Autor Figlio Karbownik Roth and Wasserman 2019 — Family Disadvantage and the Gender Gap in Behavioral and Educational Outcomes

Summary

Using linked Florida administrative data on nearly $770{,}000$ children born 1992–2002, the authors document that family disadvantage (low socioeconomic status (SES)) disproportionately harms boys relative to girls on behavioral and academic outcomes, creating a larger boy-girl gap at the bottom of the SES distribution. The core identifying variation is a double-difference: the boy-girl gap is contrasted across SES levels, so any common gender-neutral effect of family disadvantage differences out. A fetal origins falsification test (birthweight SES gradient $\approx 3$ grams per standard deviation (SD) of SES — negligible) rules out prenatal biology and confirms that the mechanism is postnatal. Behavioral outcomes — suspensions and absences — account for the dominant pathway from family disadvantage to high school non-completion, while test score differences contribute negligibly.

Key Claims

• Family disadvantage disproportionately impairs boys: each SD increase in SES closes the boy-girl suspension gap by $2.2$ percentage points (pp), the absences gap by $0.075$ pp, the kindergarten readiness gap by $1.15$ pp, and the high school graduation gap by $1.5$ pp.
• The underlying SES index (principal component analysis (PCA) of mother's education, age, Medicaid birth indicator, marital status) has eigenvalue $2.26$; the Black–white SES gap is $+1.44\sigma$ (mean Black $= -1.06$, mean White $= +0.37$).
• Fetal origins falsification: boy-girl birthweight gap $\approx 3$ grams per SD SES, effectively zero, confirming that the mechanism operates postnatally, not in utero.
• Within-sibling fixed effects produce estimates close to ordinary least squares (OLS) with SES controls — the double-difference is not driven by unobserved family characteristics that systematically differ across SES levels.
• Behavioral pathway dominates: suspensions account for $90.6\%$ of the behavioral/academic mediator path from family disadvantage to high school dropout; test scores account for negligible share.
• School and neighborhood controls reduce the family SES effect on behavioral outcomes by $\leq 25\%$, implying that at least $73\%$ of the excess boy-girl gap explained by family disadvantage operates within schools and neighborhoods, not between them.
• Race heterogeneity: the protective SES effect on suspensions is smaller for Black children than for white children (three-way interaction $\beta_5 = +0.97$ pp), consistent with differential punitiveness in schools (the Oaxaca restrictions test rejects equal gradients across races).
• Connection to Chetty-Hendren: family disadvantage effects on the boy-girl gap operate independently of the school and neighborhood quality channel, suggesting family and neighborhood are distinct (though correlated) inputs.

Concepts Introduced or Extended

• Gender Gap in Educational Outcomes
• Intergenerational Income Mobility (family vs. neighborhood as inputs)
• Marriage Market (single-parent family structure as the SES disadvantage channel)

Entities Mentioned

• David Autor
• David Figlio
• Krzysztof Karbownik
• Jeffrey Roth
• Melanie Wasserman
• Raj Chetty
• Nathaniel Hendren

Quotes

"Boys raised in disadvantaged families have substantially worse behavioral, cognitive, and educational outcomes as compared with their sisters raised in the same households, and this gap is substantially larger for boys raised in lower-SES families."

"We conclude that the gender gap in disruptive behavior is primarily a postnatal phenomenon driven by social rather than biological factors."

"Family disadvantage accounts for a meaningful share of the Black-White gap in these outcomes, and the finding that Black children respond less to family SES — particularly in suspension rates — is consistent with additional barriers faced by Black males in educational settings."

My Take

The double-difference design is elegant and transparent: by differencing on gender within family SES, the paper strips out both the gender-neutral effect of SES and any level difference in boys' outcomes, leaving only the interaction. The fetal origins test is a particularly clean falsification. The main limitation is that the SES index is measured at birth, so it conflates permanent household resources with transient conditions, and the Florida sample is racially distinctive (high share of Black and Hispanic children, high poverty) — national generalizability is plausible but not guaranteed. The finding that school/neighborhood controls explain $\leq 25\%$ of the family SES effect is consistent with Chetty-Hendren (2018) isolating neighborhood effects: the two studies together suggest family and neighborhood are genuinely separable inputs, not collinear proxies for each other.