This paper analyzes the welfare implications of imperfectly categorizing risks in the insurance industry under conditions of asymmetric information. Firms, initially unable to distinguish the risk type of individuals, are provided with imperfect information concerning risk membership. The paper then compares various Wilson-type equilibria to determine the welfare implications of firms imperfectly categorizing risks. It is shown that only in the case where the initial equilibrium is of the Nash no-subsidy type is there a Pareto-type improvement in welfare.