enrollment history. Without controlling for that history prior to the fostering, the researcher will
incorrectly measure the fostering impact. Foster children are more likely to come from households
that experienced negative income shocks (Akresh, 2004), and due to these shocks, it is likely the
child’s parents could not afford the school fees and the child was not enrolled prior to the fostering
(Thomas et al., 2004). As a consequence, if the host family maintains the foster child’s pre-fostering
enrollment status, in a cross-sectional comparison it will appear that the foster child is worse off
compared to non-fostered children. Using cross-sectional data to measure this impact would yield
misleading results if there are factors (such as school enrollment history, wealth, or network quality)
omitted from the regressions that are correlated with both fostering and school enrollment.
Lloyd and Blanc (1996) use Demographic and Health Surveys (DHS) from seven countries in
sub-Saharan Africa and show that children’s school outcomes (measured by current school enroll-
ment and grade four completion) are more influenced by characteristics of the child’s extended
family network and the household in which the child resides than by characteristics of his biological
parents. Zimmerman (2003), using 1993 household survey data from South Africa, finds that the
risk for foster children of not attending school is lower than it would have been if the child had
stayed with his biological parents. Both studies rely on cross-sectional data which are subject to
the aforementioned problems. Overall, some researchers have found similar results to Zimmerman
(Eloundou-Enyegue and Shapiro, 2004), while others have found that foster children are less likely
to be enrolled and more likely to be working (Kielland, 1999). Cichello (2004) extends Zimmer-
man’s analysis by incorporating information from a 1998 survey collected on a sub-sample of the
households used by Zimmerman. This allows him to construct a school progress measure defined as
the number of additional years of schooling attained between 1993 and 1998. He finds there are no
positive gains for foster children in terms of school progress, despite higher initial school enrollment
in 1993. Despite using the additional 1998 information to develop an improved measure of human