The problem of model estimation when the data are collinear was examined. Though the ridge regression (RR) outperforms ordinary least squares (OLS) regression in the presence of acute multicollinearity, it is not a problem free technique for reducing the variance of the estimates. It is a stochastic procedure when it should be nonstochastic and it does not satisfy the boundary condition. It is argued that the dilemmas of the RR approach stem from use of the minimum mean square error criterion. An alternative, called the variance normalization criterion is proposed. In theory this method overcomes the dilemmas of RR while at the same time preserving all the advantages of RR over OLS. On the basis of nine performance indices, it is shown that when applied to test data normalization ridge regression performs satisfactorily. Monte Carlo experiments are necessary to shed more light on the problem of selecting the most satisfactory optimal k-value for coping with the multicollinearity problem. (Author/DWH)