Number of siblings and educational attainment: application of son preference

ABSTRACT

This article examines the relation between the number of children in Taiwanese families and the educational attainment of those children. To identify the causal relation, our analysis operationalizes the traditional Taiwanese parental preference for male children as an instrumental variable to generate exogenous variations in the number of siblings. Ordinary least square estimates reveal that a larger number of siblings results in lower educational attainment. However, after addressing for the endogeneity of the number of siblings, our two-stage least square estimates result in doubt of the existence of a trade-off between child quantity and quality within a family.

KEYWORDS:

• Sibling
• educational attainment
• instrumental variable
• son preference

JEL CLASSIFICATION:

• I20
• J10

Disclosure statement

No potential conflict of interest was reported by the author.

Notes

¹ Data from the Directorate General of Budget, Accounting and Statistics, Executive Yuan, Taiwan.

² The ‘Social Stratification’ survey was held in 1992, 1997, 2002, 2007 and 2012; however, this study does not use the 1992 and 2012 data because data on the number of siblings are unavailable for these years.

³ We follow Angrist and Evans (1998) to use the sexes of firstborn and secondborn siblings as IVs. Furthermore, Tsay and Cyrus Chu (2005) suggest that if a Taiwanese woman bears female children in her first two births, then she is highly likely to aspire for a third child.

⁴ Table 2 reports robust SEs. The results are similar when using SEs and clustered SEs.

⁵ Full control includes gender dummy (female = 1), eldest son dummy (eldest son = 1), parents’ educational attainment, fathers’ employment status when the individual was 15, birth cohorts, ethnic dummies and regional dummies.

⁶ The 2SLS estimate is estimated for individuals whose number of siblings has been affected by the first two siblings’ sexes. Imbens and Angrist (1994) reported that the estimate can be interpreted as a local average treatment effect.