children are more likely to be girls.
Table 2 analyzes how children’s school enrollment changes over time before and after the fos-
tering episode. Many development organizations are concerned that after a child is sent away from
his biological parents he will stop attending school, but the data do not confirm this. Only two
percent of foster children were no longer enrolled after being sent to the host household despite
being enrolled prior to the fostering. This compares with 3.3 percent of host siblings and 2.3 per-
cent of biological siblings who discontinued enrollment after the fostering exchange. Following the
fostering, approximately the same percentage of children in each group were newly enrolled stu-
dents, with rates ranging from 4.5 to 4.9 percent. The largest difference between the three groups
is the percentage of children who were never enrolled. There are 82.7 percent of foster children and
77.9 percent of biological siblings in this category, but only 60.5 percent of host siblings were never
enrolled. I can reject the null hypothesis that the percentage of children in each transition group
(never enrolled, discontinued enrollment, newly enrolled, and enrolled both years) is the same across
host siblings, biological siblings, and foster children with a likelihood ratio y2(6) test statistic of
51.00 and a corresponding p-value of 0.00. However, the likelihood ratio y2(3) test statistic testing
for equality between foster children and biological siblings cannot be rejected with a p-value of 0.34.
3 Empirical Strategy
3.1 Empirical Identification in Previous Research
Several recent empirical papers attempt to measure the school enrollment impact of children living
away from their biological parents. Most of these papers use cross-sectional data and compare
school enrollment for children living with their biological parents with that of foster children living
without their parents. However, current school enrollment is partly a function of that child’s school