Chetty and Hendren 2018 — The Effects of Neighborhoods on Intergenerational Mobility

Summary

Using de-identified tax records covering more than $7$ million families who moved across commuting zones and counties in the U.S., Chetty and Hendren show that neighborhoods have significant causal effects on children's long-term economic outcomes through childhood exposure. The key design insight is that siblings who move to the same new neighborhood at different ages receive different amounts of exposure — and their adult outcomes diverge in proportion to the exposure gap at a rate of approximately $4\%$ per year. This reconciles the long-standing tension between observational studies (which found large neighborhood effects) and experimental studies such as Moving to Opportunity (MTO) (which found small effects on adults and older youth).

Key Claims

• $4\%$ per year childhood exposure effect: Each additional year a child spends in a commuting zone where permanent residents' income rank is $1$ percentile higher increases the child's own income rank by $\approx 0.04$ percentiles. The effect is linear in age.
• $80\%$ lifetime pickup: Extrapolating over 20 years of childhood (ages 0–20), a child who moves at birth to a neighborhood where permanent residents score $1$ percentile higher will capture $\approx 80\%$ of that gap themselves.
• No critical age: Exposure effects accumulate linearly throughout childhood; moving at age 8 instead of 9 has the same impact as moving at age 15 instead of 16. There is no sharp critical period.
• Effects cease in early adulthood: Convergence stops once children enter their twenties, consistent with childhood — not labor-market conditions — as the primary mechanism.
• Multiple outcomes: Similar $4\%$/year exposure effects hold for college attendance rates, marriage, and teen birth, not only income.
• Causal identification from three approaches:
  1. Sibling within-family comparisons: Siblings who move at different ages show outcomes that diverge in proportion to the age gap × destination quality, holding family-fixed factors constant.
  2. Displaced movers: Restricting to households displaced by aggregate shocks (natural disasters, local plant closures) yields similar estimates, ruling out selection on move timing.
  3. Cohort, gender, and quantile placebo tests: Movers converge to permanent residents' outcomes for their own cohort, gender, and income quantile — but not to those of adjacent cohorts, opposite genders, or other quantiles. This granular pattern is precisely what causal exposure would predict and is extremely unlikely to arise from omitted variable bias.
• Reconciles MTO null: The Moving to Opportunity experiment found small effects on adult economic outcomes because most participating children were already older at the time of the move (average $\approx$ age 8). With only $\approx 12$ years of exposure remaining, they could capture only $\approx 48\%$ of the gap rather than the full $\approx 80\%$.
• County-level causal estimates (Part II): The companion paper uses the same design to estimate causal place effects for each of $\approx 2{,}500$ U.S. counties, confirming that permanent residents' outcomes are predictive of causal effects on average and enabling mapping of opportunity at fine geographic resolution.

Concepts Introduced or Extended

• Intergenerational Income Mobility
• Geographic Variation in Disability Insurance

Entities Mentioned

• Raj Chetty
• Nathaniel Hendren

Quotes

"Neighborhoods affect children's long-term outcomes through childhood exposure effects: every extra year a child spends growing up in an area where permanent residents' incomes are higher increases his or her income."

"These results motivate place-focused approaches to improving economic mobility, such as making investments to improve outcomes in areas that currently have low levels of mobility or helping families move to higher opportunity areas."

My Take

The central methodological contribution is using move-timing variation among siblings to separate causal neighborhood effects from family selection — a design far more convincing than experimental voucher studies in which families self-select into the program. The placebo tests (cohort-specific and gender-specific convergence) are especially compelling: it would be extraordinary for selection bias to replicate the gender-specific and cohort-specific permanent-resident distributions exactly. The $\approx 50\%$ causality figure cited in Geographic Variation in Disability Insurance (Friedman et al. 2016, 2018) is consistent with the movers-design estimates here, which suggest that a substantial fraction of geographic variation in Disability Insurance (DI) rates reflects true causal place effects rather than population composition. The "no critical age" finding has clear policy implications: interventions that move families to better neighborhoods — or improve neighborhoods in place — are beneficial even in adolescence, not only in early childhood.