from zero. One potential explanation is that, after the reform, households may switch to better-
quality health care, leaving the total out-of-pocket expenditure unchanged.28 Taking the eects
at both the extensive and intensive margin, the Tobit regression estimates in column (3) imply a
signicant reduction in health expenditure after the reform for treating the illness.
Three conclusions emerge from the analysis thus far. First, prior to the reform, the estimated
impact of health shocks on total household income and food consumption is small and insigni-
cantly dierent from zero. Second, prior to the reform, the eects of a negative health shock were
concentrated on the earnings of household members directly aected by the shock and on out-of-
pocket medical expenditures made to treat the illness. Taken together with the implications on
total household income, this suggests compensatory labor supply behavior against the health shock
among other members of the household not experiencing the health shock. Third, after the reform,
there is some (weak) evidence that health insurance reduces the productivity loss among members
suered from the health shock. The reform also lessens the nancial burden on the household by
reducing out-of-pocket medical expenditure associated with treating the illness.
Table 8 reports the coecient estimate on the reform indicator. In column (1) of Table 8,
I report the estimated coecient b1 where the dependent variable is an indicator for having paid
positive out-of-pocket medical expenditure. Coecient b1 measures whether the likelihood of paying
for medical expenditure out-of-pocket for treating the illness changes after the reform. As shown,
the reform reduces the incidence of positive health expenditure by about 5 percentage points. The
reform does not appear to have any impact on total health expenditure at the intensive margin (i.e.,
log out-of-pocket expenditure). The estimated coecient in column (2) is insignicantly dierent