Omitted variable bias of this type is always a possibility, and the only way of burying the possibility is to include all feasible omitted variables in the model and see if they have a separate impact. Data limitations always preclude this perfectionist approach.