Intrahousehold Resource Allocation: Do Parents Reduce or Reinforce Child Ability Gaps?

Many empirical studies that have shed light on this topic (e.g., Ayalwe 2005; Behrman et al. 1994; Datar et al. 2010; Del Bono et al. 2008; Griliches 1979; Li et al. 2010; Pitt et al. 1990; Rosenzweig and Schultz 1982; Rosenzweig and Wolpin 1988). To control for unobserved family characteristics, the literature is mostly based on samples of twins (e.g., Behrman et al. 1982; Li et al. 2010; Royer 2009) or siblings (e.g., Datar et al. 2010; Loughran et al. 2008; Quisumbing 1994; Rosenzweig and Wolpin 1988), and much of the focus has been on health (such as survival rates, birth weight, or weight-for-age) of children in developing countries (e.g., Pitt et al. 1990; Rosenzweig and Schultz 1982; Rosenzweig and Wolpin 1988). However, some recent research has focused on health endowments in developed countries (e.g., Datar et al. 2010; Del Bono et al. 2008; Hsin 2009; Kelly 2009; Loughran et al. 2008; Royer 2009). More generally, see Almond and Currie (2010) for a review of this research.

The proportion of people who are left-handed equals 11 % in Canada (Bryden et al. 1997), 12 % in the United States (Ruebeck et al. 2007), and 12 % in the United Kingdom (Denny and O’Sullivan 2007).

For additional information on the validity and reliability of the HOME-SF, see the 1996 Users Guide and citations therein. In addition, see Todd and Wolpin (2007: tables A1–A3), for a comparison of the average scores by race/ethnicity for the individual items of the cognitive home scale for children in different age ranges.

We have also explored the role of gender by reestimating our IV-FE model for males and females separately. The subsamples of same-sex siblings with different handedness are quite small, and so the estimated effects are less precisely estimated; nevertheless, the estimated female effect is larger than the estimated male effect, suggesting that reinforcing behavior may be more important for girls than boys. In addition, an alternative model specification that includes birth order dummy variables (e.g., first-born, second-born) provides a statistically significant IV-FE estimated cognitive ability effect that is slightly smaller (0.337) than that presented in column 5 of Table 3.

Del Bono et al. (2008) used an instrumental variable approach in their analysis of household resource allocation, but they instrumented for differences in prenatal inputs between pregnancies using as an instrument prenatal inputs during earlier pregnancies. This is a completely different approach from that used here.

We do not use all five tests to construct our main measure of cognitive ability because children did not complete all five tests in the same years. Thus, the summary five-test measure changes in composition from year to year.

To attempt to address the potential bias in our estimates that could arise from any forced right-handedness, we have also reestimated the IV-FE model using only those siblings with a stable hand preference over time. If we restrict the sample to those children who, between first and last handedness measurements have a change of less than one-sixth, which corresponds roughly to moving on one activity (e.g., brushing, throwing, writing) one point on a five-point scale (e.g., from right hand nearly all the time to right hand more than one-half the time), then the estimate is 0.3306. Therefore, eliminating from the estimation those children who switch from left to right (or vice versa) does not alter our main conclusion.