of child labor as a way of coping with health shock: the coecient estimate of 1 is -0.006 (fourth
column), which implies that, after the reform, the net eect of a health shock on the use of child
labor is insignicantly dierent from zero (third row in the fourth column).
The evidence on human capital investment is in line with the predictions from models of child
labor in which households face liquidity constraints. For instance, in Basu and Van (1998) and
Baland and Robinson (2000), it can be optimal for households to have children, even when keeping
children in school leads to higher future wages. This is because keeping children out of the labor
force (and in school) can incur a large utility cost from foregone household consumption. As a
result, access to health insurance increases household resources following a negative health shock
and allows households to decrease the use of child labor and to increase child schooling. The eects
are amplied when attending school involves large xed costs.