these households have 10.6 members consisting of a household head, 1.5 wives, 3.6 children under
age 18, 3.2 children over age 18, and 1.3 members that might include the household head’s mother,
brothers, sisters, grandchildren, distant relatives, and individuals with no direct relationship.
The fieldwork component of the project improved on previous studies in several ways. First, I
adopted a methodology that involved locating and interviewing the sending and receiving house-
holds of each fostering exchange. For example, if a household interviewed in the initial sample
had sent a child to another family, then the receiving household was found and interviewed in the
tracking phase of the survey. Similarly, if a household interviewed in the initial sample had received
a child, then the biological parents of the child (sending household) were located and interviewed.
This is the first time that both the sending and receiving household from a given fostering ex-
change had been tracked and interviewed, and it enables a better understanding of the impact of
the fostering not only on the host siblings and the foster child (which is possible with some existing
datasets), but also on the foster child’s biological siblings who stayed behind.
Second, I asked retrospective questions covering the years 1998 to 2000 concerning the child’s
school enrollment history. This information allows me to compare enrollment before and after
the fostering exchange and to measure more accurately the impact of fostering. Most datasets
collected in Africa do not have school enrollment information covering a three year time period
and researchers must instead rely on cross-sectional comparisons using current enrollment. Third,
I collected information from the respondents about the childhood fostering status and occupational
and educational attainment of their siblings in order to measure the long-term impact of fostering
while controlling for household level unobservables.
The survey consisted of two distinct phases. The initial phase entailed interviews with 606
household heads and their 812 wives in fifteen randomly selected villages in Bazega province. In
these villages, the unit of analysis for the sampling frame was the compound, with some compounds