For many years, the annual survey of occupational injuries and illnesses by the Federal Bureau of Labor Statistics (BLS) has consistently reported, without explanation, that injury rates in the smaller establishments (< 50 employees) are substantially lower than those for midsize establishments (100-499 employees). Also during those years, a remarkable increase has been reported in the injury rate in large companies, following the imposition of stiff fines for failure to keep required injury records. The rate patterns are identical for Michigan and the country in general. We investigated possible causes for lower injury rates in small establishments since such rates are inconsistent with reports of higher fatality rates in small establishments in the mining, construction, manufacturing, and transportation industries and higher, or at least comparable, injury rates for small establishments in the mining industry. They are also inconsistent with increased turnover and decreased availability of occupational safety services in small companies. Moreover, injury severity, as measured by missed worktime, is greater for workers in small establishments. We investigated the possibility that interactions of workforce or injury characteristics with establishment size could explain the rate differences. We also reviewed the available literature to see whether differences in labor turnover rates could explain the BLS findings. Graphical and statistical analyses of the 1988 CPS Annual Demographic File, a sample of employed persons in the U.S. workforce, failed to identify any associations between workforce characteristics and enterprise size that would explain the lower rates. Similarly, graphical and statistical analyses of all Michigan workers incurring a compensable injury in 1986 failed to indicate any associations between injury characteristics and establishment size that would explain the lower rates. The potential role of labor turnover on the injury rate was analyzed from data in the literature on turnover rate by establishment size and risk of injury by time on the job. None of these analyses explains the lower injury rates reported for small establishments. This leaves underreporting of injuries from small establishments as a substantial possibility. If small establishments were subject to the same injury incidence rates as midsize establishments, then the 1986 survey for Michigan may have missed as many as 54,000 injuries (and far more nationally). We suggest that BLS undertake methodological studies to validate the completeness of reporting from small establishments.