McLanahan Tach and Schneider 2013 — The Causal Effects of Father Absence

Summary

McLanahan, Tach, and Schneider (2013) review 47 articles that use rigorous causal designs to estimate whether father absence causally reduces child well-being — going beyond the disputed cross-sectional literature. Six identification strategies are assessed (lagged dependent variable [LDV], growth curve models [GCM], individual fixed effects [IFE], sibling fixed effects [SFE], natural experiments/instrumental variables [IV], propensity score matching [PSM]), each with distinct tradeoffs between internal and external validity. The central verdict: rigorous designs continue to find negative effects of father absence, but smaller than cross-sectional estimates. The strongest and most consistent evidence is for high school graduation, children's social-emotional adjustment (externalizing behavior, substance use), and adult mental health. Effects on test scores, earnings, and marriage are weak and inconsistent. The evidence indicates that father absence operates primarily through behavioral/noncognitive pathways rather than cognitive ability, and selection explains part — but not all — of the raw correlation.

Key Claims

• 47 studies reviewed, covering primarily US data but also UK, Canada, Germany, Sweden, Norway, Australia, Indonesia, and South Africa. Unit of analysis is each separate model, not each article.
• Six causal strategies (in rough order of stringency for time-constant unobservables):
  • LDV: Adds pre-separation well-being as control; can't address time-varying confounders or pre-birth absences
  • GCM: Estimates divergence in trajectories after disruption; most credible if pre-divorce intercepts insignificant; allows falsification tests
  • IFE: Removes all time-constant individual unobservables; but requires repeated outcomes (can't use for high school [HS] graduation or earnings); very sensitive to measurement error
  • SFE: Removes family-level fixed unobservables via siblings with different exposure; assumes no age-of-disruption heterogeneity
  • Natural experiments / IV: Parental death or divorce law changes as instruments; parental death is rarely exogenous; divorce laws violate exclusion restriction by affecting who marries and intra-household bargaining
  • PSM: Balances observables; does not address unobservable confounders
• Test scores (31 analyses): Mixed — 14/31 significant. GCMs most likely to find effects; IFE and SFE rarely do (coefficients markedly smaller, not just noisier). Overall picture: selection explains much of the association with cognitive ability.
• HS graduation (9 studies): Strong — 8/9 significant, including diverse US designs. Only null result from German blended-family SFE. College graduation: weak (1/4 significant).
• Externalizing behavior (27 analyses): 19/27 significant; LDV finds most, IFE/SFE least. More pronounced for boys than girls. Behavioral/mental disorders, not cognitive deficits, are the dominant pathway to lower HS graduation.
• Substance use (6 analyses): Very robust — 5/6 significant, even in SFE designs (which rarely find effects on other outcomes).
• Adult mental health (6 analyses): 4/6 significant; findings robust across methods including SFE and propensity score.
• Labor force: Only ~14 analyses, dominated by Gruber (2004) (US divorce laws) and Corak (2001) (Canadian parental death). Employment effects fairly consistent and negative; earnings effects inconsistent. Corak finds higher probability of receiving income assistance.
• Marriage and family formation: Marriage effects inconsistent (3 studies, 3 different answers); marital stability more consistent (Corak and Gruber both find higher separation probability for offspring); early childbearing effects (2 UK SFE studies) consistent.
• Selection is real but partial: IFE/SFE/PSM estimates are generally markedly smaller than ordinary least squares (OLS) — strong evidence of positive selection bias in cross-sectional designs. But estimates are rarely zero in rigorous designs, particularly for education and behavioral outcomes.
• Early childhood > middle childhood: Consistent finding across domains that disruptions occurring in early childhood (ages 0–5) have more negative effects than those in middle childhood.
• Boys more affected: Externalizing behavior effects more pronounced for boys; some attainment effects as well. Consistent with the Autor et al. (2019) finding on family disadvantage and the gender gap.
• Behavioral > cognitive pathway: The juxtaposition of weak test score effects and strong HS graduation effects implies effects operate through noncognitive skills (self-regulation, externalizing behavior), consistent with Cunha and Heckman (2008).

Concepts Introduced or Extended

• Causal Effects of Father Absence
• Gender Gap in Educational Outcomes
• Nonmarital Fertility
• Intergenerational Income Mobility

Entities Mentioned

• Sara McLanahan

Quotes

"Studies using more rigorous designs continue to find negative effects of father absence on offspring well-being, although the magnitude of these effects is smaller than what is found using traditional cross-sectional designs. The evidence is strongest and most consistent for outcomes such as high school graduation, children's social-emotional adjustment, and adult mental health."

"The lack of strong test score effects is also consistent with findings in the early education literature that suggest that cognitive test scores are more difficult to change than noncognitive skills and behaviors… it makes sense that we find strong evidence of effects on the likelihood of high school graduation but not on test scores."

My Take

The paper's main contribution is synthetic rather than methodological — it shows that the causal identification revolution of the 2000s did not overturn the father absence literature, just moderated it. The convergence on behavioral/noncognitive outcomes as the robust causal channel is the paper's most important substantive finding and connects directly to Heckman's noncognitive skills agenda.

The methodological hierarchy implicit in the paper (IFE/SFE > LDV/GCM > OLS) is somewhat misleading, because IFE/SFE models have different external validity — they compare siblings within blended families, who are already a non-representative group. The null IFE/SFE test score results may partly reflect this sample restriction rather than genuinely zero effects.

Connection to this wiki: the family disadvantage mechanism identified here (father absence → externalizing behavior → lower HS graduation → lower labor force attachment) is the background micro-level mechanism behind Marriage Market, Nonmarital Fertility, and Gender Gap in Educational Outcomes. The Autor et al. (2019) finding that family disadvantage disproportionately harms boys aligns precisely with the McLanahan et al. (2013) finding that externalizing behavior effects are more pronounced for boys, and that father absence effects on HS graduation are stronger in the US than Europe.