Birth order and child cognitive outcomes: an exploration of the parental time mechanism

ABSTRACT

Higher birth order positions are associated with poorer outcomes due to smaller shares of resources received within the household. Using a sample of Panel Study of Income Dynamics-Child Development Supplement children, we investigate if the negative birth order effect we find in cognitive outcomes is due to unequal allocation of mother and father time investments. Exploiting the presence of siblings in the sample, we show that birth order differences in parental time are mostly driven by between-families variation rather than within-family variation. This finding suggests that birth order effects are unlikely to be driven by differences in quality time spent with either parent.

KEYWORDS:

• Birth order
• child outcomes
• time use

JEL CLASSIFICATION:

• D13
• J12
• J13
• J22
• J24

Acknowledgments

The authors would like to thank for comments and suggestions: Daniela Del Boca, Paul Frijters, Andrea Ichino, Cheti Nicoletti, Maria Rita Testa and the participants in the departmental seminar in the University of Bologna, the 6th Labour Markets and Demographic Change, the 25th Annual Conference of the European Society for Population Economics, the 2nd International Workshop on Applied Economics of Education, the 52nd Annual Conference of Italian Economic Association, the 36th Symposium of the Spanish Economic Association, and the 7th PhD Presentation Meeting with the Royal Economic Society.

Disclosure statement

No potential conflict of interest was reported by the authors.

ORCID

Chiara Monfardini http://orcid.org/0000-0002-0453-4785

Sarah Grace See http://orcid.org/0000-0002-4693-420X

Notes

1. Maternal employment has been found to have ambiguous effects on children outcomes (Blau and Grossberg 1992; James-Burdumy 2005), since maternal non-working time is not necessarily entirely spent with the children. For instance, employed mothers may compensate for work hours by spending more of their available time with their children and less time on other activities such as leisure (Huston and Aronson 2005).

2. Intact families are two-parent households, wherein parents and children are biologically related to each other.

3. Compared to the surveyed US population of which the PSID-CDS sample is representative, our sample of intact families headed by the biological parents of the surveyed children, is quite similar. Households in our sample are slightly more educated and have higher incomes. We also restrict to households with only two to five children under 18 years old, hence capture younger families with fewer children, with the children weighing slightly more than the average at birth.

4. The age standardisation process allows for a comparison of children's test scores by eliminating the discrepancy in the results due to different ages.

5. The data also contain a non-cognitive outcome called behavioural problem index, but we limit our analysis to the cognitive outcomes.

6. We allow the time investments from each parent to have differing impacts on the development process of the child. We do not disentangle the time that the child spends with both the father and the mother together. Such cases enter in both father time and mother time.

7. Price (2008) used the ATUS where only time use of the parents is recorded, implying their active engagement.

8. Birth order variables are time invarying. A first-born child maintains the same position, even when an additional child is born into the family.

9. Birth weight is used as a control for early input.

10. In Section 4.3, we check that our main findings are robust when we run the analysis on the subsample of families with only 2–3 offspring.

11. Note that the ‘timing’ of the independent variables follows the pattern of the model described above. Time received when the child is 5–18 years old is regressed on characteristics observed at the same age period. The same is true for the case of 0–12 years old.

12. The most comparable estimates are those obtained by Price for children aged 4–13 years to be contrasted with our estimates for children aged 0–12. The ATUS survey used by Price asks one adult respondent for each household. Hence, the sampled fathers and mothers are not married to each other. Our data-set instead reports time use by the child spent with his/her biological mother and biological father, who are living together.

13. We performed Hausman tests and found evidence of significant differences between OLS and FE specifications for mother Quality and mother Quality engaged time, wherein FE estimation is preferable.

14. We recognise that time allocation and children outcomes are simultaneously determined, as one referee had pointed out. Unfortunately, we do not have an appropriate instrument to identify jointly the effect of birth order and time on children outcomes, allowing for the endogeneity of time. We nevertheless performed joint estimations of the equations in a Seemingly Unrelated Regression approach, where the errors of the two equations are allowed to be correlated, capturing unobserved determinants of the dependent variables. Results, available upon request, are almost unchanged with respect to those reported in the paper.