Two specific alternatives of biased regression estimators; principal component and ridge regression have been examined and applied to a set of multicollinear consumption data with the purpose of curing the problems caused by the presence of multicollinearity. Comparisons of the empirical results obtained from the two biased estimators with those derived from least squares have shown that improvements over ordinary least. squares can be achieved when the alternative techniques are utilized in situations where severe multicollinearity is suspected. Comparisons of the parameter estimates of ridge and principal component regression and the use of some prediction error criteria has also shown that the RR with k=0.0104 should be used if one has to make a choice between RR and PCR.