A substantial number of papers have shown that as workers' compensation sta
tutory benefits increase, reported claims frequency and severity also incre
ase. While this seems to be a robust finding in the literature, it is not k
nown how these moral hazard responses affect real productivity. Moral hazar
d is usually regarded as shifting costs from group health medical to worker
s' compensation medical, or from absenteeism to workers' compensation disab
ility, with no real consequences on the output of firms. But moral hazard c
an reduce output in two ways: (1) in the misallocation of inputs due to "ba
d" cost information, or (2) in the loss of specific human capital, controll
ing for the allocation of other inputs. We employ two data sets to analyze
the impact of the second effect, reduced manufacturing output due to the lo
ss of specific human capital because of moral hazard. We find significant e
ffects: a 10 percent increase in benefits leads to an estimated decrease in
output, ranging from .3 to 4 percent, depending on specification.