However, the biological parents are probably selecting the child with the best chance to succeed
in the host household. The decision of which child the biological parents foster may be based on
factors that are unobservable to the researcher but which clearly influence how well the child does
in the host household. To control for these factors, I estimate a child fixed effects regression that
measures the impact of fostering on that child’s educational enrollment, conditional on the child’s
unobserved attributes. The results suggest foster children after leaving their parents are neither
worse nor better off relative to their host and biological siblings. Conditional on the child’s type
(via the child fixed effects regression), there is no school enrollment impact following the fostering,
as opposed to a positive enrollment impact when the biological parents have knowledge about these
unobservable factors and make an optimization decision about which child to send out.
The data allow me to compare these three groups of children (host siblings, biological siblings,
and foster children) with children who live in households that never fostered a child. In both
household and child fixed effects specifications, foster children are better off after the fostering
compared to children from non-fostering households, and the impact is larger for young children.
In addition, in the child fixed effects regressions, the host and biological siblings are better off after
the fostering compared to the non-fostering household’s children. The results provide evidence that
the institution of child fostering and the ability of a household to send out a child when it needs to
can lead to a Pareto improvement in school enrollment for all children involved: the host siblings in
the receiving family, the biological siblings remaining behind in the sending family, and the foster
child. This Pareto improvement is the major finding of this paper, and it appears to stem from the
ability of African households to ease the constraint of a purely biological notion of a household.4
theoretical work by Serra (2003) argues that demand for child labor by the host family and aspirations for human
capital investment by the biological family could simultaneously explain a given fostering exchange.
4There is a growing literature trying to measure the impact of orphanage on children’s school enrollment
(Ainsworth, Beegle and Koda, 2002; Case, Paxson, Ableidinger, 2004; Evans and Miguel, 2004; Gertler, Levine,
and Ames, 2004; Yamano and Jayne, 2004), and while parent death is one of several reasons why children are
fostered, the data used in this paper contain only 23 children who were fostered for that reason. Therefore, the