Summary
Uses the universe of Internal Revenue Service (IRS) administrative tax records (1996–2014) to link 38.4 million young adults (born 1980–1990) to their parents, documenting three facts about the long-run drivers of disability insurance (DI) receipt. DI hazard rates at ages 24–34 are 4.8× higher for children from the poorest 1% of families than the richest 1%, variation across places is almost entirely concentrated among poor children, and roughly 50% of place differences are causal. Paradoxically, the commuting zones (CZs) with the highest DI rates for poor children tend to be "good" places by conventional metrics.
Key Claims
- Intergenerational income gradient: At ages 24–34, the net DI hazard rate is 20.1 per 10,000 for children from the bottom parental income percentile vs. 4.2 for those at the top (4.8× ratio). Each 10-percentile increase in family income predicts a 0.014 percentage-point (pp) drop in the net DI hazard rate. The gradient is stable across ages 24–32 and holds even after conditioning on parents not receiving DI themselves.
- Geographic variation concentrated at the bottom: Rich children's DI rates are near-identical across places; poor children's rates vary enormously. Pennsylvania vs. California: poor children have net hazard rates of 0.24% vs. 0.12%. Springfield, MA and Manchester, NH are the highest-DI CZs (Pred25 ≈ 3.97% and 3.85%); Brownsville and El Paso, TX, and Los Angeles, CA are among the lowest (<0.88%).
- ~50% causal (Chetty-Hendren movers design): Children who move from high-DI to low-DI CZs have DI rates that converge toward the destination, proportional to years of childhood exposure (slope: −0.032 per year of exposure before age 22; slope near zero after). Family fixed-effects (FE) estimate (slope 0.024) implies ≈ 23 × 0.024 = 0.56 — roughly half of cross-place variation is causal.
- Paradox of "good" places: High-DI CZs (for poor children) have lower income inequality (Gini, top 1% income share), lower income segregation, better schools (lower student-teacher ratios, higher test scores), and higher social capital. This is the opposite of what the labor-market-distress narrative predicts, and contrasts with findings for older DI cohorts.
- Manufacturing and trade exposure correlate positively with high-DI CZs; immigration (fraction foreign-born) is the single strongest negative correlate (partly mechanical via Quarters of Coverage (QC) eligibility).
Concepts Introduced or Extended
Entities Mentioned
Quotes
"The clear fact that emerges from this analysis is that CZs that generally appear 'better' — for instance, with better schools, higher social capital, and lower income inequality — have higher DI rates."
"Extrapolating to earlier ages, the coefficient of 0.024 in Column 3 suggests that the entire causal effect of childhood exposure on DI rates is 23×0.024 = 0.561 (SE=0.108), which implies that roughly half of the differences in DI rates between places are causal."
My Take
A compact Social Security Administration (SSA) Disability Research Center (DRC) conference paper that establishes three striking empirical facts. The intergenerational gradient and geographic concentration findings are solid. The 50% causal estimate is more preliminary — the movers design rests on the assumption that conditional on belonging to a moving family, determinants of child outcomes are uncorrelated with age-at-move, which the family FE check supports but doesn't fully verify. The "good places, bad DI" paradox is the most interesting finding and motivates the 2018 follow-up, which unpacks sorting vs. causal mechanisms. The paper is superseded in most respects by Friedman et al. (2018), which extends the analysis considerably.
Sources