Black and Devereux 2011 — Recent Developments in Intergenerational Mobility

Summary

Comprehensive Handbook of Labor Economics chapter surveying measurement methods and empirical findings on intergenerational income mobility. The central methodological contribution is diagnosing three sources of downward bias in naive intergenerational elasticity (IGE) estimates and showing that corrected US IGE approaches $\approx 0.6$ (Mazumder 2005), nearly double the naive value. Cross-country evidence confirms a consistent Nordic < UK < US ranking. A third strand reviews causal identification via twins, adoptees, and instrumental variables (IV), finding that ordinary least squares (OLS) substantially overstates intergenerational transmission of education and that genetic endowments dominate the nurture channel for education outcomes.

Key Claims

• Three IGE biases: (1) Attenuation bias — short earnings windows conflate transitory autoregressive (AR(1)) shocks with permanent income; (2) life-cycle bias — the cross-sectional OLS of child earnings on father earnings is not the IGE unless both measured at the same career stage (Haider-Solon 2006: $\lambda_a \approx 0.2$ at age 25, peaks ${\sim}1.0$ at 30–45, falls to ${\sim}0.6$ by late 50s); (3) classical measurement error in self-reported father earnings attenuates estimates further.
• Mazumder (2005) correction: Using Social Security Administration (SSA) administrative earnings, US IGE rises from $\approx 0.25$ (2-year average) to $\approx 0.45$ (5-year), to $\approx 0.61$ (16-year average). The "true" long-run US IGE is substantially above the $\approx 0.4$ commonly cited in the 1990s literature.
• Cross-country IGE (Jäntti et al. 2006): Denmark $0.071$, Norway $0.155$, Finland $0.173$, Sweden $0.258$, UK $0.306$, US $0.517$. Consistent Nordic < UK < US ordering across multiple samples and methods.
• IGE vs. rank-rank slope: IGE is sensitive to the treatment of zero incomes and to earnings inequality differences across generations. The rank-rank slope is bounded $[0,1]$ and invariant to the marginal income distributions; it is the preferred measure for cross-area and cross-time comparison (see Intergenerational Income Mobility).
• Education transmission — causal identification hierarchy: IV (compulsory schooling laws) → adoptees → twin fixed effects → OLS. All causal estimates fall below OLS. Black, Devereux, and Salvanes (2005) Norwegian compulsory-schooling IV: OLS $0.22$, IV $\approx 0.04$ and not statistically significant.
• Sacerdote (2007) Korean-American random adoptees: Mother's education coefficient $= 0.32$ for biological children; $0.09$ for adoptees placed at random. Father's income coefficient $= 0.09$ biological, $\approx 0.00$ adoptees. Environmental nurture effect on education and earnings is far smaller than the genetic/selection channel.
• Björklund et al. (2005) structural twin decomposition: Genetic variance share $g^2 = 0.28$, shared-family-environment share $s^2 = 0.04$ for brothers' earnings. Non-shared environment accounts for $\approx 0.68$. Genetics explains $28\%$ of earnings variation; shared family environment only $4\%$.
• Other outcomes: IQ explains only $3.2\%$ of intergenerational income variance. Occupational persistence: $\approx 30\%$ father-son, $\approx 20\%$ father-daughter (rank correlation in occupation prestige).

Concepts Introduced or Extended

• Intergenerational Income Mobility

Entities Mentioned

• Sandra Black
• Paul Devereux

Quotes

"The estimates suggest that the United States has less intergenerational earnings mobility than most other developed countries."

My Take

The chapter's most important contribution is the wedge between naive IGE ($\approx 0.25$ with 2-year windows) and the corrected estimate ($\approx 0.61$): this gap is not a minor refinement but fundamentally changes the policy inference. US mobility is far more constrained than the 1990s literature suggested. The causal identification section is sobering: 20+ years of clever research designs converge on genetics dominating nurture for education transmission, yet the mechanism remains opaque — we know that environment matters little on average, but not why. The chapter gives short shrift to the rank-rank slope's advantages, which Chetty et al. (2014) later developed into the dominant empirical framework. The Mazumder correction also implies that cross-country comparisons using inconsistently averaged earnings are unreliable even within the IGE framework.