with previous research that shows households use fostering as a transitory risk-coping strategy in
response to negative, exogenous income shocks (Akresh, 2004). It is possible that for children with
longer fostering durations, the biological parents never recovered from the transitory shock and this
explains the child’s lower current wealth levels.
Despite controlling for observables that might influence current wealth, foster status as a child
could still be endogenous with unobservable factors correlated with fostering status and wealth
biasing the regression estimates. For example, certain households might have better quality social
networks and be more likely to foster a child and that child could have higher current wealth not
because of being fostered as a child, but because of the parent’s better social network. To address
this endogeneity, I use information about the childhood fostering status of each of the respondent’s
biological siblings. In addition, the respondents provided information about the education, occu-
pation, and location for each of his siblings. I can therefore estimate a household fixed effects
regression which compares the welfare outcomes for siblings who were fostered as children with the
welfare outcomes of siblings from the same family who were not fostered as children. The evidence
in Table 10 indicates that those siblings who were fostered as children are 9.9 percent more likely
to have attended school, are 16.6 percent more likely to have a “good” occupation with higher
earnings such as a businessman, government employee, teacher, or manual laborer, are 10.7 percent
less likely to be farmers, and are 10.0 percent less likely to live in a rural village.
The data do not contain information about current assets or income for each of the siblings, so
it is not possible to replicate the OLS regressions from Table 9 using the household fixed effects
estimation strategy. To compare the household fixed effects and OLS results using the same depen-
dent variables, in columns 1, 3, 5, and 7, I present OLS estimates measuring the impact of being
fostered as a child on education, having a “good” job, being a farmer, and living in a rural area.
The OLS point estimates are similar in sign and significance but are larger in magnitude.
22