that be a fair society? Some say yes: equal rules are what fairness requires. Others, following John Rawls (see The Veil of Ignorance in Philosophy), say that nobody deserves the talents they were born with either, so a fair society arranges its rewards to benefit the least advantaged. Others again think the goal should be a decent floor and real absolute mobility, with rank mattering less. The numbers in this piece can tell you how sticky the ladder is. They can't tell you how sticky it ought to be.