compare the school enrollment change for the foster child with the foster child’s biological siblings
who were left behind. This comparison will give a lower bound for the fostering impact, since the
biological siblings who remained behind were also affected by the fostering by having more resources
available to spread among fewer children and may themselves be better off after the fostering. A
comparison of the biological siblings with children from non-fostering households can be used as
the baseline to measure the improvement for the biological siblings relative to households that do
not engage in fostering.
In Panel B of Table 3, I compare foster children before and after the fostering episode with
their biological siblings who stayed behind. As in Panel A, the results are imprecise because not
all available information is used, specifically children who were fostered in 1998 are excluded from
this table. After the fostering, foster children have 0.8 percent higher enrollment compared to the
biological siblings from the same household.
4 Empirical Results
4.1 Household Fixed Effects Results
In Panel A of Table 4, I estimate the household fixed effects regression from equation 1 comparing
foster children with the host siblings they live with.13 Column 1 presents the baseline specification
in which foster children are 1.5 percent more likely to be enrolled after being fostered compared to
the host siblings. The coefficient is not significant, but the standard error is smaller than in Table
3, indicating foster children are unlikely to be much worse off following the fostering, with a 90
percent confidence interval ranging from a 3.3 percent drop to a 6.3 percent increase. However, foster
13All households that fostered a child in 1998, 1999, or 2000 are included in the regressions. There are 2682
observations which consists of 640 host siblings and 316 foster children measured over 3 years minus 186 observations
that were excluded because the child is under age 5 in a given year.
14