We present quasi-experimental estimates of the effect of changes in workers' compensation benefits on benefit duration and application frequency, using administrative data for California. Our design exploits two increases in temporary disability benefits occurring during the mid-1990s. We find consistent increases in the duration among injured workers whose benefits were affected by the schedule changes, and some evidence indicating that the likelihood of filing for benefits conditional on being injured is responsive to benefit levels. Finally, we evaluate whether the frequency effect on applying for indemnity benefits introduces a sample selection bias into standard quasi-experimental estimates of duration benefit elasticities.