This study measures lost earnings by comparing the earnings of all other injured workers with a comparison group, controlling for differences that could account for different earnings. The comparison group consists of workers who lost 7-9 days from work and then returned to their jobs. We studied lost time injuries reported to workers' compensation systems in Wisconsin and Florida. The method is a linear-regression version of a difference-in-differences approach. The analyzed data are for 54,309 cases in Wisconsin and 71,189 cases in Florida. Workers' compensation administrative databases supplied injury and personal characteristics. To obtain earnings and employer characteristics, we matched the workers' compensation data with data collected by state unemployment insurance and labor market information agencies. In this study, we also conducted two ethnographic interviews of workers with back injuries, including (a) 198 workers from Wisconsin stratified by gender and age (20-29, 30-54, 55 and older), and (b) 216 workers from a Florida telephone survey stratified by gender, age, and racial and/or ethnic background Black non-Hispanic, White non-Hispanic, and Hispanic). Interviews conducted about years after injury explored the workers' beliefs and attitudes about themselves as workers including their pre and post injury relationship to employment, their experience of the work injury, encounters with the medical and legal systems, and the effects of the injury on self-image and family relationships. The interviews were conducted about six years after the workplace injury.