The effects of four predictor variables-driver age, driver gender, time of day, and average annual mileage-on crash involvement rates were estimated through the use of multivariate modelling techniques. Separate models were developed for fatal, injury, and property damage only crashes. All four predictor variables proved to be highly significant in explaining variations in observed rates. Rates predicted by the models after substituting the mean average annual mileage value for all driver age/gender groups were also calculated. These 'adjusted rates' show men to have a consistently higher risk of crash involvement per mile driven than women for all six combinations of crash severity and light condition examined. This contrasts with women's higher involvement rates in non-fatal crashes compared with men in the observed data. The results of the modelling are consistent with the idea that women's typically low average annual mileage is a factor in their observed higher non-fatal crash involvement rates.