Regression models were fit to these fatality data, both in the aggregate and for a variety of subgroups, based on race and gender. Next, two, time-series models (autoregressive integrated moving average and structural time series analysis) were fit to the data, both with and without the economic indicators, to examine the relative ability of the models to forecast subsequent fatalities. Results revealed that none of the modeling techniques indicated that knowledge of yearly values by rates of employment, unemployment, and the nonlabor force leads to improved forecasts of the level of motor vehicle fatalities, suicides, or homicides in the total United States population or within various subpopulations of interest. Tables, figures, and 14 references (Author abstract modified)