where the nominator sums over the number of days of severe illness among the remaining household
members aged between 18 and 55 years (in addition to the head and the spouse of the head).
The health shocks dened in the sample are largely transitory and mean reverting. For instance,
after controlling for county-by-year xed eects, the autocovariance structure between the current
health shock and the rst- and second-lagged health shocks are -0.66 and -0.31, respectively.21
This indicates a strong mean-reverting pattern for health shocks between survey rounds.