Disability impairment ratings are used in 19 states to determine the benefits for severely injured workers; hence, these impairment ratings are implicitly a prospective assessment of wage loss. We compare 810 assessments made in 1989 for low-back injuries against the actual subsequent wage loss (through 1993), and find that the impairment ratings are statistically insignificant predictors of lost wages. Because low-back permanent partial claims form the single largest class of indemnity claim in workers' compensation in terms of cost, this research throws serious doubt on the validity of pure impairment rating schemes now used in many states' workers' compensation systems.