On the positivity of the effective action in a theory of random surfaces