The use of accident data to evaluate the crashworthiness of vehicles, and small cars in particular, is a subject with many problems. This paper identifies a number of statistical and engineering issues which have been identified in the course of data analysis efforts using current data files. Analysis results are presented comparing large and small car crashworthiness: the statistical models developed relating injury probability to crash conditions show no dependence on vehicle class; the effect of statistical and engineering issues on potential conclusions is discussed. Recommendations are made through which the issues identified may be addressed and resolved.