A game theory model with incomplete and imperfect information is proposed here to understandthe decision faced by motorists, from two identifiable groups, to drive under the influenceof alcohol. In order to assess the best implementable policy, the rational decision from a trafficpolice force to engage in a group profiling policy strategy is described. We also suggest aperfect bayesian equilibrium solution, provinding conditions of existence and uniqueness. Thepredictions from this model suggest that, if there exist disparities in the driving skills for bothgroups when motorists are impaired by alcohol, traffic police officers should stop and administratea breath alcohol test to a higher proportion of motorists from the group with the largestviolation rate. Therefore, we suggest that group profiling through a statistical discriminationprocedure is feasible. However, if there is no statistical evidence to support such disparity, onlya fair policy -that is, to stop and test motorists from both groups with the same intensity- isimplementable. In this latter case, we suggest that a biased behavior in policing is explainedby prejudice or taste-based discrimination.