Explains why an econometric practice introduced by J.C. Soper cannot eliminate the problems caused by multicollinearity. The author suggests that it can be a useful technique in that it forces researchers to pay more attention to the specifications of their models. (AM)