One of the primary objectives of Colombian social health insurance reform was to increase competition among for-profit insurers. Unfortunately, the flat capitated formula creates an opportunity for sickness funds to maximize reimbursement gains by "cream skimming"--selecting against unhealthy individuals. This paper explores sickness fund selection behavior to evaluate the efficiency losses associated with the introduction of managed competition in Colombia. Data from a 1997 Colombian household survey are analyzed with a bivariate probit model with partial observability using instrumental variables. The model yields some evidence of sickness fund selection based on health status. Public policy options to discourage risk selection by health status are discussed.