This dissertation performs a comprehensive analysis of the effect of different environmental, demographic and vehicle variables on the severity of two-vehicle crashes. The limitations associated with previous studies have been addressed by using a large crash database, properly defining the independent variables, using appropriate statistical models, and by considering the effect of factors normally unaccounted for such as crash type, impact speed, and weight or height incompatibilities between the two vehicles.The use of multinomial logit models at the individual occupant and crash levels provides the flexibility to evaluate variables that have opposing effects at different injury levels (such as airbags). Alternative formulations with interaction terms and with instrumental variables are included. An analysis of marginal probabilities and costs is also provided, which is particularly useful when discussing potential safety treatments with transportation officials, politicians and other decision makers.The findings from the different models are consistent and suggest that the type of crash has a great impact on severity. Age is the most significant demographic variable, with children and older occupants being least and most likely to be injured, respectively. Behavior also seems to be critical, as the use of seatbelts greatly decreases occupant injuries. Heavier vehicles increase the safety of its occupants but decrease the safety of occupants of the other vehicle. The effect of vehicle type is not as significant as weight, with the exception of pickups, which are both more crashworthy and more aggressive than passenger cars. Further research is needed on the effects of airbags and impaired driving, as the analyses conducted have been inconclusive.