5 estimates the regression including a gender and fostering interaction, while column 6 includes an
age and fostering interaction. Both foster girls and boys do better after the fostering relative to the
biological siblings remaining behind, but foster boys are 5.6 percent more likely to be enrolled than
foster girls, yet this difference is not significant. Relative to their biological siblings, young foster
children are 12.7 percent more likely to be enrolled after the fostering than older foster children.14
I further explore the heterogeneity in observed outcomes for foster children in Table 5 in which
I present household fixed effects regressions broken down by the reason for the fostering and the
location of the sending and receiving households. In Panel A, I compare foster children with their
host siblings, and in Panel B, I compare foster children with their biological siblings. Only the
coefficient on the term EverFosteredij * AfterFosteringjt is presented, although all regressions
also include control variables indicating age, gender, year, and if the child was ever fostered. Column
1 repeats the previous results from Table 4, columns 1 and 4.
For each foster child, the head of the biological household answered why the child was sent to
live with another family. Based on those responses, if the foster child is sent for schooling reasons,
then after being fostered the child is 36.2 and 35.3 percent more likely to be enrolled compared with
his host and biological siblings, respectively. This result contrasts with a foster child sent for child
labor reasons. These children are 6.4 and 2.5 percent less likely to be enrolled after the fostering
compared with their host and biological siblings, respectively, although the result is only significant
when compared to the host siblings. Children fostered due to a parent’s death are significantly
better off compared to the biological siblings left behind, with enrollment 16.9 percent higher after
14Appendix Table 1 presents household fixed effects regressions comparing the host and biological siblings from
a paired sending and receiving household, in which the household fixed effect is for the joint sender-receiver paired
household. Biological siblings are slightly worse off after the fostering compared with host siblings, but the result
is not statistically significant. However, in the case of children fostered due to a parent’s death (column 6), the
biological siblings are significantly worse off, with 16.0 percent lower enrollment after the fostering compared to the
host siblings. Likewise, when foster children are sent to households living in the same village, the biological siblings
are 8.4 percent less likely to be enrolled after the fostering compared with the host siblings.
16