This work presents the results of a study carried out to point up the correct interpretation of the variances
ratio in the stochastic frontier framework. We have evaluated the error which arises when the varance of the
underlying normal variable is taken instead of that which corresponds to the truncated normal in the computation
of the Gamma parameter. We have defined the real parameters on which decisions must be taken and have
demonstrated that the above mentioned error prodeces an overevaluation of more than 750%, of the inefficiency
presence. We also concluded that the evaluated error depens on the m /s ratio and on the inefficiency variable
incidence in the composed error of the model.