Mining injury surveillance data were used as the basis for assessing the severity of injuries among operator and contractor employees in the underground and surface mining of various minerals. Injury rates during 1983-2002 derived from the Mine Safety and Health Administration (MSHA) database were analyzed using the negative binomial regression model. The logarithmic mean injury rate was expressed as a linear function of seven indicator variables representing noncoal contractors, metal operators, nonmetal operators, stone operators, sand and gravel operators, coal contractors, and work location, and a continuous variable, RelYear, representing the relative year starting with 1983 as the base year. Based on the model, the mean injury rate declined at a 1.69% annual rate, and the mean injury rate for work on the surface was 52.53% lower than the rate for work underground. Compared to the coal operator mean injury rate, the noncoal contractor rate was 30.34% lower, the metal operator rate was 27.18% lower, the nonmetal operator rate was 37.51% lower, the stone operator rate was 23.44% lower, the sand and gravel operator rate was 16.45% lower, and the coal contractor rate was 1.41% lower. Fatality rates during the same 20-year period were analyzed similarly using the Poisson regression model. Based on this model, the mean fatality rate declined at a 3.17% annual rate, and the rate for work on the surface was 64.3% lower than the rate for work underground. Compared to the coal operator mean fatality rate, the noncoal contractor rate was 234.81% higher, the metal operator rate was 5.79% lower, the nonmetal operator rate was 47.36% lower, the stone operator rate was 8.29% higher, the sand and gravel operator rate was 60.32% higher, and the coal contractor rate was 129.54% higher.