log household income, and log household income excluding subsidies and transfer income. When
the health shock is dened on the household head and the spouse of the head, I also investigate
the impact on log income from other household members. Panel (a) reports the estimates when
the health shock is dened as the percentage of the past four weeks during which the head and
spouse reported severe illness. The coecient on the health shock (0) indicates that, prior to
the reform, a 10-percentage-point increase in the size of the health shock is associated with a
dierential decrease in the total income of the head and the spouse of the head of 17%. After the
reform, the negative impact on income reduces by one-half. The dierence (1) is only borderline
signicant. Therefore, without health insurance, illness facing the head and the spouse of the
head had a signicant negative impact on their productivity. There is some evidence that health
insurance dampens the productivity losses arising from a health shock experienced by the head and
the spouse of the head, supporting the hypothesis that health insurance potentially improves the
quality of treatment for illness.
The coecients on the health shock in the regressions for total household income are close to
zero and not statistically signicantly dierent from zero (second row of Panel (a)). The coecient
on the interaction of the health shock and the reform is also insignicantly dierent from zero.
Contrary to the direct productivity loss arising from the health shock, total household income
appears to be well insured against health shocks, even without access to health insurance. One
might think that transfer income and subsidies play an important role. However, the coecient
on the health shock in the regressions for total household income excluding subsidies and transfers
are also close to zero and not statistically signicantly dierent from zero (third row of Panel (a)).
This suggests that transfers and subsidies play a minimal role in mitigating the negative impact of
a severe health shock. Taken together, the impact of the shock on total income appears to work
entirely through the change in the income from other household members: the last row in Panel
(a) presents evidence that income from other household members compensates for the productivity
losses incurred by the health shock on the head and the spouse of the head. The reform appears
to reduce the extent of such compensatory behavior, although the coecient on the health shock
interacted with the reform is not very precise.
Panel (b) presents the coecient estimates for when the health shock is based on the share of
the past four weeks with severe sickness among the head, the spouse of the head, and all other
working-aged household members. Prior to the reform, a 10-percentage-point increase in the size of
the health shock was associated with a dierential decrease in the total income of the head, spouse