Summary
Using 40 million tax records covering children born 1980–1991, the paper measures intergenerational income mobility across 741 U.S. commuting zones (CZs) and finds enormous geographic variation. The preferred measure — the rank-rank slope — is 0.341 nationally, but the traditional intergenerational income elasticity (IGE) ranges 0.26–0.70 across specifications and is rejected as a robust summary statistic. Absolute upward mobility varies from 35.8 to 46.2 (in percentile-rank units) among the 50 largest CZs. The five strongest correlates of mobility are family structure, social capital, school quality, income inequality (at the bottom of the distribution), and residential segregation — with fraction of single parents as the single strongest predictor (r=−0.76). Causality is explicitly not claimed.
Key Claims
- Rank-rank slope vs. IGE: The rank-rank slope (regression of child income rank on parent income rank) is 0.341 nationally, stable across specifications and reasonably stable within CZs. The IGE (log-log) ranges from 0.26 to 0.70 depending on how zero-income households are treated — making it unreliable for cross-area comparisons. Rank-rank is the recommended measure.
- Cross-national benchmark: P(Q5∣Q1) — the probability a child born to parents in the bottom quintile reaches the top quintile — is 7.5% in the U.S., compared with 11.7% in Denmark and 13.4% in Canada. But this national average conceals vast internal variation.
- Absolute upward mobility (rˉ25): Mean child income rank when parents are at the 25th percentile. Among the 50 largest CZs, this ranges from 35.8 (Charlotte, NC) to 46.2 (Salt Lake City, UT). Cities below 40: Charlotte, Columbus, Cincinnati, Atlanta, Raleigh, Indianapolis. Cities above 44: Salt Lake City, San Jose, Washington DC, Seattle.
- Relative mobility (rank-rank slope within CZ): Lowest in the Southeast (Alabama, Georgia), highest in the Mountain West and rural upper Midwest.
- Geographic pattern: The lowest-mobility corridor runs through the Southeast and Appalachian/Ohio Valley belt. The highest-mobility regions are the Great Plains, Mountain West, and rural upper Midwest. The pattern largely survives controlling for race composition, poverty, and urbanization.
- Correlate 1 — Family structure: Fraction of single parents in the CZ is the strongest correlate (r=−0.76 with absolute upward mobility). Critically, this is a community-level effect: children of married parents living in high-single-parent CZs have lower expected mobility than children of married parents in low-single-parent CZs. Family structure captures something about local social norms and peer environments beyond individual household composition.
- Correlate 2 — Social capital: Bowls-Putnam index of social cohesion correlates at 0.641. Volunteering rates, religious participation, and civic engagement are components.
- Correlate 3 — School quality: Output-based school quality measures (test score level and gain measures, high school dropout rates) correlate around 0.60 with mobility.
- Correlate 4 — Income inequality (bottom-99 Gini): The Gini coefficient of income excluding the top 1% correlates at −0.634 with absolute mobility. The top 1% income share is nearly uncorrelated (r=−0.19). It is inequality among the broad population, not the concentration at the very top, that predicts lower mobility.
- Correlate 5 — Residential segregation: Measured via commute-time dissimilarity index (proxy for income segregation): r=−0.605. Income segregation physically distances low-income children from high-quality labor market and school networks.
- Race as a community-level factor: CZs with a higher fraction of Black residents have lower mobility for all races, including whites in the same CZ. The Black–white mobility gap is partially mediated by the social environment of high-Black-share areas rather than being solely an individual-level racial penalty.
- Causality caveat: The paper explicitly states these correlates are associations, not causal estimates. Family structure may itself be an outcome of other factors (economic conditions, manufacturing decline, religious composition) rather than an independent cause of mobility.
- Data: ≈40 million children born 1980–1991, matched to parent tax returns. Income measured in 2011–2012 tax records (ages ≈28–32 for the youngest cohort). CZs from Tolbert and Sizer (1996); introduced to economics by Dorn (2009).
Concepts Introduced or Extended
Entities Mentioned
Quotes
"The U.S. is better described as a collection of societies, some of which are 'lands of opportunity' with high rates of mobility across generations, and others in which few children escape poverty."
"The fraction of children living in single-parent households in an area has a correlation of −0.76 with upward income mobility. This is the strongest correlate of upward income mobility among all the variables we consider."
"We find that children of parents who live in areas with a high fraction of single-parent households have lower rates of upward mobility even if they themselves are not from single-parent households."
My Take
This paper is the geographic companion to Chetty et al. (2016): where that paper documents the income-mortality gradient nationally, this one documents the income-mobility gradient geographically. The geographic variation finding is the paper's most enduring contribution — the difference between Charlotte and Salt Lake City in absolute upward mobility is enormous by any metric. The five correlates are suggestive but deliberately non-causal, which the paper handles honestly. The family-structure finding is the most provocative: a community-level effect that survives individual controls. The methodological contribution (rank-rank over IGE) matters for anyone trying to compare mobility across areas or countries. The paper's weakness is that its 2011–2012 outcome measurement catches children at age 28–32 — relatively young for lifetime earnings stabilization — which the authors acknowledge and partially address with robustness checks showing stability from age 30.