A longitudinal establishment data set is used to assess the effect of changes in workers' compensation benefits on the incidence of lost-workday injury and illness cases in manufacturing for the years 1979-84. Higher benefits are found generally to increase lost-workday cases. However, consistent with theory, the benefit effect is smaller in larger, more highly experience-rated establishments. After initial estimates are obtained using ordinary and weighted least squares, several count data models are explored that are more appropriate for the integer injury and illness counts in the data. The results are consistent across the specifications.

In the 1990s, many states passed workers' compensation laws to control cost growth. Using a difference-in-differences approach, we determine the impact of these laws on the frequency of reported workplace injuries. In response to restrictions that make it more difficult to file claims, reported days-away-from-work injuries decline, accounting for between 7.0% and 9.4% of the dramatic fall in their frequency in 1991–1997. At the same time, these filing disincentives appear to account for 6.8% of the increase in cases with only restricted work activity, although the evidence is weaker for these injuries. Restricting workers' choice of medical care provider did not appear to reduce the frequency of cases in any nonfatal injury category.