The plainest test of the American Dream is whether children end up better off than their parents. Economists measured it directly: the share of children who, by age thirty, earned more than their own parents had at the same age. For children born in 1940 it was 92%, a near-universal promise. For children born in 1984, the last cohort with data, it was 50%, a coin flip. This is a decades-long structural change, driven far more by how the economy's growth is shared than by how fast it grows, and it belongs to no single president.
Share of children who grew up to earn more than their parents (at about age 30), by the year the child was born.
Why it fell, and what most Americans now expect
The economists behind this measure tested two explanations. Slower economic growth than the mid-century boom explains part of it, but the bigger driver is how unequally that growth is now shared. In their counterfactuals, restoring 1940s-style growth rates (with today's distribution) lifts mobility only to about 62%; keeping today's economy but sharing it as broadly as in 1940 lifts it to about 80%, reversing roughly 70% of the decline. The fall also hit the middle class hardest: for children of middle-income parents, the odds of out-earning them dropped from 93% (born 1940) to 45% (born 1980). Americans have absorbed the message, in a 2024 Pew survey, about three-quarters said today's children will grow up to be financially worse off than their parents.
What this measures, and what it does not. This is absolute mobility, whether children out-earn their own parents in real terms, not equality of opportunity or relative rank; it depends mechanically on overall growth and its distribution. The series is by birth cohort, not calendar year, and it ends with children born in 1984 (measured around 2014); no verified update exists for later cohorts, so this is a historical trajectory, not a live year-by-year series. The decline spans births from 1940 to 1984, incomes measured from the 1970s to the 2010s, across many administrations and Congresses, and is attributable to none of them individually. The method links parent and child tax and Census records with statistical assumptions for the earliest cohorts; the endpoints are robust, and the intermediate points here trace the published curve.