Birth Order & Educational Achievement in African-American Families (PSID Study)

Introduction / Background

• A long-standing question is whether a child's educational success is affected by their birth order (e.g., first-born, last-born).

• Older studies in economics were limited. A key study by Hanushek (1992) used data from a program for low-income Black families in the early 1970s. Hanushek made two main points:

  1. Children born earlier seemed to have an advantage, but this was only because they were from smaller families.

  2. In very large families (more than 5 children), educational achievement first dropped and then rose with birth order, suggesting it was best to be the last-born.

• This new study re-examines these ideas using a different dataset, the Panel Study of Income Dynamics (PSID) Childbirth & Adoption History File (CAHF), and focuses only on African-American families.

• Key improvements in this study:

  • It accounts for the mother's age when she gave birth (something Hanushek didn't do).

  • It uses two statistical methods: Ordinary Least Squares (OLS) with grouped errors and a more advanced "within-family" (fixed-effects, FE) method.

Data Description (PSID–CAHF)

• The CAHF tracks all births and adoptions from 1985–2001 for main family members (heads, wives, and members aged $12 \text{–} 44$).

• "Index persons" are Black individuals who have their own and at least one parent's CAHF record.

• Restrictions for the study sample:

  • The mother must have been observed past age $44$, which suggests she has likely completed having children.

  • The person studied must be at least $22$ years old by 2001, assuming their education is finished.

  • The person must have at least one sibling.

• The final groups analyzed:

  • All families with $2$ or more children: $2,887$ individuals from $929$ families.

  • Large families ($5$ or more siblings): $1,376$ individuals from $268$ families.

• Checking for data loss: The PSID tends to keep more educated people in its data over time. The authors checked if this bias created a false "first-born" advantage by including age controls and confirmed their results were still valid.

• Key information used:

  • Outcome: Total years of education completed.

  • Factors studied: Indicators for birth order (e.g., being first-born), total number of siblings, mother's age at childbirth, education and ages of mother/father, child's gender, child's age and age squared, indicators for whether all siblings and both parents provided information.

  • For the FE method, these factors were adjusted based on the family's average.

Methodological Approach

• OLS (Ordinary Least Squares) with family-clustered errors: This is a standard way to estimate the relationship between factors. The formula is:

  $\hat{Edu}<em>{if}=\beta</em>0+\beta<em>1FirstBorn</em>{if}+\beta<em>2Sibs</em>f+\beta<em>3MomAge</em>{if}+X'<em>{if}\gamma+u</em>{if}$

• Family Fixed-Effects (FE): This method compares siblings within the same family. It effectively removes any family characteristics that don't change over time (like family wealth, parental styles, etc.). The formula is:

  $(Edu<em>{if}-\overline{Edu}</em>f)=\beta<em>1(FirstBorn</em>{if}-\overline{FirstBorn}<em>f)+\beta</em>3(MomAge<em>{if}-\overline{MomAge}</em>f)+(X<em>{if}-\overline{X}</em>f)'\gamma+\varepsilon_{if}$

• Testing the "last-born" effect in large families: To see if the effect of birth order changes, the study used different indicators:

  • D_{>3}: for birth order $4$ or more.

  • $D_{\le3}$: for birth order $3$ or less.

  • This allowed the study to check if the slope (the change in education for each additional birth order) was different for earlier vs. later children.

Key Empirical Findings

1. First-Born Advantage

• When using OLS without considering the number of siblings, being first-born was linked to about $0.21$ more years of education (Table 2A, column 1).

• When the number of siblings was added to the OLS model, this advantage became insignificant (just like Hanushek found).

• However, once the mother's age at childbirth was included, the first-born advantage became significant again, showing about $0.31$ more years of schooling for the first-born.

• Using the FE method (Table 2B):

  • Across all families, the first-born gained roughly $0.17 \text{–} 0.18$ years of education, though the significance was marginal.

  • In large families (more than 5 children), the first-born gain was about $0.34 \text{–} 0.43$ years, which was statistically significant at $5 \%$

• Gender difference: First-born girls gained more education compared to their sisters than first-born boys did compared to their brothers.

2. Role of Mother’s Age at Childbirth

• There's a negative correlation between being first-born and mother's age at childbirth; earlier births tend to be to younger mothers ($corr(FirstBorn, MomAge)\approx -0.44$).

• Mother's age is strongly linked to a child's education, with each additional year of mother's age at birth associated with about $0.04 \text{–} 0.06$ more years of child's schooling.

• Not including mother's age in the analysis makes the birth-order estimates appear worse for earlier-born children than they actually are.

• Mother's age itself might be influenced by other factors (like being a single mother, having less education, or unplanned pregnancy).

3. “Last-Born” Myth in Large Families

• Hanushek's finding that education improved for children born after the 4th child disappears once the mother's age is included.

• OLS model for large families (Table 3A):

  • Without mother's age: There was a positive trend for children born at order $4$ or later ($+0.118 \text{***}$) but also a negative overall effect (dummy $-0.888$).

  • With mother's age: The upward trend became insignificant ($+0.035$); the negative overall effect remained.

  • Conclusion: There's no educational advantage to being the last-born once the mother's age is considered.

• FE model for large families (Table 3B):

  • There was a significant negative starting point for children born at order $4$ or later (about $-0.65$ years).

  • The slopes within both groups (birth order $\le 3$ and > 3 ) were not significant, meaning education either stayed flat or declined, it didn't show a U-shaped pattern.

4. Other Covariates

• More siblings consistently predict less education (about $-0.084$ to $-0.128$ years per additional sibling).

• More schooling for the mother is linked to more schooling for the child (about $!+0.14!$ years per mother's schooling year); father's schooling has a weaker effect ($!+0.05! \dagger$).

• Boys tend to have about $0.25 \text{–} 0.40$ fewer years of schooling than girls.

• Having both parents present at childbirth is linked to more education, except in very large families where this effect lessens.

• The "all siblings report" indicator was usually not significant, suggesting that data loss (attrition) didn't significantly bias the results.

Interpretation & Mechanisms

• Resource dilution: Children born earlier might receive more parental time, attention, or money, like a "first-come-first-served" approach.

• Two-parent exposure: First-borns in African-American families, where single motherhood is more common, might spend more of their early years with both parents.

• Maternal age: Mother's age serves as a stand-in for other important factors like her education level, emotional readiness, and financial stability.

• The FE results suggest that the findings are not due to unmeasured family traits that stay the same over time.

Robustness & Limitations

• The study checked other age groups for the sample (mid-20s) and found similar results.

• Similar trends were observed for White families, but the impact of family size wasn't as varied.

• The sample size for families with actively involved fathers was small ($43 \%$ of families with more than 5 siblings), which limited the precision of findings related to fathers.

• It's hard to precisely pinpoint the causal effect of mother's age because it's difficult to find a clear "instrument" (an unrelated factor that only affects mother's age but not education directly).

• The FE method removes fixed characteristics for families, which means it doesn't show the overall magnitude of mother's age or total age effects.

Policy & Research Implications

• Differences in education based on birth order do exist, especially a disadvantage for later-born children in large Black families. This points to unequal opportunities within families.

• Targeting policies based on a child's birth order isn't practical. Instead, policies should focus on factors that lead to early motherhood (like increasing education, access to contraception, and economic opportunities).

• The trend of smaller family sizes among African Americans will likely increase overall schooling by increasing the proportion of first-born children.

• Future research should look for better ways to measure the causal effect of mother's age at first birth (e.g., using twin births or miscarriages as "natural experiments") and study long-term outcomes like earnings and caregiving roles.

Connections to Literature

• This study relates to the idea of a trade-off between the number of children and their quality (education), as discussed by Becker & Lewis and tested by Hanushek (1992).

• It connects to research on sibling rivalry and how resources are shared or "diluted" among children (Birdsall 1979; Kessler 1991; Behrman & Taubman 1986).

• It builds on studies about the effects of early motherhood (Geronimus et al. 1994; Bronars & Grogger 1994; Hofferth & Reid 2003; Lopez-Turley 2003).

• It aligns with Norwegian registry data that suggest family size affects educational outcomes partly through birth order (Black, Devereux & Salvanes 2004).

Numerical Highlights & Formulas

• Education gain for first-born vs. later-born (large families, FE method): $\Delta Edu_{\text{1st}}\approx 0.34\text{–}0.43\;\text{yrs}$

• Effect of each additional sibling (OLS method): $\partial Edu/\partial Sibs \approx -0.10\;\text{to}\;-0.12$ yrs.

• Effect of each year of maternal age at childbirth (OLS method): $\partial Edu/\partial MomAge \approx +0.04$ yrs.

• Correlation coefficient between FirstBorn and MomAge: $\rho(FirstBorn,MomAge)=-0.44$.

Tables & Sample Statistics Snapshot

• Average schooling (all families with $2$ or more children): $\overline{Edu}=12.08$ years (standard deviation $\approx 1.8$).

• Average number of siblings: $\overline{Sibs}=6.01$ in the group with $5$ or more siblings.

• $70 \%$ of the sample were from the baby-boom generation; over $80 \%$ of mothers had a high school education or less.

• $47 \%$ of the children were male; $63 \%$ were born to married mothers; $43 \%$ had both parents providing information.

Ethical / Practical Considerations

• There's a risk of creating negative stereotypes if educators or policymakers label later-born children as "at risk."

• The effects of birth order are tied to race, gender, and socioeconomic status, meaning solutions need to be carefully thought out.

• Data gaps regarding fathers (like due to incarceration or absence) highlight deeper societal issues in African-American communities.