Documenting decline in U.S. economic mobility.

Economists and other social scientists have long studied intergenerational income mobility, but consistent data linking adult incomes of children and their parents at similar ages over many generations have been unavailable, which thwarted attempts to study long-term trends. Chetty et al.’s study in this issue of Science (1) is therefore a tour de force for producing historically comparable estimates of absolute income mobility—the fraction of individuals in a birth cohort who earn, at age 30, more than their parents did at roughly the same age—over the post–World War II period. Their striking conclusion is that there has been a large decline in the rate of upward mobility across successive U.S. birth cohorts, from 92% of children born in 1940 earning more than their parents to only half of children born in 1984. Although Chetty et al. find that the slowdown in Gross Domestic Product growth has played a role, they conclude that faster economic growth is insufficient to restore mobility to its immediate postwar level in light of increased income inequality—a critical insight for policy and research. Chetty et al. combine parent-child linked income information derived from Internal Revenue Service administrative tax data for recent cohorts with cross-sectional income data on representative samples of parents and children from Census Bureau household surveys for earlier cohorts, adjusting for consumer-price inflation. Their key methodological innovation is to combine information on the marginal income distributions of children and parents with plausible, empirically based assumptions about the stability of the joint distribution of parents’ and children’s income ranks (the copula) to generate estimates of intergenerational mobility even in years when longitudinal linked parent-child income data are unavailable. Their data and approach provide the most compelling evidence to date that U.S. intergenerational absolute income mobility has declined substantially. Although this dramatic decline is startling, the results fit well with what has previously been established regarding rising U.S. income inequality and stagnating real median earnings. The chart below makes this clear by displaying Chetty et al.’s estimate of absolute mobility and the difference between real median income of children and parents, both around age 30, for the 1940 to 1984 birth cohorts. The two series essentially move in parallel [correlation coefficient (r) = 0.995 (2)]. A $10,000 decline in real median income of the children’s generation relative to their parents’ generation is associated with a 9.3 percentage point decline in absolute income mobility (3). Our inference from the chart is that the well-documented stagnant growth in U.S. real median household income beginning in the mid-1970s is central to the decline in absolute mobility. As the change in median income reflects changes in economic growth and inequality (the median is resilient to increased skewness in the right-tail of the distribution), it makes intuitive sense that median income growth across generations closely tracks income mobility, if the copula of parent and child income ranks stays stable. An alternative concept of mobility is to change the reference group from one’s own parents to the median parent in the parent’s generation. Chetty et al. show that the share of children earning more than the median parent declined from 92% in the 1940 birth cohort to 45% in the 1984 cohort (4). The two mobility measures (earning more than one’s own parent versus the median parent) move almost identically across cohorts (r = 0.998). An advantage of measures using the median income of children and parents or the share of children earning more than a given quantile in the parents’ distribution is that such measures can be directly computed from standard public-use cross-sectional household survey data and do not require data that longitudinally link children to parents. Given that the children-parent cohort difference in median income accounts for 99% of the time-series variation in absolute mobility, little is apparently lost by examining trends in real median income rather than directly measuring mobility because Chetty et al. show the copula is stable for recent cohorts and not important for mobility trends for earlier cohorts. This need not have been the case and was unknown before Chetty et al.’s research. Further research could examine whether children-parent cohort differences in median income predict cross-state differences and regional trends in absolute mobility. A focus on real median income by generation (as long as the copula remains stable) could allow one to extend the Chetty et al. study to examine absolute mobility for subDocumenting decline in U.S. economic mobility