Brand and Davis 2011 — The Impact of College Education on Fertility

Summary

Brand and Davis extend the Brand-Xie (2010) propensity-score hierarchical linear modeling (HLM) framework from earnings returns to fertility effects, asking whether women who are least likely to attend college also experience the largest fertility-decreasing effects of college. Using National Longitudinal Survey of Youth 1979 (NLSY79) data on $2{,}013$ women, they find evidence for negative selection in both treatments (timely college attendance by age 19 and timely college completion by age 23): the fertility-decreasing effect is large and significant for women in the lowest propensity strata and attenuates — and in the completion analysis fully reverses to a fertility-increasing effect — as propensity for college rises.

Key Claims

• Negative selection in fertility effects confirmed: Level 2 slopes are positive and significant for both treatments and both outcomes. For college attendance on number of children: Level 2 slope $= +0.10$ ($p < 0.05$), ranging from $-65\%$ (Stratum 1) to $+8\%$ (Stratum 6). For college completion on number of children: Level 2 slope $= +0.17$ ($p < 0.01$), ranging from $-48\%$ to $+42\%$.
• Full reversal for college completion: High-propensity women who complete college by age 23 have significantly more children by age 41 than comparable high-propensity non-completers — consistent with advantaged women using college as a complement to, rather than substitute for, family formation.
• Average effects mask heterogeneity: Under homogeneity assumption, college attendance reduces number of children by $12\%$ (significant) and completion has no significant average effect. Both are averages of very large effects at the bottom and small or reversed effects at the top.
• Mechanism — counterfactual deprivation again: The large effect at the bottom arises because disadvantaged non-college women have high fertility ($2.1$ children, stratum 1) while disadvantaged college-educated women have low fertility ($0.8$ children). The gap closes moving up the propensity distribution as even non-college high-propensity women have moderately low fertility ($\approx 1.6$). Employment supports the mechanism: low-propensity college women have the highest employment rates — $17$ percentage points (pp) higher than low-propensity non-college women — consistent with strong economic motivation driving both college attendance and fertility postponement for disadvantaged women.
• Attendance vs. completion: The fertility-decreasing effect is larger for attendance than for completion, suggesting attending even some college produces a consequential life-path shift for disadvantaged women. The completion effect reverses more dramatically among high-propensity women because college completion for them adds social resources enabling childbearing rather than creating a fundamental opportunity-cost shock.
• Event-history results: For low-propensity women the college effect on first birth concentrates in their early-to-mid-20s. For high-propensity women, college attendance has no effect on first birth timing; college completion increases first birth probability in the 30s.
• Normative ambiguity: Unlike earnings, lower fertility is not straightforwardly a "benefit" of college. The reduced fertility for low-propensity women may represent an escape from early disadvantaged family formation (single motherhood in young adulthood) that would otherwise characterize their life path, but also forecloses desired childbearing.

Data and Method

• Data: NLSY79, nationally representative, $N = 2{,}013$ women (restricted to those 14–17 in 1979, with at least a high school diploma by age 19 and complete covariate data). Followed to age 41 (through 2006). Excludes high-school dropouts.
• Outcomes: (1) Number of children by age 41 (Poisson regression); (2) probability of first birth (discrete-time event-history model, ages 19–41 for attendance, 23–41 for completion).
• Method: Three-step propensity-score HLM (Xie and Wu 2005; Brand and Xie 2010). Probit regression → propensity scores → balanced strata. Level 1: stratum-specific Poisson/event-history effects. Level 2: variance-weighted least squares regression of Level 1 slopes on propensity stratum rank.
• Identification: Selection-on-observables (ignorability). Pre-college covariates include race, family income, parental education, family structure, siblings, ability (Armed Services Vocational Aptitude Battery, ASVAB), college-prep track, parents' and friends' college expectations, and pre-college fertility history (key addition vs. Brand and Xie 2010).

Concepts Introduced or Extended

• Heterogeneous Treatment Effects

Entities Mentioned

• Jennie Brand
• Dwight Davis
• Yu Xie

Quotes

"Disaggregating the effects of college shows that comparatively disadvantaged women who attend or complete college have lower and delayed fertility than similar women who did not attend or complete college."

"Recent research on economic returns to college suggests that individuals who are least likely to obtain a college education benefit most from college (Brand and Xie 2010). We offer evidence for a corresponding result for fertility: women who are least likely to obtain a college education experience the largest effect of college on fertility."

My Take

A natural and well-executed extension of the Brand-Xie (2010) framework to a non-earnings outcome. The evidence for negative selection in fertility effects is consistent and robust across two treatments, two outcomes, and multiple life-course stages. The fertility context adds a useful theoretical complication: the "benefit" of a large treatment effect is normatively ambiguous here — lower fertility may represent liberation from a disadvantaged life path or deprivation of desired childbearing, depending on the woman's preferences. The paper is candid about this.

The main methodological limitation is the same as Brand and Xie (2010): selection-on-observables. Pre-college fertility is included as a covariate here (a meaningful improvement), but unobserved career and family orientation could differentially bias estimates across strata. The age restriction to timely college attendance/completion is necessary to avoid reverse causation (early births can disrupt college) but excludes the most disadvantaged women who take non-linear paths and may be where the effects are most policy-relevant.